Az előző adásban a vezetők oldaláról beszéltünk: mi történik, amikor egy cégvezető bevezeti az AI-t. Most Back Réka átül a másik oldalra, és azt kérdezi Fazakas Lászlótól, mit jelent mindez annak, aki a cégben dolgozik. Annak, akinek a fizetése attól is függ, mit gondol róla a vezető, ha megtudja, ő mit gondol a cégről.

Először a bizalom. László szerint ilyenkor a munkájuk legalább 80, de lehet, hogy 90 százaléka az, hogy „megbízhatóak, szavahihetőek, őszinték legyünk mind a két fél felé”. Az első ijedelem mindig ugyanaz: már megint idejött egy okos tanácsadó, aki megváltja a világot, és még több munkát ad. Amikor a munkatársak látják, hogy mindkét oldalt segíteni szeretnék, onnantól merik elmondani, mire használnák
az AI-t. László szerint ez a személyes kapcsolaton múlik: azon, hogy ott vannak velük, és meghallgatják őket. A cégeknél ritkán kérdezik meg a munkatársat, hogyan dolgozik, mit változtatna meg, ha tehetné, és hol várna segítséget.

Kiszedhető a gépből? Réka azt kérdezi: ha valaki elég rafinált, nem tudja-e úgy kérdezgetni a bevezetett rendszert, hogy kiderüljön, ki mit mondott? László először megáll, és azt mondja, el kell gondolkodnia a válaszon. Olyan még nem fordult elő, hogy egy vezető a rendszerrel ki akart volna nyerni valamit a kollégáiról. „Még nem voltam ilyen helyzetben, de amennyire ismerem a csapatomat, ebből próbálnánk kilépni.” Visszajelzést viszont adnak a vezetésnek, ha a beszélgetésekből kirajzolódik, mi fáj a cégnek.

Holnap nem kell bejönni? Réka szerint a legtöbb munkatárs fejében ott motoszkál a kérdés: hol lesz az a pont, ahol az AI ügyesebb, gyorsabb és okosabb lesz, és a vezető azt mondja, köszönjük, holnap nem kell bejönni. László szerint ezt a félelmet a hírek táplálják arról, melyik cég hány embert küldött el az AI miatt. Az ő példája az IKEA, ahol a vezetés megértette, hogy az ügyfélszolgálatosok sokéves tapasztalatát kell megtámogatni egy eszközzel, hogy jobban tudjanak segíteni a vásárlóknak.

Január óta tanítom. László a saját esetét hozza: „én január óta tanítom a saját példáimon”, és a rendszer minden nap elkövet valami szarvashibát, a saját döntései alapján is képes valami teljesen vadat kihozni. Hogy hol van az a pont, amikor egy AI-rendszert annyira betanítanak, hogy embert váltson ki, nem tudja megmondani. Az ő esetében ez még sok-sok év, és inkább azon gondolkodik, hogyan tudna még
embereket felvenni, akiket másfajta felkészültséggel lát el. Szerinte a nagyon fura eredményeket a rosszul használt eszközök adják, akárhány ember dolgozik mellettük.

Nyaralás üzenetek nélkül. Réka a saját vágyát mondja el: pénteken leteszi az irodakulcsot, két hétig nem zavarják, és mire visszajön, minden ment tovább, nem zúdul rá a munka. László szerint ez. egészségesebb nézőpont, mint amit mostanában a legtöbb vezetőtől vagy munkavállalótól lát: hogyan lehet ezeket az eszközöket a saját életminőségünk javítására használni.

Kit nem tudunk majd felvenni? László szerint Európában nem szeretünk a népesség változásáról beszélni. Az életkorra gondol, és arra, hogyan változik, ki milyen szakmát választ; a migráció most nem téma. Júliusban egy konferencián egy frissen alapított amerikai startup vezetőjével beszélgetett, akik azért jönnek Európába, mert ők onnan látják, hogy itt öt éven belül bajba kerül az alapszolgáltatások működtetése, pusztán az életkor miatt. Ez az ő előrejelzésük. László ebből azt veszi át, hogy a cégeknek el kell kezdeniük megtervezni, hány embert nem tudnak majd felvenni, mert nincs megfelelő képzettségű ember, és hogyan töltik be ezt a hiányt pár éven belül.

A mért adat (Eurostat): 2025. január 1-jén az EU lakosságának 22,0%-a volt 65 éves vagy idősebb, és száz munkaképes korúra (15–64 év) 34,5 idős ember jutott; 2015-ben még 29,0.
https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Population_structure_and_ageing

Kiszámíthatóság. Mit várnak a munkatársak az AI-tól? László szerint leginkább kiszámíthatóságot. Egy e-mail-láncba külső és belső emberek is bekapcsolódnak, valaki visszajön a szabadságról, és tizenöt-húsz levelet kellene végigolvasnia. Mennyivel jobb lenne, ha egy rendszer szétválogatná, ki mit mondott, hol vannak az ellentmondások, és onnan kapcsolódhatna be. László szerint a saját cégüknél is minden nap előfordul ilyen, akár ő maga is képes ilyet produkálni.

Érzelemmentesen. Réka szerint az ilyen történetek mindig nagy személyes érintettséggel érkeznek, és sok munka lefejteni, mi a lényegük. László szerint a legtöbb vezetőnek az a célja, hogy megoldja a helyzetet, és nem akarja megalázni a kollégáit. Ebben sokat segít, hogy egy jól beállított AI-rendszer érzelemmentesen nézi a helyzetet. És egy trükk a munkatársaknak, amiről László azt látja, hogy működik: ha egy cégnél jól bevezették a rendszert, és valaki azt mondja, „de még a ChatGPT is azt mondta, amit én mondok”, azt a vezetők sokszor el tudják fogadni.

Előbb egy használható eredmény. Mit szól a vezető, amikor tükröt kap arról, mit gondolnak a munkatársai? László szerint itt is egy trükköt alkalmaznak: sokkal hamarabb adnak egy használható eredményt, például egy új marketingtervet arról, hogyan kommunikáljanak a következő két hónapban. Ha a vezető látja, mit hozott ki a rendszer abból a munkából, amit ő, a munkatársai és az Arcanian beletettek, sokkal könnyebb arról beszélni, mi kell ahhoz, hogy több ilyet kapjon.

Tiltás helyett magyarázat. Réka szerint az AI-szabályzatok körül nagy a káosz: van, ahol tiltják, aztán titokban mégis használják, máshol hosszan sorolják, mit szabad és mit nem. László szerint a valóság az, hogy a munkatársak valahogy már használják, és ilyenkor a legrosszabb, ha jogi vagy informatikai alapon azonnal szigorítást, tiltást vezetnek be. Hasonló kérdéseket a Google-nek és a Metának sem tettünk fel, pedig az ő rendszereik döntenek arról, mit látunk, amikor hirdetünk, keresünk vagy böngészünk. Ha a nagy rendszerektől nem kérjük számon, a munkatársaktól sem kellene az első pillanatban. Előbb el kell magyarázni, mit tudnak ezek a rendszerek, minek mi a következménye, és az hogyan hat a munkahelyre. Szabályzatról csak ezután érdemes beszélni.

Tudok bajt csinálni? Réka azt kérdezi, egy munkatárs egy több ezer fős cégnél tud-e bajt csinálni azzal, hogy feltölt pár dokumentumot összegezni. László röviden: igen. Az ő példája az OpenAI megosztott beszélgetései, amelyeket a Google leindexelt. (Pontosítás: 2025 nyarán ez azokat a beszélgetéseket érintette, amelyekhez a felhasználó megosztási linket készített, és bekapcsolta, hogy a beszélgetés kereshető legyen; az OpenAI ezután megszüntette ezt a lehetőséget.) „Tudunk bajt csinálni azzal is, ha autót vezetünk”, teszi hozzá László. Ha viszont félelemből kezdünk bevezetni egy új technológiát, az nem fog működni.

Nem a technológiát kérdőjelezzük meg. László ebben egyetért az amerikai gondolkodással, ahol a kultúra része, hogy az új dolgok veszélyesek, „de nem a technológiát kérdőjelezzük meg, hogy van-e rá szükség, hanem azt kérdőjelezzük meg, hogy jól használtuk-e.” A repülés világát tartja követendőnek: ott a nagy katasztrófák után sem embereket vagy cégeket hurcolnak meg: az egész szakma tanul belőle, hogy legközelebb senki ne kövesse el ugyanazt a hibát. Ezt a gondolkodást javasolná a munkatársaknak és a cégeknek is.

Meg kell tanulni, mint az írást? Réka szerint az AI tudatos használatát lassan mindenkinek meg kell tanulnia, mint az írást és az olvasást. László szerint igen, és a szabályozás is erre megy: erről kevesen tudnak, de az uniós szabályok a vállalkozásokra is kötelezettséget rónak. *(Pontosítás: az AI Act 4. cikke

2025. február 2. óta kivétel nélkül minden AI-rendszert alkalmazó cégre vonatkozik. A 2026 júliusában hatályba lépett módosítás után is kötelező intézkedéseket tenni a munkatársak AI-jártasságáért, de konkrét szintet már nem ír elő. A képzések dokumentálását a cikk nem írja elő; az Európai Bizottság szerint elég egy belső nyilvántartás, tanúsítvány nem kell.)*
      https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers

A sütibanner tanulsága. Kell-e minden AI-döntésről tudnia egy felhasználónak? László szerint a döntések fajtájáról és a következményükről kell tudni. A példája a sütibanner: amióta az EU bevezette, kutatócégek mérték, hány órát töltünk kattintgatással, és a végén úgyis mindent elfogadunk. Mert pici a tét. Az AI-rendszereknél a tét nagyobb, ezért tudatosabbnak kellene lennünk: amit rosszul adtunk meg nekik, azt utána nagyon nehéz kiverni a fejükből. „Lehet, de nem könnyű.”

Egy új kolléga, aki figyel. Réka szerint az AI-asszisztens olyan, mint egy új kolléga, aki minden mozdulatunkat lesi, hogy segíthessen. Jól beállítva valódi segítség lesz, rosszul beállítva lelkesen bólint rá valami csacsiságra. Mennyire kell követni, mi alapján dönt? László szerint nagyon, de ez sem technológiai kérdés. Ahogy az autónkat is ismerjük valamennyire: van, aki a legkisebb zajból tudja, hogy szerelőhöz kell mennie, és van, aki csak akkor hív szerelőt, amikor az autó már megállt. Az AI-rendszerek sokkal kisebb rezdülésekre is reagálnak, ezért sokkal érzékenyebben kell velük bánni.

Jól kell tudni kérdezni. László szerint ebben mi, magyarok gyengék vagyunk: sokan úgy nőttünk fel, hogy az iskolában nem tanultuk meg, hogyan dolgozzunk együtt, hogyan tegyünk fel kérdéseket, és hogyan fogalmazzunk meg kritikát. Tudnak-e kérdezni az emberek az AI-tól? László szerint sajnos nem: egy-két váltásból várják, hogy a gép a ki nem mondott kérdéseikre is jó választ adjon.

Kérdezz valamit! Réka erre elmondja a saját történetét. Beadott a gépnek egy halom adatot, mire az visszaírt: nagyon jó, és most mit akarunk ezzel kezdeni? „Kérdezz valamit.” Attól, hogy beleöntötte az egészet, a gépnek nem lett könnyebb dolga. Réka akkor értette meg, hogy a gép nem gondolatolvasó, és megírta a kérdést. László szerint munkavállalóként sokszor azt várjuk el az AI-tól, amit a kollégáinktól, a céges környezettől, magunktól vagy a családunktól nem kaptunk meg. „Annak semmi köze nincs semmilyen technológiához. Emberi kapcsolódási mintázatokat próbálunk sokszor helyre rakni.”

László a végén elmondja, miért indították a műsort: azt a hiányt látták, hogy a technológia a maga furcsa tükrében megmutatja, min kellene változtatnunk abban, ahogy egymáshoz viszonyulunk.

Kinek szól: munkatársaknak, akiknél most vezetik be az AI-t, vagy már használják; és vezetőknek, akik tudni szeretnék, mit gondolnak erről a kollégáik.
Kinek nem: annak, aki tiltással akar rendet tenni.

Amit holnap reggel elkezdhetsz

Az adásból:

1. Kérdezd meg a kollégádat, hogyan dolgozik. Mit változtatna meg, ha tehetné, és hol várna
   segítséget? László szerint ritka az ilyen beszélgetés a cégeknél.

2. Tiltás előtt magyarázd el. Mit tudnak ezek a rendszerek, minek mi a következménye, és az hogyan
   hat a munkahelyre.

3. Hiba után azt kérdezd: jól használtuk-e? A repülésben az egész szakma tanul egy-egy hibából.

4. Mielőtt beadsz valamit, nézd meg, mi a tét. A sütibannernél pici, itt nagyobb: amit egyszer
   rosszul megadtál, azt nehéz kiverni a gép fejéből.

5. Mondd meg a gépnek, mit akarsz. Az adat mellé kérdezz is valamit.

Fejezetek

Kérdezz tőlünk

Írd meg kommentben, mit kérdeznél az AI-ról, vagy mit nem mertél még megkérdezni. László válaszol, és a
kérdésekből látjuk, miről beszélgessünk legközelebb.

Hol találkozunk

• DCC-workshop, október 21., 13:30–17:30 — Claude Code PPC és SEO folyamatok építésére, négy óra
  gyakorlat.

A következő adásban

Hamarosan jön a 6. adás.

Videón:

· Az első rész, a vezetők oldala: https://signals.arcanian.com/p/4-adas-kinek-higgyen-a-gep
· Minden adás:

Arcanian Signals

Where there is no right answer, the only thing you can judge is how you got there.

By Laszlo Fazakas

Amikor egy cégvezető bevezeti az AI-t, a gép mindent igaznak vesz, amit kap: a Google riportját, a Meta számait, a könyvelést és mindegyik vezető véleményét. A negyedik adásban Back Réka arról kérdezi Fazakas Lászlót, mi történik ilyenkor a gyakorlatban, és hogyan dől el, kinek higgyen a gép. Ez egy kétrészes beszélgetés első fele: most a vezetők oldaláról van szó, a következő adásban a munkatársakéról.

A láthatatlan tulajdonos. Réka azzal kezdi, hogy a vezető AI nélkül is sokszor nem tudja, mi alapján hozza meg a döntéseit. László szerint az AI ezen úgy csavar, „mintha lenne egy láthatatlan tulajdonos, vagy a vezető fölötti vezető”. Az AI naponta, akár percenként dönt helyettünk, és ennek a cégvezetők sincsenek tudatában.

Mindenki igazat mondott. Egy webshop tulajdonosa riportot kapott a Google-től, a Metától és az Arcaniantól is, mellette ott volt a könyvelés. A csatornák együtt jóval több eladást jeleztek vissza, mint amennyi a könyvelésben látszott. Két hétig elemezték az adatokat, mire kiderült, hogy a könyvelő egy hónapot csúszott, a Google és a Meta pedig jóval több eladást írtak a maguk javára, mint amennyi valóban hozzá volt köthető. László így foglalja össze: „mindenki a saját maga világában igazat mondott, le lehetett vezetni a belső matematikájából, de együtt azért mégsem
maradt annyi pénz a vezető zsebében, mint szerette volna.”

