A producer at Cash App asked me this in week two, about twenty minutes after the room had watched a 3D artist rebuild a campaign asset in the time it takes to get coffee. She waited for the demo to end, and then asked the thing everyone had been circling: "Do we have to tell people this was made with AI?"
The room went quiet in a specific way. Not confused. Uncomfortable. Because everyone in it already had a private answer, and the private answers didn't match.
That team went on to cut production time to roughly 10% of what it had been, with an estimated $3.5 million in year-one savings and about 30% more output. But the question that afternoon had nothing to do with any of that. It was about what the disclosure would imply. And I've now heard some version of it in every engagement I've run, which tells me it isn't a policy gap. It's an unresolved belief sitting in the middle of the team.
The norm is unsettled, so people default to hiding
Every's Context Window ran a piece on August 12 about a small version of this problem: whether it's acceptable to talk to an AI out loud around other people. One of their writers, Arielle Shipper, said she types instead of using voice around colleagues because doing it in the open "feels slightly socially embarrassing". The behavior is the interesting part. The norm hasn't settled, so the default is to do it quietly.
Creative teams are running the same play at a much higher price. Nobody has told them what the standard is, so each person invents one. The senior art director discloses nothing. The junior designer discloses everything, nervously, on a client call. The producer avoids the topic in the SOW. Three positions inside one team, and the client eventually meets all three.
The question under the question
Here's what the disclosure question actually is: if I say this was made with AI, does the client think it's worth less?
That's the real fear, and it's worth taking seriously instead of talking people out of it. Because for a lot of AI-assisted work floating around right now, the answer is yes, and correctly so. If the output is generic, the disclosure isn't the problem. The work is.
So I don't start with a disclosure policy. I start with the standard. Once a team has a quality bar they can actually defend — a review process, a brand system the model is working inside, a named person who holds the line on craft — the disclosure question mostly dissolves. You are not confessing to a shortcut. You are describing a method.
Three different things get called "disclosure"
Most of the argument in the room happens because people are debating three separate things at once.
One: the contractual kind. Rights, licensing, talent likeness, music, training-data provenance, anything going into a paid media buy. This isn't a philosophical question and it isn't optional. It lives in the SOW and in your legal review, and if you don't know your position, that's the first thing to fix this quarter. Nothing else on this list matters if that's loose.
Two: process transparency. How the work got made. This should be ordinary, the same way you'd say which retoucher you used or that a shoot was done in-camera. It's a description of method, not an admission.
Three: the apology. "We should mention some of this was AI." Nobody asks for this one, and it's the one teams accidentally do. It signals that the team believes they cut a corner, and the client will believe them. This is the version worth eliminating entirely.
Separate those three and the conversation gets much shorter.
What I tell teams to actually do
Decide it once, at the org level. Not per project, not per person. An individual quietly deciding their own disclosure posture on a client call is the outcome you're trying to avoid, and it's the outcome you get by default when leadership stays silent.
Say it in the pitch, not in the delivery. Timing carries almost all of the meaning. Raised while you're selling the work, it's capability. Raised while you're handing over the work, it sounds like something you were hoping not to mention.
Lead with the standard, not the tool list. Clients don't care which model you used, and the answer expires in a quarter anyway. What holds up is the process: here's our quality bar, here's what a human decides, here's where review happens, here's what we won't do. That's a durable answer. "We use Midjourney" is not.
Make sure the most junior person has the same answer as the CD. This is the whole-team part, and it's the part that gets skipped. If you train two enthusiasts and leave everyone else out, you have not just created a skills gap. You've created a story gap, and the client hears the inconsistency long before they see the work.
Hiding costs more than telling
Three costs, in the order they show up.
The team's usage goes underground. If disclosure feels risky, so does experimenting, so the good work happens on personal accounts and never turns into shared method. Nothing compounds.
You forfeit the pitch. At Herman Scheer, what pushed the engagement forward was clients asking whether the agency had AI capability. That question is now being asked in rooms you're not in. A team with a real answer wins those; a team hoping it doesn't come up loses them quietly and never finds out why.
And the client finds out anyway. They always do. The difference between "they told us how they work" and "we figured it out" is the entire relationship.
The disclosure question only feels dangerous while the work is secretly worse. Fix the standard first, out loud, across the whole team. Then the answer to "do we have to tell them?" turns into a much better question: why wouldn't we lead with it?