At Herman Scheer, the thing that surprised me wasn't how little AI the team was using. It was how good they were at everything else. This is a Los Angeles branding agency whose work sits on shelves in every supermarket in America. Craft was not the problem. Their clients had started asking whether they had AI capability, and the honest answer was no — no shared method, no quality bar for AI-assisted work, nothing they could sell.
So we trained the whole team, not two enthusiasts. Before the training was finished they were putting client-ready image, copy, and video into live CPG launches. The estimated year-one production savings came to $4.5 million.
Here's the part worth writing down. Almost nothing I taught in that room was new information. Most of it was permission to stop doing four things the team had been doing correctly for fifteen years.
The unlearning problem
Every's Context Window recently ran an "unlearning" series from three Maven instructors, arguing that cheap AI prototyping forces practitioners to drop habits that used to be professionalism: waiting for certainty, holding untested opinions, asking permission, treating execution as expertise. Their audience is mostly product and technical people.
The creative version is sharper, because our habits aren't just workflow. They're identity. A creative director's authority comes from having taste before the work exists. A producer's comes from protecting a schedule. A designer's comes from being the one whose hands can actually make the thing. Every one of those is a habit built for a world where making something took a week. When making something takes twenty minutes, the habit stops protecting quality and starts blocking it.
That's why training two people never works. Two people can learn a tool. Only a whole team can change a reflex, because the reflex lives between people — in what gets scheduled, what gets approved, what gets respected.
Four habits, in the order they usually surface.
1. Waiting until you're sure before you start
The old sequence was research, align, brief, then make, because making was the expensive part. You didn't want to spend a week building the wrong thing, so you spent two weeks making sure.
Now making is the cheap part. The most expensive thing in a creative process is a two-week alignment phase producing a document nobody can see. I watch teams protect that phase hardest, because it feels rigorous. It isn't rigorous anymore. It's just slow.
The replacement: start before you're sure, and use the artifact to get sure. Not "here's my hypothesis about the campaign platform." Three visual directions on screen, made that afternoon, so the conversation happens about something real. Being wrong on Monday afternoon costs nothing now. Being uncertain until Friday costs the week.
2. Having opinions you could have tested
This is the one senior people find hardest, and I say that with sympathy — I do it too.
Someone says "that tone won't land with this audience." Someone else disagrees. The room debates it for forty minutes, and the most senior voice wins, which is usually the right outcome under the old physics. Under the new physics it's absurd. That question was answerable in fifteen minutes: generate both versions, put them in front of a well-constructed audience persona, look at the difference.
Seniority stops meaning "my opinion is the tiebreaker." It starts meaning "I know which questions are worth testing and I can read the result." That's a real skill and a harder one. But the meeting where three experienced people trade instincts about something they could have checked is the most expensive hour in an agency.
3. Asking permission to try something
Every team I train has a version of this: someone quietly wanting to run a campaign through a new workflow, waiting for a moment when it feels sanctioned. That moment never arrives, because nobody's job description includes approving experiments. So the enthusiast does it on a personal account, off the books, and their learning never reaches anyone else.
You fix this with structure, not encouragement. Telling people to "be curious" does nothing. Giving them a standing ten-minute slot every Monday where somebody demos one thing they tried last week does a lot, because it converts permission into an expectation. If nothing shows up in that slot for three weeks running, that's your real adoption data, and it's more honest than any survey.
4. Treating execution speed as your expertise
The hardest one. For most creative careers, the thing that made you valuable was that you could execute — faster, cleaner, better than the person next to you. That was the identity.
Across the engagements I've run, production time drops by about 90% on the work we rebuild. If your professional worth is measured in execution hours, that number reads as a threat, and people respond to threats by hiding. This is exactly why adoption stalls in the middle of an org and not at the top or bottom: middle-weight people have the most execution identity to lose.
The honest reframe isn't "AI frees you for strategy." Everyone can smell that. It's narrower: your expertise moves from making the thing to specifying and judging the thing. Knowing what to ask for. Knowing, in four seconds, why the output is wrong and what specifically to change. That judgment is worth more than execution ever was, and it's genuinely rarer. But it has to be said out loud, by leadership, in terms of how people will be evaluated — otherwise the team hears "you're being made faster so fewer of you are needed" and behaves accordingly.
What actually installs
None of these four unlearn from a memo. They unlearn when the same room watches the old habit fail and the new one work, on real client work, with something at stake.
That's the whole reason our work is a half-day Audit Workshop and then an 8-week Operating Model rather than a course. The audit finds which of the four is actually costing you — usually one dominates. The eight weeks are how long it takes for a new reflex to survive a bad week and a difficult client.
At Herman Scheer, the capability existed inside the team the entire time. What changed was what the team no longer felt obligated to do first.