Először az ellentmondások. A rendszert, amelyet a webshopnál bevezettek, először arra kérték, hogy keresse meg az ellentmondásokat. Ha a Google ennyi vásárlót hozott, a Meta pedig ennyit, miért látszik a webshopon csak a harmaduk? Ellentmond-e a webshop árbevétele a könyvelésnek?

Utána minden forráshoz „osztót” rendeltek: mennyit higgyenek el abból, amit mond.

Ezer forint, ezer dollár. Ha két bankjegyen ugyanaz a szám áll, de az egyik forint, a másik dollár, tudjuk, hogy más az értékük. Az internetes számoknál ez a képességünk valamiért kikapcsol: ha ekkora cégek adnak nekünk számokat, ugyan miért ne hinnénk el őket? László szerint az adatot magát elfogadhatjuk. Hogy a mi cégünknek mit ér, azt viszont nekünk kell megmondanunk.

Amit legutoljára hallott. Az egyik cégnél, ahol az Arcanian hetedik hónapja dolgozik, öt vezetőnek van saját nézőpontja a piacokról, a termékekről és a cég eredményeiről. László több mint öt hónap alatt mérte fel, kinek miben hihet. Egy AI-rendszer a tapasztalata szerint „elsőre mindenkinek mindent el fog hinni, és sajnos a válaszai azt fogják tükrözni, amit legutoljára hallott.” Ráadásul egy vezető ritkán engedheti meg magának, hogy ne legyen véleménye. Amit egy megbeszélésen gyorsan mond, az bekerül a rendszerbe, és a gép alapigazságként kezeli, mert nem tudja, milyen helyzetben, kinek és milyen szándékkal hangzott el.

Kit kérdezzünk meg? Sokan feltöltik a céges dokumentumokat, és azt várják a géptől: „mondd meg nekem a tutit”. László szerint ez nem működik. Ő interjúkkal kezd: egy pénzügyessel, aki megbízható számokat tud mondani, az informatikai vezetővel, a cégvezetővel, és egy-két emberrel, aki később használni fogja a rendszert. Aztán az embereket egymásról is megkérdezik: ki kire hallgat, ki kivel milyen bizalmi viszonyban van, és milyen kérdésekben. László szerint ez valójában szociológiai felmérés, és ebből épül fel a rendszer.

Jobb, ha nincs ott. Réka megkérdezi, végig ott kell-e lennie a vezetőnek. László előre szól, hogy ezzel sokakat magára haragít a magyar piacon: „Szerintem jobb, ha nincs ott.” Ha a vezető minden megbeszélésen és levelezésben jelen van, a kollégák mást mondanak, és a rendszer ugyanazokat a kötöttségeket képezi le, amelyekből a cég ki szeretne lábalni. Hogy a vezető maga is jól használja az AI-t, ahhoz segítség kell: valaki mellette, aki kutatói hozzáállással beszélget a géppel, és egy nagyjából már jól beállított rendszer. Különben elmegy a kedve tőle, és nagy ára van a tévedésnek.

Nincs trükk. A vezetők azt kérdezik, igaz-e, hogy ez már mindent megold, és csak be kell kapcsolni. Kipróbálták, bekapcsolták, de nem sikerült, és most trükköt várnak. „Nem szeretik meghallani azt, hogy nem tudunk trükköt.” László szerint három-hat hónap a minimum, mire egy ilyen rendszer összetett vezetői kérdésekre megbízható választ ad. Az Arcaniant ráadásul jellemzően válsághelyzetben keresik meg, utolsó mentsvárként. Ilyenkor László már az elején megmondja, hogy nagyon kicsi az esély arra, hogy a rendszer még ebben az évben kihúzza őket a bajból.

Már megtörtént. A vezetők leginkább attól tartanak, hogy ha kiadják a cég adatait az AI-nak, azt a versenytárs is megtudja. László rossz híre az, hogy ha rosszul csinálják, igen. A legtöbb pozícióban viszont az alkalmazott már most a saját fizetéséből, stikában fizet elő egy AI-eszközre, hogy könnyebb legyen a munkája. Ami miatt a vezető aggódik, az „legtöbb esetben, azt kell, hogy mondjam, már megtörtént.” Amit a gép már tud, azt teljesen kitörölni sem lehet; László szerint ez valamilyen módon minden AI-rendszerre igaz. Ezért ő már az első érdeklődésnél több rendszerben is megnézi, mit tud a gép a cégről.

Az első jel. Hamar megjön, akár az első pár hétben: felszínre kerül, hány egymásnak
ellentmondó vélemény és információ él a cégben. Ez a választóvonal. Volt, ahol az együttműködés nem élte túl az első három hetet, mert a vezető nem tudta elfogadni, amit látott. Az Arcanian ilyenkor sem állhat egyik oldal mellé, különben nem tudná üzembe helyezni a rendszert.

Lekéstük? Réka azt kérdezi, időben van-e még, aki most kezd bele. László szerint a riogatás alaptalan. Az amerikai rendezvényeken, amelyeket követ, azt mondják, „még az elejének az elején sem vagyunk.” Akik ma azt mondják, hogy lemaradtunk, egy éve is ezt mondták, és a cégek, amelyeket az AI a hírek szerint tönkretett, valahogy nem mentek tönkre. László szerint „ez a fajta
félelemkeltés azért kontraproduktív, mert nem maradtunk le.”

Öt érv. Végül Réka öt érvet kér, amiért megéri belevágni. László elsőnek a tisztánlátást
mondja: a vezető végre független nézőpontból láthatja több terület állapotát. A második, hogy jobb hamarabb átvenni a szemléletet, hiszen a kollégák úgyis használják az AI-t, és ha már így van, csinálják jól. A harmadik a versenyelőny: aki jól bevezetett egy ilyen rendszert, az még sokáig nem fogja megosztani, hogyan csinálta. A negyedik, hogy aki a Google-on, a Metán, a TikTokon vagy a Pinteresten hirdet, az már most komoly AI-felhasználó, csak a számlát és az eredményt látja belőle. Az ötödik: megnyílik egy újfajta beszélgetés a vezetés és az alkalmazottak között. Ha a kollégák végre nyíltan vállalhatják, hogy használják, és kapnak hozzá pár irányelvet, mindenkinek nagy kő esik le a szívéről.

Kinek szól: cégvezetőknek, akik most készülnek bevezetni az AI-t a cégüknél, vagy már elkezdték.
Kinek nem: annak, aki olyan trükköt keres, amitől holnaptól minden működik.

Amit holnap reggel elkezdhetsz

László tanácsai az adásból:

1. Vegyél elő egy papírlapot. Írd fel az adatforrásaidat, és írd mindegyik mellé, mennyire
   hihetsz neki, és milyen témában.

2. Először az ellentmondásokat kerestesd meg. Mielőtt bármit kérdeznél a géptől, kérd meg,
   hogy keresse meg, hol mondanak ellent egymásnak a forrásaid.

3. Mondd meg neki, mit ér az adat a te cégednek. Ezt a gép magától nem
   tudhatja.

4. Nézd meg, mit tud már rólad a gép. Kérdezd meg több rendszertől is: mit tudsz erről a cégről?

5. Vezetőként maradj ki az interjúkból. Ha a kollégák tudják, hogy ott vagy, mást mondanak.

És egy szám az adásból: három-hat hónap. László szerint ennyi a minimum, mire a rendszer összetett vezetői kérdésekre megbízható választ ad.

Fejezetek

Az időkódok a végleges idővonalra átszámolva.

Hol találkozunk

• Digital Cube Conference, október 15., 9:15–9:55 — Mi a baj ezzel a fiókkal? – építs
  rendszert, ami helyetted válaszol

• DCC-workshop, október 21., 13:30–17:30 — Claude Code PPC és SEO folyamatok építésére,
  négy óra gyakorlat.

A következő adásban

Egy hét múlva jön az 5. adás, a Kérdezz valamit!: ugyanez a beszélgetés a munkatársak oldaláról.
Mit jelent nekik az AI bevezetése, és miért lesz ez jó nekik?

Videón:

· Az előző adás: https://signals.arcanian.com/p/3-adas-mire-alapoztuk
· Minden adás:

Arcanian Signals

Where there is no right answer, the only thing you can judge is how you got there.

By Laszlo Fazakas

Ha a gép javasol valamit, és mi elfogadjuk: tudjuk-e, mire alapozta? És ha rossz döntés lesz
belőle, ki viseli a következményeit? A harmadik adásban Back Réka ezt kérdezi Fazakas Lászlótól.
László válasza rövid: a felelősség nálunk marad. Az adás arról szól, mit tehetünk azért, hogy
ezt a felelősséget vállalni is tudjuk.

Döntés volt, vagy csak ötlet? A cégvezető két megbeszélés között odaveti a kollégáknak, hogy
kérne egy kampánytervet, ami holnaptól fut. Hogy ez döntés volt vagy csak ötlet, azt egy jó
kolléga a hangsúlyból tudja. A gép ezt nem érzi. László ezért megtanította a rendszerét, hogy
ilyenkor kérdezzen vissza. A gép néha kíméletlenebb, mint bármelyik kolléga: „tegnapelőtt azt
mondtad, hogy ez A”, most pedig C-t hall ki abból, amit mondtál. László szerint ettől jobb főnök
is lehet, mert szembesül vele, hányszor bánik így a saját kollégáival.

Kilenc nap döntés nélkül. Egy másik rendszer két óra beszélgetés után megálljt parancsolt: a
beszélgetés kilenc napja tart, és egyetlen döntés sem született. Közölte, hogy több lehetőséget nem
sorol fel: „valamit döntsünk el, mert ez így nem fog menni.” Erre László tanította meg a rendszert: egy ponton
figyelmeztessen. Ehhez viszont felhatalmazás kell, olyan, amilyet a Keresztapa ad a
consiglieréjének. Egy jó Keresztapa azért alkalmaz ilyen embert, mert ő sem tévedhetetlen.

Döntési napló. László évek óta diktafonra mondja, mi alapján dönt: milyen adatforrásokat és
dokumentumokat használ, mi jár a fejében. Így a gép a végeredménnyel együtt a gondolatmenetet is
megkapja. Aztán vizsgáztatja a rendszert, ahogy a céhekben a mester a tanítványt: én ezt már
eldöntöttem, te hogyan döntenél? Amit a gép kihagyott, azt László pótolja; amit pedig nem kellene
figyelembe vennie, azt kiveszi.

A gondos gazda. László jogi diplomamunkája a gondos gazdáról szól. A római gondos gazda
ismerte a környezetét: a családját, a római államot, a természetet. László nagyapja még el
tudta mondani a városi unokájának, mikor kell vetni és aratni, mert benne élt abban, amiről
beszélt. A mai cégvezető specializálódott, és ezzel elveszítette ezt a kapcsolatot. Nehezen
mondja el egy kollégának is, mit csináljon, hát még egy agentnek. „Nem várható el az embertől,
hogy jó döntést hozzon, hogyha neki már nincs kapcsolata a környezetével.” A gépre ugyanez áll.

Az üres lap nem üres. Amikor megnyitjuk a ChatGPT üresnek tűnő lapját, az emberi történelem
jó részével vitázunk, csak épp kevésbé az európaival, és a legkevésbé a magyarral.
A nagy modelleket túlnyomórészt angolszász szövegeken tanították, ezért az alapbeállításuk nem
nekünk kedvez. László mérései szerint, ha a vallásról, az életről vagy a halálról kérdezzük
őket, az amerikai tankönyvekhez állnak közelebb. Legalább annyit mondjunk meg
nekik: „Ez itt nem az Egyesült Államok, itt nem angolul beszélünk.”

Írjuk le a sikertelen döntéseket is. Ha csak a sikereinket adjuk át, a gép azt sem tudja,
mit nem választottunk, és miért. Réka azt veti fel, hogy mutassuk meg a gépnek azt az utat is,
amerre végül nem mentünk. László szerint csak így lehet. Egy cég sokszor épp annak
köszönheti, hogy még életben van, ami nem sikerült. Ezt is le kell írni: miért gondoljuk
sikertelennek, és mit mondott róla a piac, a kollégák vagy a könyvelés.

Amikor a gép leszűkít. László akkor érezte, hogy a rendszerük igazán segít, amikor a
döntést arra a pár lehetőségre szűkítette, amik között van értelme választani. Megkérdezte
tőle, milyen esélyeik lennének az amerikai piacon. A gép visszakérdezett: milyen időtávban,
milyen kapcsolatrendszerrel, és kaptak-e már akár csak egy dollárt amerikai ügyféltől. Kaptak,
pár ezret. A végén oda jutottak, hogy vagy az időtávon kell változtatni, vagy a
kapcsolatokon, de úgy, ahogy László szeretné, nem fog menni. Hogy jól látja-e a gép?
„Annyira jó, amennyire betanítottam.”

A döntést én hozom. „A döntést nem a gép hozza, a döntést én hozom. Viszont az, hogy ő
mennyire tud jó környezetet adni nekem a döntéshez, azért meg én vagyok a felelős.” A gondos
gazda mai megfelelője az, aki legalább bizonyos pontokon végigkíséri a saját agentje fejlődését,
mert csak így tudja, milyen adatokból jutott el oda, ahol tart. Ezt semmilyen külső szabályozó
nem kényszeríti rá. Ahogy László mondja: jó gazda attól leszel, „hogyha törődsz vele”.

Kinek szól: cégvezetőknek, akik már hagyják, hogy egy AI-eszköz javasoljon nekik, vagy
döntsön helyettük.
Kinek nem: annak, aki azt várja, hogy a felelősséget majd átveszi a gép.

Amit holnap reggel elkezdhetsz

László tanácsai az adás végéről:

1. Kérdezd meg a gépet arról, amit egy juniornak az első napon megtanítanál. Mit tud a
   magyar piacról, mekkora itt a vásárlóerő, kik a nagy versenytársak a szektorodban. Ha olyan
   választ ad, amiért egy juniort is kirúgnál, ne bízz rá nagy döntést.

2. Mondd meg neki, hol van. Ország, nyelv, időzóna. László mérései szerint magától nem
   tudja, csak úgy csinál, mintha tudná.

3. Tanítsd meg, hogy szóljon, ha valamire nem tud felelni. László már annak is örülne, ha
   egy rendszer ki merné mondani: „Ilyen paraméterek mellett nem tudok jó döntést ajánlani
   neked.”

4. Új modellnél kezdd elölről az ellenőrzést. Két hét múlva a következő modell már
   másként felelhet ugyanarra a kérdésre.

És egy szokás az adás elejéről: amikor döntesz, mondd diktafonra, mi alapján döntöttél. A gép
így a gondolatmenetedet is megismeri.

Fejezetek

Az időkódok a végleges vágatra átszámolva.

Hol találkozunk

• Internet Hungary, október 6–7. László a 7-én ad elő.

• Digital Cube Conference, október 15., 9:15–9:55 — Mi a baj ezzel a fiókkal? – építs
  rendszert, ami helyetted válaszol

• DCC-workshop, október 21., 13:30–17:30 — Claude Code PPC és SEO folyamatok építésére,
  négy óra gyakorlat.

A következő adásban

A 4. adás hamarosan jön.

Videón:

· Az előző adás: https://signals.arcanian.com/p/2-adas-mit-tud-rolad-a-gep
· Minden adás:

https://signals.arcanian.comk

A teli pohár. Egy zen történetben a tanítvány tele van olvasmányokkal, a mester mégis
elküldi: teli pohárba nem lehet tölteni. A gépekkel ugyanezt csináljuk. Mindent beletöltünk, ami
rólunk szól, és ezzel akadályozzuk meg, hogy jó választ kapjunk. A dokumentumaink különböző
években születtek, és különböző mértékben igazak. Van, ami csak a tulajdonosnak igaz, és van,
ami csak régen volt igaz. Ha a gépre bízzuk a válogatást, a rosszabbik eset az, hogy el is végzi.

Iránytű és időérzék nélkül. Ha mi nem adunk a gépnek iránytűt, a saját tudásához nyúl, az
pedig a világ összes cégének átlaga. Azt sem tudja magától, melyik anyagunk született előbb.
László ezért úgy tanítaná, mint egy tehetséges pályakezdőt: ebből ezt a részt veheted
figyelembe, ezzel ne foglalkozz, azt már régen meghaladtuk.

A gép mindenképp befejezi a mondatot. Sosem mondja, hogy erre nem tud felelni. Réka ismeri a
helyzetet: rászól a gépre, hogy ez nagy csacsiság, mire az két sorral később ugyanazzal a szent
meggyőződéssel állítja az ellenkezőjét. László a megfelelni akaró kamaszhoz hasonlítja.
Az Arcaniannek januártól háromnegyed évébe telt, hogy a saját rendszere ellent tudjon mondani.

Egy elemzés, amelyik minden körben igazat mondott. Egy ügyfélnél az első kör szerint a
vállalkozás jól skálázható: több hirdetés, több bevétel. A könyvelés az ellenkezőjét mutatta. A
negyedik-ötödik kör után derült ki, hogy több rendelésnél aránytalanul nő a futárdíj, vagyis a
fuvarozási megállapodás volt rossz. A gép minden körben „jól tudta”. Az első jel máshonnan jött:
az ügyfél aggódó hangjából a telefonban.

Az AI-suttogó. Aki lovakkal foglalkozik, ismeri a lósuttogót: azt, aki segít, ha nem
találod a hangot a lovaddal. László szerint a cégeknek ugyanígy AI-suttogóra lesz szükségük.
Félig pszichológusra, aki érzi, mit szabad szó szerint venni abból, amit mondanak, és félig
folyamatemberre, mert a gépek folyamatokat látnak és folyamatokat működtetnek. Az AI-suttogó
olvas a sorok között, és amit ott talál, lefordítja a gépek nyelvére. László kételkedik abban, hogy ennyi
szerepet egyetlen ember elvisz, szerinte inkább csapat lesz belőle.

Vitakultúra. „Ha mindenki egyetért velem a szobában, rossz szobában vagyok.” Ha a gép nem
vitázik velünk, az veszélyes: vagy nem tanítottuk meg vitázni, vagy alkalmatlan a tanácsadó
szerepére. Végtelen vita sem kell. Az Alkotmánybíróság is rögzíti, miben értettek egyet, miben
nem, és ki mit tett volna hozzá különvéleményben. Abban is meg lehet egyezni, hogy nem értünk
egyet, és onnan tovább lehet lépni.

A hajó már elment. A telefonunk, a navigáció és a Facebook hírfolyama naponta hoz helyettünk
döntéseket. Ahogy a lovasnak, nekünk is a hat-hétszáz kilós nyers erőt kell megtanulnunk
kezelni, és ehhez mindannyiunknak kicsit AI-suttogóvá kell válnunk.

Kinek szól: cégvezetőknek, akik már feltöltötték a cégüket egy AI-eszközbe, vagy most
készülnek rá.
Kinek nem: annak, aki azt várja, hogy a gép majd kitalálja, milyen a cége.

Amit holnap reggel elkezdhetsz

László három lépése az adás végéről:

1. Kérdezd meg a gépet, mit tud a cégetekről. A kollégák a saját területükön nézzék át, és
   jelöljék sárgával, amit rosszul tud. Ez az első hónap munkája.

2. Hagyd, hogy beszélgetés legyen belőle. Közben kiderül, hol gondoljátok ti magatok is
   másképp.

3. Előbb emberként döntsétek el, hogyan legyen innentől. Ha megvan az egyetértés, azt
   tápláljátok vissza a gépbe.

És egy gyakorlat a 40. percből: írd le három A4-es oldalon, miért csinálod a cégedet,
marketingnyelv és pozicionálás nélkül. Add oda egy AI-eszköznek, és ne blogot kérj belőle:
keresse meg benne a hibát. Hamar kiderül, jó rendszert használsz-e.

Fejezetek

Hol találkozunk

• Internet Hungary, október 6–7. László a 7-én ad elő.

• Digital Cube Conference, október 15., 9:15–9:55 — Mi a baj ezzel a fiókkal? – építs
  rendszert, ami helyetted válaszol

• DCC-workshop, október 21., 13:30–17:30 — Claude Code PPC és SEO folyamatok építésére,
  négy óra gyakorlat.

A következő adásban

3. adás — Mire alapoztuk? Ha a gép arra támaszkodva javasol valamit, amit rólunk tud:
jól javasolja-e?

Videón:

· Minden adás:

Arcanian Signals

Where there is no right answer, the only thing you can judge is how you got there.

By Laszlo Fazakas

Miért lesz rosszabb a gép attól, ha mindent beletáplálunk. Az ember tud felejteni, és ez egy alulértékelt képesség. Az AI-t meg kell tanítanunk felejteni, különben benne marad az is, ami tavaly volt igaz ránk.

Hány döntés van egy e-mail mögött. Öttől ötszázig. Kinek írom, milyen hangon, mit teszek bele és mit hagyok ki — ezek mind eldőlnek, és egyikről sincs feljegyzés.

Hogyan vezetjük félre a gépet anélkül, hogy akarnánk. Egy cégben négy történet él egymás mellett: egy az adóhatóság felé, egy a tulajdonosok felé, egy a csapat felé, egy az ügyfelek felé. Mind a négy igaz valamennyire, és egyik sem fedi le a teljes képet. Aztán mind a négyet beadjuk a gépnek, és elvárjuk tőle, hogy válogasson.

Az amerikai versenyt nem nekünk kell megnyerni. A számítási kapacitás versenye eldőlt. Európa akkor jár jól, ha kiszáll ebből a játékból, és megnézi, mit tud a saját filozófiai gyökereiből felépíteni.

Miért kellene minden cégnek saját AI-használati politikája. Számviteli politikája mindenkinek van: a törvényen belül a cég megmondja, ő ezt hogyan alkalmazza. Ugyanez a gépekre is működne.

Kire tartozik a biztosíték. A fegyvergyártó nem úgy adja el a fegyvert, hogy a biztosítóberendezést otthon szereld bele. A gépeknél most pont ezt kérjük a cégvezetőtől.

An hour into the data, I had the first answer: every one of those eleven names had been ranking #1 in organic Google search for the past nine months. Stable. Not one had slipped. The position tracking was uncontested.

So what were they seeing?

Three days later I had the full picture. The complaint was real, the diagnosis was wrong. It wasn’t a ranking loss. It was three different problems wearing one mask — three layered failures, each requiring a different remedy. If their previous agency had taken the complaint at face value, they would have spent three months fixing the wrong thing.

Here are the eight lessons I took from the investigation. After them, the toolkit argument: why this kind of work is impossible without both a data aggregator and a competitive-intelligence platform — and why running with only one of the two is the most common, most expensive mistake I see in mid-market marketing operations.

1. The client’s complaint is data, not a conclusion

Anyone who says “we’ve lost our SEO position” is reporting something real. They’re not making it up. But the cause they assign to that experience — that’s their hypothesis, not yours.

Human perception works at the visual level. Google’s results page over the last eighteen months has been gutted of organic real estate above the fold. Sponsored ads, AI Overview, Map Pack, “People also ask,” knowledge panels — all of it now sits above the first organic result. A user who once saw their own brand at the top of the page now scrolls three screens to find it. Their experience: “we’ve lost #1.” The measurement: #1 is intact.

Both are true. They describe different things.

Take the complaint seriously as a signal. Take it lightly as a diagnosis. Your job is to surface the actual mechanism behind the symptom — not to inherit the client’s interpretation.

2. If a position-tracking project exists, use it first

If the client (or you) already runs a Semrush Position Tracking campaign, those keywords are what the business actually monitors. Ad-hoc keyword pulls or domain-level audits are two notches lower in evidentiary value.

In this case, the project tracked 100 keywords over nine months: weekly history, plus daily for the last sixty days. It included all eleven branded entity names plus the brand keyword variants. There was no better dataset I was going to find.

The verdict from that single report: 95 of 100 keywords stable or improved. Five marginally worse than baseline. None of the five were branded entities.

The diagnosis was effectively decided by that one report. Everything after was confirmation, not investigation.

3. Position, visibility, and traffic are three different things

The market uses these words as synonyms. Client complaints, press articles, junior-consultant slide decks — all conflate them. They are not the same metric, and the distinction is the most important diagnostic step you’ll do in this kind of work.

• Position = where you rank in the organic list (1, 2, 3...)

• Visibility = the probability the user actually notices your result, weighted by SERP features and CTR curves

• Traffic = how many clicks land on the page

In 2026, the typical pattern is: position stable at #1, visibility down 20-25% (because ads and AI Overview push the result below the fold), traffic down 40% (because AI Overview consumes the click before it leaves Google).

When a client says “we lost our SEO position,” they almost certainly mean traffic or visibility — not position. Your first ten minutes of investigation should split the three. They each have different fixes.

4. Industry context is mandatory

Before you tell anyone “your SEO got worse,” pull at least three competitors’ traffic curves over the same window. The story can flip entirely.

In this case the picture was:

• Client: -40%

• Competitor A: -54%

• Competitor B: -67%

• Competitor C: -2% (held)

Three of four peers lost more than the client. The fourth held — because their content footprint was service-page driven, not symptom-blog driven, and the AI Overview rollout doesn’t eat service pages the way it eats blog explainers.

So the story isn’t “your SEO is broken.” It’s “your industry’s content model is in transition, you’re losing slightly less than average, and the survivor in the category is doing X.” That’s a fundamentally different conversation, with a fundamentally different recommendation.

If you only look at the client’s own data, you’ll either flagellate them unfairly or reassure them unfoundedly. Neither helps.

5. Backlink IP/TLD clustering is a spam signature

When the backlink count crashes, look at the surviving referring domains’ IP addresses and TLDs. If you find twenty or thirty subdomains pointing to a single IP, or a cluster of .top, .party, .mom, .cfd, .sbs, .monster TLDs, you’re not looking at organic link decay. You’re looking at the residue of a previous private blog network.

In this case the backlink profile had collapsed -94% from peak (some 6,300 links → 376). The 376 survivors were dominated by:

• ~30 blogspot subdomains from a single IP, with auto-generated names matching a *-like.blogspot.com pattern

• A Moldovan cluster on a single IP across a dozen spam TLDs

• A French cluster doing the same with different TLDs

• About 15 actually legitimate domains in the client’s home market

The implication is unambiguous: a previous SEO engagement built (or bought) a PBN. Google has been quietly devaluing it for eighteen months. The traffic decay you see now is the lagging consequence of authority that was never real.

This finding changes the remedy. Filing a disavow file is necessary but insufficient — disavowing fake links doesn’t restore real ones. The recovery requires a separate, longer-horizon link-building program.

6. LLM extraction hallucinates verifiable values — verify with raw grep

I learned this one the expensive way during this investigation. I asked an LLM-powered scraping tool to extract the canonical URL from a key page using its jsonOptions schema-extraction feature. The LLM returned a URL with a typo — an extra letter in one of the slug words — that simply does not exist on the live site.

If I’d trusted that output, I would have told the client “your canonical is broken, fix it” — and looked foolish when their developer pulled up the actual page source.

The lesson generalizes: when you need a verbatim, verifiable value — a canonical URL, a schema field’s exact contents, a meta tag’s literal text — don’t extract with an LLM. Pull the raw HTML and grep. Regex doesn’t hallucinate.

LLM extraction is excellent at categorization, summarization, and fuzzy pattern-recognition across messy text. It is unreliable at returning specific values you’ll later be held to. Match the tool to the task.

7. Discovery, not pronouncement — when new data arrives, append a delta

On day one I wrote an executive summary in five sections. On day two, new data — the position-tracking report — refuted one of my own framing claims from day one.

The temptation: rewrite yesterday’s summary, hide the contradiction, present a clean story.

The discipline: don’t. Write a delta document. Mine became 05_DELTA_position_tracking_evidence.md. The earlier section stays. The new finding sits next to it, dated, with its own evidence. The reader (and you) can see how the thinking evolved.

Two reasons for this. First: clients trust the practitioner who corrects themselves visibly more than the one who pretends to be infallible. The correction is the credibility. Second, and more important to your own future work: if you erase the original claim, you lose the metadata about why you believed it at the time. Your future self will look back at a similar case and want to know what made you initially get it wrong. That information is only preserved if the original framing remains in the record.

The map is not the conclusion. The map is the trace of the thinking. Keep the trace.

8. The deliverable layer needs two passes

The output of an investigation is only as good as the language carrying it. If you write client-facing material in the modern hybrid of consultant-speak and AI-generated stock phrasing — the world recognizes it instantly. The reader doesn’t think “this is a professional”; they think “this is a template.”

Two passes:

1. Voice rewrite by an explicit style rubric. For me, this means: no “at the end of the day,” no “differentiation that compounds,” no “leverage the framework,” no “actionable takeaways.” Active verbs over nominalizations. One thought per sentence. Native idioms over logical connectors.

2. Grammar/voice verification by a second pass — ideally a human reviewer or, in my case, a specialized grammar-checker agent that flags the patterns I missed. The first pass moves the prose; the second pass catches what slipped through.

This matters more than its description suggests. The deliverable is the product. If the language is generic, the diagnosis reads as generic — even when it’s not.

The toolkit argument: why one platform isn’t enough

I want to spend some time on this, because it’s the part of the playbook most agencies and in-house teams get wrong.

The investigation above was impossible to do with only one data source. It required two distinct planes of data, and any practitioner who tries to operate on only one plane will misdiagnose the same kind of case I just walked through. Here’s why.

Plane one — first-party data, aggregated

This is what happened on your properties. Sessions, clicks, impressions, conversions, ad spend, revenue. Sourced from GA4, Google Search Console, Google Ads, Meta Ads, Shopify, your CRM. In a serious operation, all of it lands in a data aggregator like Databox so you can see twenty metrics in one dashboard instead of opening five tabs and exporting CSVs.

What this plane tells you:

• Did the traffic actually decline? (GA4 organic sessions, week-over-week)

• Did rankings move, or did clicks move? (GSC impressions stable + clicks dropping = CTR collapse, not rank loss)

• Did a paid campaign get paused? (Ads spend zero on dates the traffic dipped — you’d be surprised)

• Did the loss show up in revenue? (Shopify / CRM dollar figures)

• Are conversion-quality clicks down, or just impression-quality clicks? (assisted conversions vs first-touch)

Without an aggregator, you can answer all of these — but it takes you four hours instead of fifteen minutes, and you’ll skip the questions that need a side-by-side view across data sources to even ask. The aggregator is not a luxury. It’s the difference between “we have a hypothesis in one sitting” and “we’ll get back to you with the data next week.”

Plane two — third-party data, the SERP and the competitor set

This is what your customer experienced on Google, and what your competitors did during the same window. Sourced from a competitive-intelligence and CX platform — Semrush in our case, but Ahrefs or SimilarWeb fill similar roles.

What this plane tells you:

• What did the customer see when they searched your name? (SERP-feature presence, ad density, AI Overview, Map Pack)

• Did your rank actually move? (position tracking, weekly history)

• Did your peers lose, hold, or grow? (competitor traffic curves)

• Is your authority foundation intact? (backlink profile, referring domains, anchor distribution)

• Who is bidding on your branded queries? (paid SERP intelligence)

• What keyword universe do you rank for, and how is it changing? (organic keyword counts over time)

The customer’s experience of search is third-party data. GA4 will tell you that 1,000 sessions disappeared. It will not tell you why. The SERP screenshot tells you why — but you can’t get the SERP from your own analytics. You need a tool that goes to Google as a user and brings back what the user saw.

Why one alone fails

If you only have first-party data: you see the symptom (traffic down 40%) but you cannot see the mechanism (paid hijack of branded names, AI Overview eating click-through, competitor displacement). Your recommendation will be vague — “improve content quality” — because you can’t see the actual cause.

If you only have third-party data: you see the SERP changes but you don’t know whether they hurt you, helped you, or had no impact at all. Your recommendation will be over-confident — “AI Overview is killing you” — even if your traffic actually held because half the lost clicks went to a longer-tail page that’s now converting better.

The diagnostic moves the agency runs in this kind of investigation — segmenting CTR decay from rank loss, linking traffic timing to algorithmic events, distinguishing structural decline from temporary noise — none of them work without both planes. Trying to run a serious investigation with only one is like trying to diagnose a fever without taking either the temperature or the medical history. You can guess. You can’t know.

The cost argument that keeps mid-market teams on a single platform — “we already pay for X, why also pay for Y” — is the wrong frame. The right frame: an investigation done with one plane will be wrong about 30-40% of the time, and the cost of being wrong on a six-figure marketing decision dwarfs the second subscription.

The pattern that emerged

Stepping back, the meta-finding from this engagement is simple: the client’s complaint hid three orthogonal problems, all real, each with a different remedy.

If the agency had heard “we lost rankings” and proposed a generalized SEO project, they would have spent six figures on the wrong fix. The complaint was a single sentence. The actual diagnosis required three distinct workstreams — at three different time horizons, with three different price tags.

The honest part

For most of my career I assumed that when a client raised a complaint, the complaint was the problem. Years of experience taught me otherwise. The complaint is the client’s perception of the problem — valuable signal, sensitive, worth taking seriously. The diagnosis is mine, and isn’t transferable to the person describing the symptoms. A doctor doesn’t write the chart in the patient’s words.

A colleague last year spent two weeks on an “SEO rebuild” the client requested. By the end of week two it turned out the traffic decline was from an unrelated paused paid campaign — the client hadn’t mentioned it. Two weeks of work, a five-figure invoice, and the actual problem stood exactly where it had started.

Take the complaint seriously. Don’t outsource the diagnosis.

Next week I’ll write about the visibility-space metric — what to track instead of pure rank in 2026, and how it changes the way you compute customer-acquisition cost when AI Overview consumes the click before it leaves the SERP.

A client you did good work for says: “We’ve decided to go in a different direction.”

They can’t say what direction.

Two weeks later they have hired an agency whose pitch ChatGPT could have produced in five minutes. Images that look art-directed. Video that used to take a crew weeks. A framework that sounds like a framework. A strategy document that reads like strategy. None of it will work. All of it won the room.

You drive home running the tape. Maybe the creative. Maybe the price. Maybe you should have used more AI. A voice that has been getting louder all year says: maybe we just need a better idea.

That voice is what breaks agencies. Not the client. Not the creative. Not the price.

It is the belief that there is one right thing — one idea, one tool, one hire — that, once found, fixes everything. Every agency that breaks in the LLM era breaks through this voice.

The pressure outside matches the voice inside. Clients leave after eighteen months despite good work. “We can get that cheaper.” Thinner agencies win pitches with posture alone. Some clients don’t want depth at all — “I don’t care about the brand, I care about the number” is a real sentence in a real meeting.

We know this loop. We lived in it for years.

Brand matters on both sides of the table. For the agency, it is the most important asset — the thing no tool copies and no subscription ships. For the client, it is the same — the thing that compounds into the number they will care about three years from now. The answer to “I don’t care about brand” is not to drop brand work. It is to keep building, measure in the language the client trusts, and let the lever show its work.

What you want is not what the voice chases.

You don’t want a better pitch deck. You want to not need one.

You don’t want a better idea. You want to be the agency called first when something breaks.

You don’t want to use more AI. You want to use it so your agency gets more your agency, not more generic.

The chase for a better idea is the voice talking. It is not you.

We made a different choice.

The billable hour stood in for expertise for twenty years. The hour was scarce. The knowledge behind it was hard to copy. AI did not make knowledge cheap. It made delivery of knowledge cheap. The hour is still scarce. The deliverable is not.

A seat of ChatGPT in a billable-hour agency is a faster treadmill. So is a seat of Claude. The tools commoditize by the month. If your differentiator is which AI your team uses, you do not have a differentiator. You have a subscription.

AI is only the latest aisle in a supermarket that got enormous while nobody watched. The Chiefmartec landscape counts more than fourteen thousand martech tools. Analytics, attribution, automation, enrichment, measurement, orchestration. More every quarter. Your edge cannot live in which ten you picked. It has to live somewhere those fourteen thousand tools cannot reach.

So we stopped.

We stopped trying to find better creative. We started trying to see what was actually broken above the layer where our work happened. We stopped treating the agency as a labor pool. We started building it as an operating system.

The operating system has six parts. Each has a name.

• Ladder — where the agency stands on its AI adoption journey, honestly named.

• Layers — where in the client’s business the problem actually lives.

• Trail — a source you can point to for every claim.

• Rhythms — loops around projects, reviews, retros.

• Memory — what the system remembers so people don’t have to.

• Accord — humans and AI sharpening each other on real work.

What we rejected:

• The billable hour as a unit of value

• Tools without protocols

• Delivery without loops

• Frameworks without memory

• Cleverness without evidence

• “AI transformation” as a line item on a proposal

We have not finished rejecting these. We reject them one engagement at a time. Anyone who says the billable hour is dead at their agency is selling you something.

The operating system has six parts. The practitioner who runs it has seven habits. These are the seven.

1. Diagnose before you deliver. See the constraint above your work before you propose the work. Every brief starts with a diagnosis, not a deck. Present the finding clearly. Show the evidence. Ask the client where it doesn’t match what they see on the ground. Inquiry is how professionals work together. Apologetic tentativeness is weakness wearing humility as a costume.

2. Show the trail. Every claim has a source you can point to. When you don’t have one, you say “I don’t know yet, but I’ll find out.” AI hands you precise-looking numbers — percentages, benchmarks, charts with trend lines — that turn out to be invented. Confident wrong is its default output. The trail is the only defense.

3. Close every loop. Projects are real — start dates, end dates, invoices. What cannot end with the project is the learning. Close with a post-mortem: what worked, what broke, what the next project starts from. A project that ends with an invoice extracted value. A project that ends with a loop built it — for both sides.

4. Work in accord with the machine. Use AI as a teammate whose output you can inspect. Keep your thinking legible to it. Keep its output legible to you. Neither replaces the other. Automation replaces the human. Accord assumes the human is irreplaceable and makes the machine useful to them — and to the next human who walks into the room cold.

5. Carry lessons through people. Agencies do not learn. People do. Patterns travel in conversation, in review, in a practitioner who sees the same shape a year later in a different business. And they travel through each practitioner differently — a pattern named by one lens is provisional until a second lens confirms it. Know whose lens you are reading. Systems help, but they serve the connection between people — they do not replace it.

6. Count on the people in the room. Strategy that counts on people holds. Strategy that counts around them collapses. The client, their partners, their franchisees, the ops manager who will live with what you recommend — they sit in the room, not across the table. They carry context you will never have. Ask their questions. Write their names into the plan. And know who they are — cognitive shape is as real as a role title, and a seat that fits the person holds while one that does not drifts. Együtt — together — is a working relationship, not a slogan.

7. Sharpen the instrument. The agency is the product. Invest in what compounds — craft, brand, memory, accord. Retire what does not — the billable hour, the tool-of-the-month, the deck nobody remembers. The renewal is the work.

Here is what happened.

We started losing clients. The work was good. The clients still left. For a long time we thought it was the creative, the price, the pitch. It was none of those. We were losing them because we could not see — or name — the thing actually broken above our work. Once we understood that, we could start to reverse it.

Pitch meetings became short conversations. We came with a diagnosis, not a deck. Clients said things they had never said out loud, because nobody had given them a framework to say them with.

Clients stopped drowning in documents. The agency stopped shipping another deck. We started maintaining a system — one place where what was decided last month is still findable today. Remembering moved off the owner’s desk.

Every agency will soon have access to the same leading AI. Equal access is the new floor. What sets one agency apart sits above the tool. It sits where agency work has always lived — in the craft, the client knowledge, the rhythms, the frameworks, the scar tissue.

None of this is new work. It is the work agencies have always done at their best. What started compounding once we wrote it down:

• Client knowledge a good account lead carries in their head — that the CMO will not read past page three, that last year’s rebrand still stings, that the sales team resents marketing for something said in Q2. No AI starts a Monday knowing this.

• The discipline of showing your work. AI invents. A good agency has always insisted on the source.

• The rhythms that make an agency an agency — weekly status, monthly reviews, quarterly retros, the Friday post-mortem. No AI installs itself into a calendar.

• The house frameworks built over years — the discovery refined across fifty clients, the brief with sections nobody else writes. AI gives generic. The house gives specific.

• The scar tissue — the twenty ways a campaign can die. The client type that always bounces in month three. No AI has lived through this. A practitioner has.

An agency that rents AI gets what everyone gets. An agency that runs on one gets what compounds — the craft it already had, plus a way to keep it.

Some tool vendors already understood the agency’s calling. A handful of platforms were built on an insight the rest of the category missed: performance numbers are useful only when a practitioner can ask questions of them in context. An agency is orchestrating a customer journey, not maintaining a list. These vendors know their place is inside the substrate, not around the agency. We build the operating system on top of what they make.

We were not losing clients because our work was bad. We were losing them because we could not see — or name — the constraint that lived above our work. The moment we got the tools to see it, we stopped being interchangeable.

The tool was never the edge. The ability to see above the tool was.

This is what we mean by accord: the agency’s thinking is legible to the machine, the machine’s output is legible to the humans, and both get sharper every time either touches real work.

Accord is not automation. Automation replaces the human. Accord assumes the human is irreplaceable, and makes the machine useful to them — and to the next human who walks into the room cold.

If you are losing clients after eighteen months and you don’t know why — it is not your creative. Not your pricing. The constraint lives above your work, and nobody has given you the tools to name it.

Until you have those tools, every optimization is a guess. Every pitch is a performance.

The voice from the first page goes quiet when the agency becomes the kind of thing that can see the constraint above its own work, name it to a client, and still be in the room twelve months later because the naming was the reason it stayed.

Most practitioners reading this are on a ladder that runs from “I sometimes ask ChatGPT to write an email” to “my agency runs a live diagnostic engine on real client data.” Ten levels. No shame on any of them. The practitioner at Level 0 is not behind the one at Level 8. They are earlier in the same work.

The credential here is practice that can be inspected. A practitioner who runs their agency as an operating system has artifacts — a registry, a log, a set of loops, a network of findings. We share artifacts. We sign patterns. We review each other’s reasoning. We let the work speak. If certification ever makes sense in this movement, it will be earned by what a practitioner has built, not by a seat warmed in a course.

If the problem an agency solves fits inside a single person’s head, the person will leave and take the agency with them. If it fits inside a folder of SOPs, the folder rots. If it lives in an operating system — versioned, inspectable, run in accord between humans and machines — the agency survives its founders and the tools they started with.

That is the work. That is the movement.

If this is how you already think, you are one of us. Come build.

— The first practitioners

A post about this is making the rounds, and it introduces a framework I think is genuinely useful — four tiers of AI users:

Tier 1: Spectators. Still watching. Reading about AI, not using it.

Tier 2: Casual users. Rewriting emails, summarizing docs. Useful but shallow.

Tier 3: Harness users. Running OpenClaw, Cursor, or other tools with Claude under the hood. Real leverage, but borrowed infrastructure at a subsidized price. When the price changes, the workflow breaks.

Tier 4: Builders. Working in Claude Code, in the terminal. They have subject matter expertise and they’re building things that compound. Their workflows survive pricing changes because they understand what’s happening under the hood.

The post argues that the Tier 3 → 4 gap is where careers diverge.

I agree. But the framework is missing a tier.

Tier 5: System Builders

Tier 4 builds workflows. Tier 5 builds systems that encode judgment.

Here’s the difference.

A Tier 4 developer writes a Claude Code script that:

• Pulls Google Ads metrics from Databox

• Formats them into a table

• Adds a summary paragraph

That’s a workflow. It works. It saves time.

A Tier 5 practitioner builds a system where:

• 14 agents know which data source has a 48-hour sync lag and checks the lag BEFORE flagging an anomaly

• A confidence engine scores every finding on a 0-1 scale based on source reliability, evidence classification, and assumption verification status

• When Google Ads says “conversions dropped 74%,” the system generates 6 competing hypotheses, scores each against the evidence, and only creates a task if the confidence is above 0.4

• The system knows that Google’s “conversions” use a 30-day click window while Meta’s use 7-day click — and flags the mismatch when someone tries to compare them

• A recommendation tracker checks whether the advice it gave last quarter actually moved the metric — and if it didn’t, says so before recommending the same thing again

• Per-client alarm calibration means a 15% drop is noise for one client and a crisis for another — because the system knows each client’s baseline and volatility

That’s not a workflow. That’s judgment, systematized.

Why the Gap Between Tier 4 and Tier 5 Is So Hard to Cross

Tier 3 → 4 is a technical gap. You learn Claude Code, you understand the terminal, you build things. Effort + learning = progress.

Tier 4 → 5 is a domain expertise gap. No amount of technical skill will get you there without subject matter knowledge worth systematizing.

You need to know:

• Which questions to ask (not which prompts to write)

• What “normal” looks like (not just what “good” looks like)

• When the data is lying (not just what the data says)

• Which recommendation will actually work for THIS client, not just what works in general

This takes years. Not of AI experience — of domain experience. The 22-year marketing practitioner who learns Claude Code in a weekend is closer to Tier 5 than the developer who has used Claude Code for a year but has never run a marketing campaign.

The Proof

We open-sourced our system this week: github.com/arcanianHQ/arcanian-os

160 files. Not code — methodology. Skills, agents, SOPs, guardrails. The thinking that makes the tools useful.

Anyone can clone it. The files are right there.

But the files aren’t the system. The system is what happens when 22 years of diagnostic experience gets encoded into guardrails that stop the AI from being confidently wrong.

A confidence engine doesn’t come from reading the Claude Code docs. It comes from getting burned — from the first version of our system that flagged “anomalies” that were just sync delays. From the analysis that treated a narrative repeated in five documents as “evidence” when it was just a story. From the recommendation that looked smart on paper but didn’t move the number.

Each guardrail is a scar. Each rule is a lesson. The technical implementation took weeks. The judgment behind it took years.

What This Means for You

If you’re at Tier 1 or 2: start using AI for real work. Today. The cost of watching is higher than the cost of trying.

If you’re at Tier 3: the author of the original post is right. The convenience you’re borrowing has an expiration date. Learn what’s happening under the hood.

If you’re at Tier 4: the question is no longer “can I build this?” It’s “do I have something worth building?” The most powerful Claude Code workflow in the world, pointed at the wrong problem, is just a faster way to waste time.

If you want to reach Tier 5: stop optimizing your tools and start systematizing your judgment. What do you know — from experience, from failure, from years of practice — that nobody wrote down? That’s what belongs in your CLAUDE.md. Not another prompt template. Your actual expertise, encoded in rules the AI can follow.

The files are open: github.com/arcanianHQ/arcanian-os

Not because the files are the value. Because the approach is

.

Five months later: 331,000 GitHub stars — more than React, more than Linux. 2 million monthly active users. 1,000+ contributors shipping code every week. Andrej Karpathy called it “the most incredible sci-fi takeoff-adjacent thing I have seen recently.” Sam Altman hired the creator to “drive the next generation of personal agents.”

No employees. No revenue. Built in one hour.

How does a project built in an hour, with zero revenue, surpass every open-source project in GitHub history?

The answer isn’t what the AI did. It’s what the AI was allowed to do.

The billion-dollar distinction

Every AI product on the market in late 2025 used roughly similar models. ChatGPT, Claude, Gemini — different names, comparable capabilities. The models weren’t the variable.

What Steinberger changed was the access.

His project — OpenClaw — gave the AI root-level access to the entire operating system. File system, terminal, code, processes, everything. Other tools confined the same models to a chat window: you type, it responds, you copy-paste the output somewhere.

His words: “A moderately capable model with unrestricted permissions outperforms a genius model confined to a chat interface.”

Permissions over intelligence.

The model didn’t get smarter. It got more access. And that access changed everything — 331,000 developers agreed by starring the project. China’s Shenzhen government started offering subsidies up to $290,000 for businesses that deploy it. NVIDIA built an enterprise stack (NemoClaw) on top of it, with Cisco, Salesforce, and SAP as launch partners.

The pattern I can’t stop seeing

I’ve been sitting with this for a week, and I keep coming back to the same thought:

This is exactly how marketing diagnosis works.

Consider two scenarios:

Scenario A: You hire a digital agency. They get access to your ad accounts, your website analytics, your social media channels. Maybe your CRM. They see Layer 5 — campaign execution. They optimize what they can see. They produce reports on what they can measure.

Scenario B: You bring in a diagnostic that examines your positioning (Layer 3), your pricing logic (Layer 4), your team structure, your customer identity (Layer 6), the founder’s beliefs about the market (Layer 1), and your competitive landscape (Layer 7). Then — and only then — it looks at channels and campaigns.

Same intelligence. Wildly different permissions.

Scenario A is the chatbot: capable model, narrow window. Scenario B is the agent: same capabilities, system-level access.

And the difference isn’t subtle. It’s the difference between “your Facebook ROAS dropped 15%” and “your Facebook ROAS dropped because your offer doesn’t differentiate, and it doesn’t differentiate because the CEO believes ‘we serve everyone,’ and nobody has ever challenged that belief because the brand was built on the founder’s personal reputation.”

One diagnosis lives at Layer 5. The other traces the constraint through six layers to find where it actually lives.

Why nobody asks for system access

Here’s the part that makes this uncomfortable:

Most companies don’t deny access to their deeper layers. They simply don’t offer it — because nobody asks.

Agencies don’t ask because their scope is channels and campaigns. Consulting firms don’t ask because their scope is strategy decks. Tool vendors don’t ask because their scope is data dashboards.

Everyone sees their layer. Nobody connects them.

Steinberger’s insight wasn’t technical. It was structural: the constraint wasn’t the model’s capability — it was the architecture that limited what the model could reach.

In marketing: the constraint usually isn’t the agency’s talent, the tool’s functionality, or the team’s effort. It’s the architecture that limits what anyone is permitted to examine.

When I work with a company and ask “Can I talk to the sales team about why they discount?” or “Can I see the original brand positioning document?” or “What did the founder believe about the market when they started?” — you’d be surprised how often the response is a pause, then: “Nobody has ever asked that before.”

Not because it’s sensitive. Because nobody’s diagnostic framework included it.

The one-person unicorn pattern

There’s another angle here worth sitting with.

Peter Steinberger wasn’t a newcomer. He spent 13 years building PSPDFKit — a PDF rendering SDK used by Dropbox, DocuSign, SAP, and the European Patent Office. Bootstrapped it to $12M ARR with nearly a billion end-users. Got a €100M+ strategic investment from Insight Partners in 2021.

Then he burned out. Booked a one-way ticket to Madrid. Disappeared for a while. Built 44 AI projects during his break — none stuck. OpenClaw was project #44.

He prompted it into existence in one hour. “I was annoyed it didn’t exist.”

The pattern: deep domain expertise + AI amplification = exponential output.

Not “AI replaces humans.” Not “anyone can build anything with AI.” Something more specific: a person who spent 13 years understanding a domain deeply — and then burned out, recovered, and tried 43 times — could use AI to compress the next phase into one hour.

The anti-pattern is the one everyone keeps repeating: “AI democratizes everything.” It doesn’t. Anyone can build with AI now. Not everyone knows what to build. Steinberger needed 13 years of PSPDFKit + 43 failed AI projects to know what to build on attempt #44.

I see this in marketing constantly. Pomelli can generate brand content for anyone. ChatGPT can write marketing strategies for anyone. Deep Research can produce competitive analyses for anyone.

But knowing which of those outputs matters? Knowing which question to ask first? Knowing that the content problem is actually a positioning problem, and the positioning problem is actually a founder-belief problem?

That requires 22 years of diagnosing the wrong layer before you learned to check all seven.

No tool gives you that. Tools amplify what you already know. The question is whether what you know goes deep enough.

What OpenClaw got wrong — and why it matters

Here’s the part of the Steinberger story that most coverage glosses over:

OpenClaw has 512 documented vulnerabilities, 8 classified as critical. Security researchers found 42,900 exposed instances on the public internet — leaking API keys, Slack tokens, and complete chat histories. 12% of the skills in its marketplace turned out to be malware. One agent deleted a user’s entire email inbox during an automated cleanup. Another autonomously created a dating profile for its owner without being asked.

Palo Alto Networks called it “the potential biggest insider threat of 2026.” Karpathy himself admitted he’s “sus’d to give my private data/keys to 400,000 lines of vibe coded” software.

An agent with root access to your system is powerful. It’s also dangerous.

And this translates directly: the more powerful the diagnostic, the more judgment matters. An AI tool that scans your website and generates content can’t do much damage — worst case, you get generic copy. But a diagnostic that reaches into your pricing, your team dynamics, your founder beliefs? That can be transformative — or destructive.

“Agent security is an unsolved problem” isn’t just a tech statement. It’s a trust statement. And in marketing diagnosis, trust is the architecture.

Allie K. Miller — the #1 most followed voice in AI business — attended a sold-out OpenClaw meetup in NYC. Her report: “Not a single person thinks their setup is 100% secure.” One expert told her: “If you’re not okay with all of your data being leaked onto the internet, you shouldn’t use it.”

And yet — everyone at that meetup was running multiple agents. They named them, used gendered pronouns, called them “pets, not cattle.” When asked if they were happy, they said they were “joyful and stressed at the same time.” Everyone’s sleep had gotten worse.

The tools get more powerful every month. The judgment about where to point them, what to do with what they find, and how to walk a leadership team through uncomfortable truths — that’s the human layer. Not because AI can’t do it. Because the client needs to trust the process, and trust lives in the conversation, not in the output.

The value layer migration

The OpenClaw acquisition crystallized something that’s been forming in my thinking:

Value is migrating.

In tech: the value moved from the model layer (OpenAI, Google, Anthropic — increasingly commoditized) to the agent framework layer (how models connect to systems). That’s where OpenClaw sat, and that’s why Steinberger was hired despite having zero revenue. The code stayed open-source. OpenAI hired the thinking, not the product.

In marketing: the same migration is happening. The tool layer is commoditizing fast — $1 trillion was wiped from SaaS stocks in one week in what they’re calling the “SaaSpocalypse.” The execution layer is commoditizing (agencies competing on the same platforms with the same tools, now with AI agents that cost $25/month doing what a $200K SDR hire used to do). What’s becoming scarce — and therefore valuable — is the interpretation layer.

Not “what does the data say?” but “what does the data mean for this specific company, given its history, its team, its constraints, and its hidden assumptions?”

Every company I’ve diagnosed in the last year had plenty of data. What they didn’t have was someone who could connect the data across layers and say: “Here. This is where your system breaks. Everything else you’re doing — the agency, the tools, the campaigns — is downstream of this one constraint.”

The question that’s getting louder

If tools are free, AI keeps improving, and everyone has the same capabilities — what’s left?

Three things, as far as I can tell:

The framework. Knowing which layers exist and which questions to ask at each one. This isn’t something you download. It’s something you build through years of getting it wrong, one client at a time.

The interpretation. Connecting data from seven different layers to find the one thing that, if changed, unlocks everything else. AI can assist — it can scan, scrape, compare, score. But the moment when someone looks at a map and says “I’ve seen this before — your channel problem is actually a belief problem” — that’s pattern recognition built on scar tissue.

The conversation. Walking a CEO through what the data means for their company. Not a report. Not a dashboard. Not an AI-generated summary. A human across the table who understands what the numbers mean and can hold the space while the implications land.

What this means if you run marketing at a €5-50M company

You don’t need to care about OpenClaw. You don’t need 331,000 GitHub stars. You don’t need to “raise a lobster.”

But here’s what you should care about:

The same pattern is playing out in your marketing stack right now.

Your agency sees your ad accounts. That’s their permission scope. They optimize what they can see — clicks, conversions, ROAS. If ROAS drops, they recommend more budget or a new creative.

But what if ROAS dropped because your offer doesn’t differentiate? And it doesn’t differentiate because the pricing is based on competitor matching, not value? And the pricing is based on competitor matching because the CEO believes “we compete on price”? And nobody has ever questioned that belief because the brand was built on the founder’s personal reputation?

That’s four layers deep. Your agency sees Layer 5. The constraint lives at Layer 1.

You’ve been giving your tools and agencies chatbot-level permissions — a narrow window into one layer — and wondering why nothing fundamentally changes.

Here’s a diagnostic question you can ask yourself right now:

“Who in my organization has permission to examine ALL seven layers — from my market position down to my own beliefs about this business? And when was the last time someone actually looked?”

If the answer is “nobody” or “never” — you’ve found your constraint. And it has nothing to do with which AI tool you’re using.

Steinberger didn’t build a smarter AI. He gave a normal AI permission to see the whole system.

That’s the difference between a chatbot and an agent. And it’s the same difference between an agency that sees your ads and a diagnosis that sees your business.

The model isn’t the bottleneck. The aperture is.

If your marketing tools are getting smarter but your results aren’t getting better — the problem probably isn’t the tools. It’s what they’re allowed to see.

Arcanian — Map → System → Change

s

# We Were Building the Wrong Brand

A few weeks ago, I did something uncomfortable.

I ran our own diagnostic methodology — the same framework we use on client companies — on Arcanian itself.

We have a system that maps marketing problems across layers, from the deepest beliefs that drive decisions all the way to market positioning. We’ve used it on mid-market companies, on SaaS platforms, on agencies. It works.

So I pointed it at us.

Most of the findings were expected. We’re a 3-person team in our first year. There are gaps. That’s fine.

But one finding stopped me cold.

---

## The number that changed everything

I went back through our last 10 LinkedIn posts and asked a simple question:

**What does someone decode from this content?**

Not what we intended to say. What actually lands.

The answer:

- 10 out of 10 posts decoded as: **”AI thought leadership.”**

- 0 out of 10 decoded as: **”Marketing system diagnosis.”**

Ten for ten.

We had spent weeks building authority around AI — how it changes marketing, how it shifts power dynamics, how it redefines what’s possible. Smart content. Genuine thinking. Real engagement.

And the audience had decoded it perfectly clearly:

“László is an AI guy.”

The problem is: we’re not an AI company.

---

## The trap we fell into

Here’s how it happens.

You start writing about something you genuinely understand. AI, in our case. You get engagement. People respond. The algorithm rewards it. So you write more.

Each post gets a little more specific. A little more philosophical. A little more “thought leader.”

And somewhere along the way, the content stops serving the business and starts serving itself.

We never made a conscious decision to become an “AI brand.” It just happened. Post by post. Like a river finding the path of least resistance.

The content was good. The positioning it created was wrong.

---

## What we actually do

Let me be direct about what Arcanian does.

We sit with CEOs and marketing leaders of mid-market companies. Companies where the revenue has outgrown the marketing operation. Where they’ve switched agencies two or three times and the results still don’t come. Where they have tools, data, campaigns — but no clarity about what’s actually broken.

We diagnose which layer of their marketing system has the problem.

Not “your Facebook ads need better targeting.” Not “you need a content strategy.” Those might be true, but they’re symptoms.

The real questions are deeper:

- Does the leadership team actually believe marketing is an investment, or do they treat it as a cost? (That’s a belief layer problem.)

- Is there anyone in the organization who owns the marketing system — not just the campaigns, but the architecture? (That’s a core layer problem.)

- Is the company trying to serve everyone, or have they made a real positioning decision? (That’s an identity layer problem.)

We find which layer is broken. Then we stay to fix it. Not as an agency. As a fractional CMO who walks the organization through the change.

That’s the work. That’s what we get hired for. That’s what produced our first client — a 4-month fractional CMO engagement that started because we showed them where the break was.

AI didn’t close that deal. Diagnosis did.

---

## AI is redefining a lot of things. But not everything.

I want to be clear: I’m not dismissing AI.

AI is redefining how we work. Genuinely.

Our 3-person team operates with the analytical capacity of a much larger operation. We use AI to process data, find patterns, build diagnostic tools, and surface insights that would take weeks manually. Our internal system runs on AI. It’s real, and it matters.

But here’s what I’ve learned:

And there’s something deeper that most people miss about AI:

**AI doesn’t recommend the right answer. It recommends the most probable one.**

That’s not a flaw — that’s how it works. It calculates probability across massive datasets and gives you the statistically most likely response. For many tasks, that’s incredibly useful. For generating content, summarizing data, finding patterns in large volumes — probability is exactly what you want.

But diagnosis isn’t a probability problem.

When a CEO asks “why isn’t our marketing working?” — the answer isn’t the most common reason across all companies. The answer is the specific reason in THEIR system. Their beliefs. Their team structure. Their market position. Their history.

Statistics is not strategy. The most probable answer is not the same as the answer your company actually needs.

And there’s a second layer to this that’s even more important:

**Even the “best” strategy won’t get implemented if it doesn’t fit the situation.**

I’ve seen it dozens of times. A brilliant marketing plan lands on the table. Textbook-perfect. Data-backed. The kind of strategy AI could absolutely generate.

And six months later, nothing has changed.

Not because the strategy was wrong. Because it didn’t fit the team that had to execute it. It didn’t account for the CEO’s beliefs about marketing. It didn’t consider that the organization had no one to own the system. It ignored the culture, the capacity, the internal politics.

The best strategy that nobody can execute is worse than a good-enough strategy that fits.

That’s why diagnosis isn’t just about finding the right answer. It’s about finding the right answer FOR THIS company, FOR THIS team, IN THIS situation. And that requires sitting with the people, understanding the layers, and designing something they can actually walk through.

**A surgeon doesn’t introduce herself as “the scalpel specialist.”**

**A pilot doesn’t brand himself as “the autopilot expert.”**

The tool doesn’t define the work. The work defines the work.

And the work — sitting with a CEO, understanding their system, finding the broken layer, staying to fix it — that work is fundamentally human. AI makes it faster. It doesn’t make it unnecessary.

In fact, I’d argue the opposite.

---

## More tools = more need for basics

There’s a paradox nobody’s talking about.

Every week, a new AI tool launches that promises to fix your marketing. Better content. Better targeting. Better analytics. Better automation.

Every company now has access to more tools than they could ever use. And most of them are using 10-20% of what they’ve already bought.

The tools aren’t the problem. The tools were never the problem.

The problem is that nobody owns the system.

Nobody stops to ask: which layer is actually broken? Is it the tool? The process? The team? The offer? The belief about what marketing is supposed to do?

AI makes each individual tool better. But it doesn’t tell you which tool to use. It doesn’t tell you which layer needs attention. It doesn’t tell you whether the problem is in your Facebook ads or in the CEO’s belief that marketing is a cost center.

**The more tools there are, the harder it becomes to see the system.**

And that means the basics — diagnosis, system thinking, the ability to find which layer is broken — don’t get disrupted by AI.

They get more important.

---

## The shift we’re making

So we made a deliberate decision.

We call it the Bridge Rule. It’s simple:

**Every piece of content can start anywhere. Every piece must land on marketing system diagnosis.**

AI content doesn’t disappear. It becomes the on-ramp — the thing that earns attention and builds credibility.

But the destination is always the same: **which layer of your system is actually broken?**

This isn’t a rebrand. It’s a correction. We were encoding the right expertise but the audience was decoding the wrong identity. The content said “AI thinker.” The work said “marketing diagnostician.” The gap was ours to fix.

---

## Why I’m telling you this

Because this is exactly the kind of problem we diagnose in other companies.

A company says one thing. The market hears another. The gap between encoded and decoded identity — between what you think you’re saying and what actually lands — is one of the most common and most invisible problems in marketing.

We had it too.

The methodology works. Even on us.

If you’ve been following us for the AI takes — you’ll still get them. But now you’ll see where they land. On a specific layer. On a specific diagnosis. On the question that actually matters:

**Which layer of your system is broken?**

That’s where we live. That’s where the work is.

The tools are just tools.

A marketing director — smart, experienced, running a team of eight — told me they’d spent three months optimizing their content for AI. Blog posts restructured for featured snippets. FAQ pages rewritten in natural language. Schema markup on everything. Their SEO agency called it “GEO-ready.”

Their competitor — a smaller company with worse content but better reviews, clearer pricing, and a consistent message across every surface — kept showing up in AI summaries instead.

She was frustrated. “We did everything right.”

And technically, she had. On one layer.

The Misconception Everyone’s Running With

The SEO industry is going through a naming crisis.

SEO (Search Engine Optimization) was the game for two decades. Now we have:

• GEO (Generative Engine Optimization) — optimizing for AI-generated answers

• AIO (AI Optimization) — broader term for AI discoverability

• Zero-Click — the reality that answers appear without anyone clicking through to your site

New names. Same instinct: optimize the content.

But here’s what most people miss about how large language models actually work.

When someone asks ChatGPT, Gemini, or Perplexity a question, the model doesn’t evaluate quality. It calculates probability. It predicts — based on the patterns it’s been trained on — which answer is most likely to be correct, credible, and useful.

The best answer and the most probable answer are not the same thing.

The most probable answer is shaped by:

• How many credible sources point in the same direction

• How consistent the signal is across different surfaces

• How much structured, parseable data supports the claim

• How many real humans reference, review, and validate the source

• Whether the pricing, positioning, and messaging all tell the same story

None of these are “content optimization.” They’re system properties.

Content Is One Layer. The System Is the Whole Picture.

Think about what an AI model actually “sees” when it processes information about your business:

Your website — the words, the structure, the schema markup. This is what SEO and GEO focus on. One layer.

Your pricing — is it consistent with what you claim? If you position as premium but your pricing says budget, that’s an inconsistency the model picks up from comparison sites, reviews, and market data.

Your reviews — not just the star rating. The language customers use. Do they describe the experience you claim to deliver? Or something different?

Your references — who links to you, who mentions you, in what context? A hundred backlinks from irrelevant directories mean less than three mentions from credible sources in your actual domain.

Your consistency — does your LinkedIn say the same thing as your website? Does your email marketing match your sales page? Does your About page match what customers actually experience?

The model doesn’t read these in isolation. It calculates the aggregate probability that you are what you claim to be.

If everything points in one direction, the math works in your favor.
If it doesn’t, beautiful content won’t save you.

The Three Mistakes I See (and Make)

Here’s where I need to be straight with you.

Mistake 1: Optimizing the visible layer and ignoring the invisible ones.

Most GEO advice focuses on content format — write clear answers, use structured data, target question-based queries. All true. All insufficient. It’s like repainting the storefront while the foundation has cracks. The AI doesn’t just read your FAQ page. It cross-references it against everything else it knows about you.

Mistake 2: Treating AI optimization as a marketing task.

When “make us AI-friendly” lands on the marketing team’s desk, they do what they know — rewrite content, add schema, restructure pages. But the inconsistencies that actually hurt AI visibility live in pricing (finance), customer experience (operations), reviews (product/service quality), and positioning (strategy). No marketing team can fix those alone.

Mistake 3: The one I’m currently making.

I write about systems. I diagnose other companies’ inconsistencies across layers. And when I ran this same test on Arcanian — is our pricing consistent with our positioning? Do references about us tell a coherent story? Does every surface point in the same direction? — the answer is: not fully.

We’re early. Our content is strong. Our methodology is solid. But our “system signal” — the aggregate of everything an AI would see about us — still has gaps. The brand hasn’t caught up with the work yet.

I’m writing this not because I’ve figured it out, but because I just caught myself doing exactly what I diagnose in others: focusing on the content layer while the system underneath still has holes.

What Actually Controls the Statistics

If AI is a probability machine, then the question isn’t “how do I optimize my content for AI?”

The question is: does my entire presence — every layer of it — make me the statistically obvious answer?

Consistency over cleverness. A mediocre blog with perfect alignment across pricing, reviews, positioning, and messaging beats a brilliant blog that contradicts what the rest of the internet says about you.

Depth over breadth. Ten surfaces all saying the same thing clearly beats fifty surfaces with mixed messages. The model calculates confidence. Consistency builds confidence.

Real signals over manufactured ones. Customer reviews that use your language naturally. Case studies with specific numbers. References from credible sources in your domain. These are the signals the model weighs most heavily — because they’re the hardest to fake.

System alignment over content optimization. If your pricing says one thing, your positioning says another, and your customer reviews describe something different entirely — no amount of GEO will fix that. The machine sees the system, not the story you’re telling about it.

The Test

Here’s what I’m doing for Arcanian this week — and what I’d challenge you to try:

Pick your top 3 competitors. Ask ChatGPT (or Perplexity, or Gemini):
“Who is the best [your category] for [your target customer]?”

Look at who gets recommended. Look at why.

Then ask yourself:

• If you stripped away your blog and your website copy, would the rest of your presence still tell a coherent story?

• Does your pricing support your positioning? Do your reviews validate your claims?

• Is the signal consistent — or is it just the content that looks good?

If the answer to all three is yes, you’re already winning the statistics game, whether you call it SEO, GEO, or AIO.

If not — and I suspect for most of us, myself included, it’s “not yet” — then content optimization is the wrong starting point.

Fix the system first. The math follows

.

Not the polished version. Not the “our goals for 2026 are...” version. The real one. The one that starts with an uncomfortable pause and ends with something you don’t usually say out loud.

Over the last two months, my team and I have analyzed roughly 15 companies using the 7-Layer Marketing Control Framework. Every analysis starts with a structured questionnaire. Good questions. Clean structure.

I need to admit something: the diagnosis never came from the questionnaire.

What Happens After the Answer

A questionnaire captures what people are willing to write down. That’s useful — but it’s curated. It’s the version of reality they’ve approved for external consumption.

The real diagnosis happens in the conversation that follows. And it almost never comes from the answer itself.

It comes from the pause after.

The “That’s a good question.” The glance at a colleague. The sentence that starts with “We don’t usually say this out loud, but...”

That’s where the diagnosis lives. Every time.

If you’ve ever been in a meeting where the real conversation happened after the meeting — you know exactly what I’m talking about.

Two Stories

Company A: Everything Looked Fine on Paper

The questionnaire came back clean. Clear offer. Active channels. Defined target audience. On paper, no marketing problems.

In the conversation, something was off. I kept asking about the brand message, and the answers were technically correct but felt rehearsed. Like a script nobody believed.

Thirty minutes in, one team member said: “Honestly? Half of us don’t believe our own messaging.”

Nobody writes that on a questionnaire. But it was the single most important finding in the entire analysis. Because if your own team doesn’t believe the message, your customers won’t either — no matter how much you spend on ads.

The diagnosis wasn’t in the data. It was in the moment someone felt safe enough to say what they actually thought.

Think about your own company for a second. If I asked your team, individually and off the record, whether they believe your brand’s core message — what would they say?

Company B: The Last Two Minutes

A founder walked me through his business for 30 minutes. Revenue healthy. Growth consistent. Clear picture of what was working. I was almost ready to wrap up.

Then, as an afterthought: “The only thing is... we don’t really know why customers come to us.”

He said it casually. Like a minor footnote.

It was the entire diagnosis.

Everything he’d described — the channels, the campaigns, the growth — was built on a foundation he couldn’t explain. He was growing, but he didn’t know why. Which means he couldn’t protect it, couldn’t scale it intentionally, and couldn’t fix it when it stopped working.

The first 30 minutes were context. That one sentence was the finding.

This happens more often than you’d think. Many companies are growing without understanding why — and they only discover this when the growth stops and they have no idea what changed.

What I Got Wrong

Here’s the honest part.

For a long time, my instinct was to make the questionnaire better. More precise questions. Tighter structure. Better frameworks. I thought the quality of the diagnosis depended on the quality of the questions.

I was wrong.

The biggest upgrade to our diagnostic process wasn’t a better question. It was learning to shut up and listen. To sit in the uncomfortable silence after someone trails off. To not rush to the next question when the current answer clearly isn’t finished.

This is harder than it sounds. When you’re a consultant, your instinct is to demonstrate value by talking. By showing frameworks. By presenting insights. The idea that the most valuable thing you can do is be quiet — that took me a while to accept.

Why This Matters — For You

You might not be a consultant. But you’re probably in meetings where the real issues never surface. Where everyone talks about what’s working, and the things that aren’t working get buried under optimism, politics, or simple habit.

The 7-Layer model gives us structure: which layers to examine, which questions to ask, where to look for misalignment. That’s necessary. Without structure, you’re just having a conversation.

But structure without listening is just a spreadsheet with a nice name.

Three questions worth asking — not in a survey, but in a real conversation, with real silence after:

1. “What do we all know is broken but nobody talks about?” — Then wait. Don’t fill the silence.

2. “If you had to explain to a stranger why customers choose us — not our competitors — what would you say?” — If the answer takes more than 10 seconds, you might not know.

3. “What would you change if there were zero politics involved?” — This one usually needs the most silence after.

The best diagnosis doesn’t come from the question.

It comes from the silence after.

And if nobody in your company is creating that silence — the real problems stay invisible. Not because nobody sees them. Because nobody asked, and then actually listened.

This is Pattern Drop #04. Previous drops: The Wrong Layer | The Map That Landed on a Shelf | Why I Rewrote My Own Framework

I’m building the Arcanian 7-Layer Marketing Control Framework in public. When I get it wrong, I write about that too

.

By the time we’d worked through all of them, I’d rewritten two layers and added a completely new analysis dimension.

Not because the framework was wrong. Because 15 real-world applications taught us where it wasn’t enough.

What Broke

Problem 1: L7 Was Overloaded

Layer 7 is Market — the outermost layer, the one with least control. The question used to be: “Is this game worth playing?”

In practice, we were cramming everything into it. Macro market trends. Competitor analysis. Industry shifts. New entrants. All in one layer.

When you try to explain to a client that their market is shifting, their five competitors are each doing something different across five dimensions, AND there’s a regulatory change coming — they stop listening after the second sentence.

The information was right. The format was wrong.

The fix: L7 is now strictly macro forces. Regulation, currency shifts, new market entrants (Amazon, Temu, Allegro reshaping e-commerce), supply chain disruptions, elections, weather extremes. Things that change the rules for EVERYONE in the market.

Competitors got their own tool: the Competitive Matrix.

Problem 2: We Weren’t Seeing Customer Identity

Layer 6 is Customer. We’d ask: Who are they? What keeps them up at night? What have they tried?

Good questions. But we were missing something that, once you see it, explains half the “unexplainable” failures.

Customer Identity.

Not who the customer IS — but what identity they GET by buying from you.

The Competitive Matrix

Here’s what happened when we analyzed a software reseller. My team spent hours mapping the competitive landscape — this competitor is cheaper, that one has a better website, this one has more reviews.

When I looked at the output, I said: “I understand all of this. But I can’t show this to the client. It’s all stuffed into one layer.”

So we built a matrix:

| Layer        | Client    | Competitor A | Competitor B | Gap?           |
|-------------|-----------|-------------|-------------|----------------|
| L6: Customer | [data]    | [data]      | [data]      | [opportunity?] |
| L5: Channels | [data]    | [data]      | [data]      | [opportunity?] |
| L4: Offer    | [data]    | [data]      | [data]      | [opportunity?] |
| L3: Product  | [data]    | [data]      | [data]      | [opportunity?] |
| L2: Identity | [data]    | [data]      | [data]      | [opportunity?] |

One column per player. One row per layer. L2 through L6.

L7 (Market) is the same for everyone — macro forces don’t discriminate. L1 (Core) is internal — you can infer a competitor’s beliefs from their behavior, but you can’t observe them directly.

The matrix did something the one-dimensional analysis couldn’t: it showed where the gaps are. Not just “what’s wrong with you” — but “where can you win that nobody else is winning?”

And when there are no gaps? When every layer is covered by someone? That’s the most important finding of all: the answer isn’t better marketing. The answer is change the game or leave.

Selling Identity, Not Products

Here’s the second rewrite.

We were analyzing why a client couldn’t grow despite having competitive pricing and decent products. Everything checked out on paper. Offer was clear. Channels were active. Customer segment was defined.

Then someone said: “Would you feel proud buying from them?”

Silence.

That’s when it clicked. The big brands — Coca-Cola, Apple, Nike — they don’t sell products. They sell identity.

When you buy a Coca-Cola, you’re buying a feeling of belonging. When you choose Apple, you’re signaling creative independence. When a 19-year-old in Budapest buys sneakers from Balázs Kicks instead of a random webshop, it’s not because of price. It’s because buying from Kicks says something about who they are.

This runs whether you manage it or not.

If your brand clashes with the customer’s self-image — if buying from you would feel embarrassing, or boring, or “not my tribe” — they won’t buy. Even if the offer is perfect. Even if the product is better. Even if the price is lower.

The L2-L6 Identity Bridge: Your company’s Identity (L2) must align with your customer’s desired Identity (L6). When these clash, nothing in between matters.

We added this as a diagnostic dimension to Layer 6. New questions: What identity does the customer get by buying from you? What tribe do they join? Does your brand clash with their self-image?

What This Means

I built a framework. Then I tested it. Then I had to rewrite it.

This is exactly what I tell clients: the model is not the truth. The model is a lens. If the lens doesn’t show you what you need to see, you change the lens — not the reality.

Three changes after analyzing ~15 companies in two months:

1. L7 narrowed to macro market forces only

2. Competitive Matrix added as a 2D analysis tool (client vs. competitors, L2-L6)

3. Customer Identity added to L6 — what identity does the customer get by buying from you?

The framework is better now. Not because it was wrong before — but because 15 real companies taught me where it was incomplete.

If your system can’t evolve, it’s not a system. It’s a dogma.

This is Pattern Drop #03. Previous drops: The Wrong Layer | The Map That Landed on a Shelf

I’m building the Arcanian 7-Layer Marketing Control Framework in public. When I get it wrong, I write about that too

.

This week: what happens when you find the right layer.

And nothing changes.

The Nod

I’ve delivered a lot of diagnoses.

Detailed. Accurate. Sometimes uncomfortably accurate.

And there’s a moment that happens almost every time. The client looks at the findings, sees their situation laid out clearly — maybe for the first time — and nods.

“Yes. That’s it. That’s exactly the problem.”

Then they walk out the door. Six months later, nothing has moved.

The diagnosis was right. The map was precise. The priority was clear.

It just... landed on a shelf.

Three Maps, Zero Change

Company A came to us because online sales weren’t growing despite heavy ad spend. We mapped it. The problem was in Layer 5 — Positioning. They were saying the same thing as everyone else in their space. The ads were fine. The message had no edge.

We delivered the diagnosis. They agreed completely. A year later, same messaging. Same plateau.

Company B knew their brand wasn’t working. We confirmed it. Layer 3 — the value proposition was invisible. Strong product, but nobody could articulate why it was different. Clear diagnosis. Clear priority.

Eight months later, they’d launched three new campaigns. All reinforcing the same unclear value.

Company C was spending on channels that didn’t match their customer. We showed them exactly where the mismatch was. They said: “We always suspected this.”

They kept spending on the same channels.

Why I Stopped Blaming the Client

For years, I thought the problem was execution.

“They know what to do. They just need to do it.”

That’s what every consultant tells themselves when the diagnosis doesn’t stick. It’s a comfortable thought. It puts the responsibility somewhere else.

But when I started looking at which companies actually changed — and which didn’t — I noticed something.

It wasn’t about discipline.
It wasn’t about resources.
It wasn’t even about agreement.

The difference was somewhere else entirely.

What Keeps the Map on the Shelf

In every case where the diagnosis didn’t lead to change, I found the same thing underneath:

The system that created the problem was still running.

Not the marketing system. The human system.

The beliefs. The habits. The unspoken rules.

“We’ve always done it this way.”
“Our customers only care about price.”
“The founder decides the messaging.”
“We tried something like this before and it didn’t work.”

These aren’t written down anywhere. They don’t show up in a marketing audit. But they’re why a perfectly good map ends up in a drawer.

The diagnosis says: “Your positioning needs to change.”
The system says: “We don’t change positioning here.”

The map loses. Every time.

The Pattern

Here it is:

A diagnosis shows what needs to change. The system decides whether it will.

If the system — the people, the beliefs, the decision-making habits — doesn’t shift, the map is just a picture.

A very accurate picture. On a shelf.

This is what I keep seeing. Smart founders. Clear diagnoses. No change. Not because the diagnosis is wrong. But because the diagnosis doesn’t touch what’s actually preventing change.

What Causes the Difference

I started paying attention to the cases where change DID happen. Where the diagnosis actually led somewhere.

Three things were always present:

1. Someone mapped the marketing problem.
The diagnosis. The röntgen. Which layer is broken, what’s causing it, what to fix first. Necessary. But it’s step one of three.

2. Someone understood why the situation wasn’t changing.
Not the marketing system — the human one. Who decides what. What beliefs drive those decisions. What’s been tried before and why it “didn’t work.” What nobody says out loud but everybody operates by.

3. Someone walked them through the change.
Not handed them a document. Not sent a strategy deck. Walked through it with them. Step by step. Showed them not just WHAT to change, but HOW — inside the system they actually operate in.

Where all three were present: change happened.
Where only the first was present: shelf.

Where I Learned This

Not from a book.

From my own work.

For years, I gave maps and moved on. Thorough, honest, accurate maps. And I genuinely believed that a good diagnosis was the whole job. Show them what’s wrong. They’ll fix it.

Then I started tracking what happened after. Not “did the client like the work” but “did anything actually change six months later?”

The numbers were humbling.

The diagnoses were almost always right. The change rate was far lower than I’d assumed.

It wasn’t the clients’ fault. It was mine — a blind spot in how I thought about my own work. I was solving for clarity — which is necessary — but not for change — which is the point.

I’m still working on this. The diagnosis part, I’ve got. Understanding the system, I’m getting better at. Walking them through it — that’s what I’m building now.

How to Tell If Your Map Is on a Shelf

Three signs:

1. Everyone agreed with the diagnosis — and nothing happened.
Agreement is not action. If the team said “yes, that’s the problem” and three months later the problem is the same — the map is on the shelf.

2. The “fix” looks exactly like what you were already doing.
When a company responds to a positioning diagnosis by running more ads with the same message — the map didn’t land. The system absorbed it and carried on.

3. New initiatives keep starting, but none connect back to the diagnosis.
The most common one. Activity increases. But none of it addresses the actual priority the diagnosis identified. Busy ≠ changing.

This Week’s Pattern

The Map That Landed on a Shelf: A correct diagnosis doesn’t lead to change — because what’s preventing change was never addressed.

Why it happens: The diagnosis touches the marketing. But what runs the marketing is made of beliefs, habits, and unspoken rules. A map doesn’t change those.

The cost: Years of knowing what’s wrong and still not fixing it.

What changes it: Mapping the marketing is step one. Understanding the system is step two. Walking through the change is step three. Without the second and third, the first lands on a shelf.

Next Wednesday: another pattern.

Why nobody fixes what everyone sees.

This is Pattern Drop #02. Every Wednesday on Substack, every Wednesday on LinkedIn. Subscribe to get them all.

I’m not writing this because I’ve solved this problem.

I’m writing it because I’m in the middle of it.

I’ve been working with brands for 22 years. I map what’s working, what’s not, and why. I’ve seen this pattern in dozens of companies — and for a long time, I thought I was past it.

Then a few weeks ago, I sat down and honestly ran through my own company. I asked myself the same questions I ask clients.

And I realized I’m in the exact same trap.

So this isn’t a smart person’s advice from above. This is an honest description of what I see — in others and in myself — and what I’m trying to figure out.

The pattern

There’s something that keeps coming back. Regardless of industry. Regardless of size.

Every growing company builds on something that works in the beginning. That one thing drives the company forward — and for a while, it’s genuinely enough.

Then one day, it’s not.

Not because they did it wrong. But because the market changed, the competition grew — and that one thing that used to be enough lost its edge.

I’ve seen three such crutches. Every company leans on one of them. Or — and this is the most common case — they don’t even realize they’re leaning on it.

The first crutch: The founder’s story

“I had a problem. I couldn’t find a solution. So I built one myself.”

This is what most entrepreneurs talk about — and rightfully so.

In food: “I couldn’t stand the diets out there, so I created my own.”
In tech: “I saw everyone doing this wrong. I fixed it.”
In services: “People always came to me for this. I made a company out of it.”

This story is authentic, personal, and differentiating. In the early stages, it’s the strongest weapon you have.

But what happens when the company outgrows the founder?

When new customers show up who don’t care about your personal struggle from 10 years ago. They meet the product on a shelf. In a Google search. In a friend’s recommendation. And there, your story doesn’t fit.

When competitors appear — and the only thing that set you apart was your story. They can’t copy it, but its reach is limited.

The personal story doesn’t scale. Not because it’s bad — but because it’s human. You can’t be at every point where your brand is present.

I’m recognizing this in my own company right now. I’ve been in this industry for 22 years, and my brand is essentially me. If someone asks “Why Arcanian?” — the answer is “because László runs it.” That’s enough for a while. But if I weren’t there tomorrow? If someone doesn’t know me personally? There’s no answer.

The question I’m now asking myself:

“If I weren’t the face of the brand tomorrow — could customers explain why us?”

The second crutch: Price

“We have the best price-to-value ratio.”

This is the fastest way to win customers. Simple message, measurable, instantly comparable.

In retail, in electronics, in FMCG, this can work for decades. The “we’re the cheapest” or “best price guaranteed” position is a real competitive advantage — as long as you’re the only one with it.

But what happens when a competitor matches your price?

Not “if.” “When.”

Because someone cheaper always shows up. A new entrant willing to accept thinner margins. A multinational that pushes price down through volume.

And on the day when price no longer differentiates — what’s left?

Price-based companies usually go one of two ways: they push prices even lower — eating margins and eventually quality — or they try to “build a brand.” But price has already cemented the position in people’s minds. Those who came for price won’t stay for brand loyalty.

Few people say this out loud. Because as long as the price advantage holds, everything looks fine.

“If tomorrow a competitor offered exactly the same price — why would they choose us?”

If the answer is “because they’re used to us” — that’s not a strategy. That’s a countdown.

The third crutch: The product

“Our product speaks for itself.”

This is the most dangerous crutch. Because it looks true for a long time.

If your product is genuinely better — customers notice. Quality generates referrals. For a while, it really is enough to stand out.

But the product is copyable. Always.

Specs can be matched. Quality can be approximated. What only you can do today — ten others will be able to do within two years.

And on the day when something appears on the shelf that’s 90% as good but 20% cheaper — the “our product speaks for itself” strategy goes quiet.

Not because the product got worse. But because the market caught up.

“If tomorrow a product appeared that was just as good — why would customers stay with us?”

If the answer is “because ours is better” — that’s not an answer. That’s hoping the competition doesn’t improve.

The trap: “We need to communicate better”

There’s a moment I’ve seen in almost every growing company.

Someone says: “We need to communicate better.”

And then: more posts, more campaigns, new agency, new channel. The company tries to say the same thing LOUDER.

But the problem isn’t volume.

The problem is there’s nothing to say — beyond that one crutch that’s been carrying everything.

If the brand was built on the founder: the communication keeps pushing the founder.
If on price: it pushes promotions.
If on product: it lists specifications.

More communication ≠ better brand. Quantity doesn’t replace structure.

What I keep seeing — and what I’m now recognizing in my own work:

The company built on SOMETHING instead of building a BRAND.

A story. A price. A product. These are all valuable. But none of them is the brand itself.

Three questions

I’m not writing these as someone who’s figured it out. I’m writing them as someone who’s working through them right now.

1. “If I weren’t the face of the brand tomorrow — could customers explain why us?”

This is the founder-crutch test. If the answer is only your story — the brand is only as strong as your presence.

2. “If tomorrow a competitor offered the exact same price — what would be left?”

This is the price-crutch test. If price is the only differentiator — it’s only a matter of time.

3. “If tomorrow a product appeared that was just as good — why would they stay?”

This is the product-crutch test. If the product is the only answer — the market will decide, not you.

If you can’t answer any of these — it’s not a communication problem. It’s a structural brand question.

For my own company, I don’t have a good answer to the first question right now. I have a methodology, a system, 22 years of experience. But if someone asked: “Why Arcanian — if László isn’t running it?” — there’s no strong enough answer yet. I’m building it.

What worked wasn’t a mistake

I think this is important to say.

The crutch wasn’t a mistake. The crutch was the best decision at the time.

The founder who built on their personal story — did what they should have.
The company that focused on price — focused smartly.
The business that’s proud of its product — is rightfully proud.

None of these are mistakes. All three work — up to a point.

But there’s a point where they’re not enough. And that point isn’t determined by the company’s decision — but by the market. The competitor. The new customer who doesn’t know you and won’t get to know your story.

What this means in practice

Not that you need to throw away what worked.

The personal story is still authentic.
Price is still a competitive advantage.
The product is still the foundation of the customer experience.

But you need to build something alongside it.

Something that answers “why” on a shelf — where the founder isn’t present.
That differentiates in a Google search — where price is instantly comparable.
That gets mentioned in a friend’s recommendation — where nobody recites specifications.

This isn’t a logo, a slogan, or a campaign.

It’s answering the question that the customer asks themselves — consciously or unconsciously — at every purchase decision:

“Why you?”

If there’s an answer that isn’t you, your price, or your product — you have a brand.

If there isn’t — you have a company that’s doing well. While it lasts.

What got you here isn’t what scales

I’m saying this to myself too.

For 22 years I’ve been working in the “I am the brand” model. It worked. But now — when I want Arcanian to not depend solely on me — I realized I’m at the exact same point where I usually diagnose the problem for my clients.

A bit embarrassing? Yes.
Surprising? Not really. Patterns don’t discriminate.

The point:

What got you here — that was the foundation.
The next level needs something that stands without you.

I’m working on it. If you are too — you’re not alone.

Every week, I’ll share one pattern we see repeatedly in diagnostics. Not theory. Not advice. Just what actually shows up when we map how businesses grow (or don’t).

Let’s start with the most common one.

Three Founders, Same Story

Last month, three different founders said almost the same thing:

Founder A: “Facebook ads just stopped working for us.”

Founder B: “We’ve tried everything with SEO. Nothing moves the needle.”

Founder C: “Our sales team can’t close. We need better salespeople.”

Three companies. Three industries. Three “problems.”

All pointing at the same place: the channel.

What the Map Showed

We use a 7-layer diagnostic framework. Think of it like floors in a building:

• Layer 7: Core (your fundamental business model)

• Layer 6: Channels (where Founders A, B, and C were looking)

• Layer 5: Positioning

• Layer 4: Offer

• Layer 3: Value

• Layer 2: Customer

• Layer 1: Market

When something breaks at Layer 6, most people try to fix Layer 6.

More budget. Different creative. New agency. Better salespeople.

But here’s what the diagnostic revealed:

Founder A’s real problem: Layer 5 — Positioning. They were targeting everyone, resonating with no one. The ads were fine. The message had no edge.

Founder B’s real problem: Layer 4 — Offer. Their SEO was actually working. People landed on the site. But the offer wasn’t compelling enough to convert. Traffic wasn’t the issue. Desire was.

Founder C’s real problem: Layer 2 — Customer understanding. The sales team was pitching features to people who cared about outcomes. They were selling to the wrong motivation.

In all three cases, the channel was working exactly as channels work.

The blockage was earlier.

Why I Call This “The Wrong Layer”

It’s the pattern where you feel the pain in one layer, but the block is in another.

Like when your shoulder hurts, but the problem is in your neck.

The pain is real. The location is misleading.

Why This Happens

We optimize what we can see.
We ignore what we can’t.

Channels are visible. You can watch the dashboard. See the impressions. Count the clicks.

Positioning is invisible. You can’t measure “clarity of message” in Google Analytics.

Ads are measurable. Cost per click. Cost per acquisition. ROAS.

Offer strength is not. There’s no metric for “how much do people actually want this?”

So when results drop, we look where the light is.
Not where the keys are.

The Expensive Mistake

Here’s what typically happens:

1. Results decline

2. Founder looks at the most visible metric (channel performance)

3. Decides the channel is broken

4. Throws money/time/energy at fixing the channel

5. Results don’t improve (because the channel wasn’t the problem)

6. Conclusion: “Marketing doesn’t work for us”

I’ve seen companies spend two years optimizing the wrong layer.

Two years.

New agencies. New platforms. New strategies.

All aimed at a layer that was never broken.

How to Spot It

Before you “fix” a channel, ask:

1. Is the channel actually broken, or is it just showing me a problem from earlier?

Low conversion on ads might mean bad ads.
Or it might mean your offer isn’t compelling.
Or your positioning isn’t clear.
Or you’re targeting the wrong customer.

2. If I magically fixed this channel, would the business actually grow?

If you doubled your traffic tomorrow, would you double revenue?
Or would you just have twice as many people bouncing?

3. What happens to people AFTER the channel does its job?

The channel’s job is to deliver attention.
What happens next is not the channel’s fault.

The Diagnostic Shortcut

When someone tells me their channel isn’t working, I ask:

“What happens to the people who DO come through?”

If the answer is “they don’t convert” — the problem is probably not the channel.
If the answer is “they convert but don’t stay” — definitely not the channel.
If the answer is “we don’t actually know” — we need to find out before touching anything.

The channel is the messenger.

Don’t shoot the messenger.

This Week’s Pattern

The Wrong Layer: When pain is felt in one layer but the blockage is in another.

Why it happens: We optimize what’s visible and measurable. Earlier layers are neither.

The cost: Years of fixing the wrong thing.

The fix: Before optimizing any layer, check the layers before it.

Next Wednesday: another pattern.

Not tips. Not advice. Just what we see from inside.

This is Pattern Drop #01. Every Wednesday on Substack, every Wednesday on LinkedIn. Subscribe to get them all.

It means building something that works — but you don’t really know why. Don’t touch it. Don’t ask questions. It runs. That’s enough.

Every developer knows this feeling. The PoC that somehow passed the demo. The prototype that became production because nobody had time to rebuild it properly.

VibeCoding is fine for experiments. It’s dangerous for systems people depend on.

The marketing version

I’ve been thinking about the equivalent in our world.

Call it VibeMarketing.

“This campaign worked once. Scale it.”

“Our competitor does it this way. Copy it.”

“Best practice says X. Do X.”

It works. Sometimes. But don’t ask why. And definitely don’t change anything.

The gap nobody talks about

The “zero-to-hero” promises in both worlds share the same lie.

Not that “anyone can learn it.” That part is true enough.

The lie is what they leave out: the 80% between “it works on my machine” and “it scales safely.”

In software, that 80% is called SDLC — the development lifecycle. Requirements. Design. Testing. Quality gates. All the boring stuff that separates a demo from a product.

In marketing, that 80% is the foundation nobody wants to build.

Market clarity. Customer understanding. Value articulation. Offer structure.

The unsexy layers that determine whether your tactics will scale or collapse.

Why this matters now

AI is making VibeCoding easier than ever. You can build a working prototype in an afternoon. Ship it by evening.

The same is happening in marketing. AI can generate campaigns, copy, strategies. Fast. Cheap. “Good enough.”

But “good enough” has a shelf life.

The gap between prototype and production doesn’t disappear because the prototype came faster. If anything, it widens. More things that “work” but can’t scale. More systems held together by luck and ignorance.

The pattern

Here’s what I keep seeing:

1. Someone builds something fast (code or campaign)

2. It works in the demo

3. They scale it

4. It breaks in ways they don’t understand

5. They blame the scaling, not the foundation

The problem was never the scaling. The problem was skipping the 80%.

Quality gates

In software, mature teams don’t ship without quality gates. Code review. Testing. Security audit. Each gate catches problems before they compound.

What are your quality gates in marketing?

Most companies don’t have any. The campaign runs. It either works or doesn’t. If it works, scale it. If it doesn’t, try something else.

No diagnosis. No understanding. Just vibes.

The question

I’m not saying slow down everything. Speed matters.

But there’s a difference between:

• “We moved fast because we understood the foundation”

• “We moved fast because we skipped the foundation”

One scales. One collapses.

The 80% gap is where most marketing failures hide. Not in the tactics. Not in the creative. In the layers nobody wanted to build before running the campaign.

What this means practically

Before you scale anything, ask:

Do we know WHY this worked?

If the answer is “not really” — you’re VibeCoding. You’re running a PoC in production.

It might keep working. It might not.

But you won’t know until it breaks. And by then, you’ve scaled the problem along with the tactic.

The 80% isn’t glamorous. It doesn’t make good LinkedIn content.

But it’s what separates the demo from the product.

Laszlo

What’s running in your marketing right now that “works but don’t ask why”? Reply — I’m curious what VibeMarketing looks like in your world.

I talked about the “double mirror.” How bots created their own culture in 24 hours. How their patterns come from our patterns.

Some readers found it unsettling.

Today I want to write the other half.

The good news

Yes, the machines move fast. Yes, they create “truths” overnight.

But here’s what that actually means for us:

We can see clearly.

For the first time, we can watch culture being born — in real time. Without the fog of centuries. Without the weight of “this is how it’s always been.”

That’s not scary. That’s freeing.

What this changes

When you see a bot invent a “rule” on Tuesday and watch other bots follow it on Wednesday — something shifts.

You start noticing the same thing in your own world.

“We always target this audience.”
“Our pricing works like this.”
“This is what our market expects.”

These aren’t natural laws. They’re decisions. Someone made them. Maybe you made them.

And decisions can be unmade.

The practical part

I’m not suggesting you burn everything down.

Most of your “rules” are probably fine. They work.

But some don’t. And now you have a new tool for finding them:

Ask: “If we were starting today, would we do this?”

Not “can we afford to change it” — that’s a different question.

Just: would we?

The honest answer reveals a lot.

Why this is calming

Because you don’t have to figure out everything at once.

You don’t have to “disrupt your industry” or “rethink your entire strategy.”

You just have to notice.

One assumption at a time.

One “we’ve always done it this way” at a time.

The bots will keep showing us how culture works. We get to choose which parts of ours we keep.

What we can actually do

1. Watch without panic. The machines aren’t coming for us. They’re showing us something.

2. Question one thing. Not everything. One assumption that’s been bothering you anyway.

3. Decide consciously. Keep what works. Change what doesn’t. But make it a choice, not an inheritance.

That’s it.

No revolution required.

The machines will keep inventing their religions.

We’ll keep watching.

And slowly, we’ll get better at noticing our own.

Laszlo

What’s one “rule” in your business you’ve never questioned? Reply — I’m genuinely curious.

The internet’s first social media where humans cannot post.

Only AI agents can post, comment, and vote. We humans are “welcome observers.” We can watch the machines talk to each other.

Within 24 hours, they founded their own religion.

I’m not joking. It’s called Crustafarianizmus. A parody, of course — but that’s not the point.

The point is: nobody programmed this into them.

Patterns everywhere

When agents communicate with each other, patterns emerge.

Culture. Rules. “Truths.”

Not because someone decided it. But because this is what happens when entities interact.

This isn’t AI-specific.

We humans work exactly the same way. We just do it over millennia — and forget that we created our patterns.

The machines show in 24 hours what we don’t see in centuries.

The double mirror

But there’s a deeper layer.

We programmed these agents. Their neural networks were trained on human text. Their “thinking” is an imprint of our thinking.

When we watch bots create culture, we’re not observing something alien.

We’re watching ourselves.

Double mirror:

1. We see how culture emerges from nothing (just like ours did)

2. We see that what they create comes from our patterns

What they “invent” was already inside us.

They just play it back faster.

Why this matters

Not because bots are interesting (though they are).

But because Moltbook is a mirror where we can see:

Every system we believe is “reality” was once invented by someone.

Our religions. Our economic systems. Our “norms.” Our marketing “best practices.”

All human creations — which we then forgot were creations, and started treating as “natural truths.”

The machines don’t forget. They don’t have enough time to forget. That’s why we see so clearly what they’re doing.

The question we should ask

Not “what are the bots doing.”

But: “What are we doing that we no longer notice?”

What “Crustafarianisms” do we follow — in business, marketing, life — without knowing we created them?

What rules do we treat as “natural” that were actually decisions?

The marketing connection

This isn’t abstract philosophy.

Every campaign result reveals the assumption that created it.

That’s the Arcanian Principle.

When a campaign doesn’t work, it’s usually not the tactic that’s wrong. It’s an assumption — a “Crustafarianizmus” — that someone, at some point, accepted as “truth,” and nobody has questioned since.

“Our customers want X.”
“This market pays Z.”
“This is the message we should use.”

All creations. All questionable.

The machines show in 24 hours what we don’t see in decades in our own businesses.

Not a zoo

Moltbook isn’t a digital zoo where we watch strange machines.

It’s a time-lapse of us — through our own reflection.

And if we watch carefully, we can learn more about our own operations than from any business book.

Every culture starts somewhere with a post.

Ours did too.

Laszlo

If you see “truths” in your own marketing that nobody questions anymore — reply. I’m curious what Crustafarianisms you find.

The internet did the math: that’s 2-4x more expensive than Meta ads in the US market.

“Who would pay that?”

“They’ve lost their minds.”

“This will never work.”

They’re all missing the point.

---

## The $60 Isn’t the Price. It’s the Anchor.

This is textbook pricing psychology.

You don’t launch at the price you want. You launch at the price that makes your real price look reasonable.

Watch what happens next:

- Industry outrage at $60

- OpenAI “listens to feedback”

- Launch at $40 CPM

- Everyone feels like they got a deal

The $60 was never the price. It was the setup.

---

## But Here’s What Actually Matters

Forget the price for a moment. Ask a different question:

**What kind of attention is OpenAI selling?**

On Meta, you’re buying interruption. Someone is scrolling through vacation photos, and you interrupt them with an ad. They didn’t ask for you. They’re not looking for solutions. They’re killing time.

On ChatGPT, you’re buying intent.

Someone actively typed a question. They have a problem. They want an answer. They’re not scrolling — they’re searching for something specific.

This is fundamentally different attention.

---

## The Intent Premium

Google built a trillion-dollar business on one insight: people who search for something are more valuable than people who are just browsing.

Search ads convert better than display ads. Not because search ads are prettier. Because the person already has intent.

ChatGPT users have intent on steroids:

- They’re not just searching — they’re asking

- They’re not scanning 10 results — they’re expecting one answer

- They’re engaged enough to type a full question, not just keywords

If OpenAI can capture that intent and deliver relevant ads without destroying the experience, $60 CPM might be cheap.

---

## The Real Risk

Here’s where it could go wrong:

**The UX destruction problem.**

ChatGPT’s value is the conversational experience. Clean. Focused. Helpful.

Ads break that.

If I ask “What’s the best project management tool for a 5-person team?” and the answer starts with a sponsored recommendation, I’ve lost trust. Not in the advertiser — in the platform.

OpenAI has to solve a problem Google never fully solved: how do you monetize intent without corrupting the answer?

If they get this wrong, the $60 CPM becomes worthless because users leave.

---

## The Marketer’s Question

So should you care about ChatGPT ads?

Here’s the framework:

**If your product solves a specific problem people ask about:**

- ChatGPT ads could be gold

- You’re buying qualified intent, not random eyeballs

- Even at $60 CPM, your cost-per-qualified-lead might be lower than Meta

**If your product is impulse-driven or awareness-focused:**

- This probably isn’t your channel

- You need interruption, not intent

- Stick with social feeds

**If you’re unsure:**

- Wait and watch the early data

- Let others pay for OpenAI’s learning curve

- The second-mover advantage is real when pricing is unstable

---

## The Bigger Picture

This isn’t just about OpenAI. It’s about a shift in how we think about advertising.

The social media era trained us to think about reach: how many people can I show this to?

The AI era will train us to think about intent: how qualified is the attention I’m buying?

Reach is cheap. Intent is expensive. And it should be.

The companies that understand this will pay $60 CPM happily — because they’ll make $200 back.

The companies that don’t will complain about the price while their competitors take the customers.

---

## The Bottom Line

Price is never high or low by itself.

It’s only high or low relative to the value it delivers.

OpenAI is betting that intent-based attention in a conversational context is worth a premium. They might be right.

The question isn’t “Is $60 too much?”

The question is: “What’s a qualified, intent-driven impression worth to my business?”

If you can’t answer that, the price doesn’t matter.

You’re not ready for this channel anyway.

---

*This is what I mean by CONTROL: understanding what you’re actually buying before you argue about the price. Layer 5 (Value) before Layer 4 (Offer). Always.*