Most AI design transformations are theater.

The deck always looks great

I have watched a lot of AI transformations get announced. The slides are beautiful. There is a roadmap with arrows. There is a number on the wall about hours saved. Everyone claps. Then six months later the team is working the same way it always did, and the tool nobody asked for is gathering dust in a tab nobody opens.

That is theater. It looks like change. It is not change.

I have spent 15 years leading and doing design, and I am still hands-on in Figma, in the copy, in the front-end. I have run the real version of this inside a multi-billion-dollar eCommerce org, and I have watched the fake version up close more times than I can count. The difference between the two is not the tools. The tools are the easy part. They have always been the easy part.

The hard part is getting a team to actually change how it works, and keep working that way long after the announcement energy is gone.

Why the tools are not the problem

Buying AI is simple. You pick a vendor, you pay the invoice, you get logins. A smart team can wire up a model and a workflow in a week. None of that is where transformations fail.

Transformations fail at the part nobody puts on a slide. The designer who has a way of doing things that works fine and does not want to relearn it. The reviewer who does not trust the output and quietly redoes it by hand. The manager who says the right words in the all-hands and then measures the team on the same things as before. That is the real terrain. Friction is where transformations go to die.

If you treat AI as a software rollout, you get a software rollout. A login, a training video, a Slack channel that goes quiet. If you treat it as a change in how people work, you have to deal with people. That is slower, messier, and the only thing that actually sticks.

Start with the work people already hate

Here is the mistake almost everyone makes. They point AI at the interesting work first. The creative work. The stuff people are proud of and protective of. Then they are surprised when the team pushes back.

Do the opposite. Find the work people already hate. The repetitive, soul-draining, copy-paste, do-it-again-every-week work. That is where AI is welcome, because you are not threatening anyone. You are rescuing them.

Inside that multi-billion-dollar org, I built an AI agent that took a recurring workflow from around 320 hours down to about 1 hour. That was not a clever demo. It was a job everyone dreaded, done over and over, and the agent simply made it disappear. Nobody fought me on that one. They thanked me.

That is the wedge. When the first thing AI does is delete a chore, people stop seeing it as a threat and start asking what else it can take off their plate. I wrote more about the specific build inside a multi-billion-dollar design org if you want the longer version.

Build AI into the path of least resistance

People do not adopt tools because the tools are good. They adopt tools because the tool is the easiest way to get the next thing done. If using AI means opening a separate app, copying context over, remembering a prompt, and pasting the result back, it will lose. Every time. The old way wins because the old way is already in their hands.

So you have to make the AI path shorter than the manual path. Put it inside the tool they already live in. Pre-load the context so they are not feeding it from scratch. Make the good prompt the default, not something they have to recall. The goal is that doing it with AI is simply less work than doing it without, in the literal moment of the task.

When I think about how to build a team around this, the AI is not a side quest. It is a member of the workflow. At Story Genie, the company I founded, agents write and illustrate personalized hardcover children’s books, and humans supervise and approve quality. The agent is not bolted on next to the work. It is in the path. The work flows through it by default, and a person catches it before it ships.

I go deeper on the org side of this in how to build a design team, because the structure you choose decides whether AI is in the path or off to the side.

Be clear about where AI is trusted and where it is not

The fastest way to kill trust in an AI workflow is to be fuzzy about where the AI is allowed to make the call. If people do not know the boundary, they assume the worst. They assume it is coming for the parts they care about, and they will protect those parts by refusing to engage.

So draw the line out loud. Say plainly where AI is trusted to run, where it drafts and a human approves, and where it does not touch the work at all. Clear standards are not red tape. They are what let people relax and actually use the thing.

At Story Genie the line is bright. Agents do the writing and the illustration. Humans supervise and approve the quality. And the customer sees the entire finished book before they pay a cent, in about 60 seconds. That last part matters. The standard is not we trust the AI. The standard is the customer judges the real output before any money changes hands. The machine does the volume. The human owns the bar. Nobody is confused about who is responsible for quality.

The same logic works inside a team. A designer needs to know that AI can churn out twenty layout starts but the taste call is still theirs. Once that is clear, the AI stops feeling like a replacement and starts feeling like leverage. Which is the honest version of what it is.

Make the early wins loud

Quiet wins do not change a culture. If the AI agent saved someone three days last week and nobody heard about it, it did not happen, as far as the organization is concerned. Theater is loud about plans. Real transformation has to be loud about results.

So broadcast the small, true wins. Name the person. Name the hours. Show the before and the after. Not a projection on a slide, an actual thing that happened to an actual teammate. That is what pulls the skeptics in. People do not move because of a strategy memo. They move because the person at the next desk just got their Friday afternoon back.

I have run this loop in a few places. At The Mysterious Package Company we shipped a full Shopify rebuild and ran weekly A/B tests. At Supply Drop we ran monthly homepage A/B tests on a predictive subscription model. At Kinjo we built a shared design system across two iOS learning apps. In every case, the thing that built momentum was not the framework. It was showing a real result fast, out loud, and letting the next person want in.

How to read the first ninety days

The first ninety days tell you everything, but only if you watch the right signals. Most leaders watch the wrong ones. They count announcements, logos on a vendor slide, and hours saved in a deck. Those numbers feel good in a review and tell you almost nothing about whether the change took hold. I have built this on real teams. The signals that matter are quiet, and they show up in how people work when no one is asking them to.

Here is the simplest test I know. Real adoption looks like a habit, not an event. Theater looks like an event with no habit behind it. If the only proof you have is a launch email and a training session, you have an event. If people are reaching for the tool on a Tuesday afternoon because it is the fastest way to get their work done, you have a habit. Spend your ninety days hunting for the habit.

The strongest leading indicator is unprompted use. Watch for people using the agent without a manager telling them to. Nobody assigns it. Nobody reminds them. They just open it because it saves them an hour. The second indicator is even better. People start asking for the next workflow. They come to you and say can we do the same thing for this other awful task. That pull is gold. It means they trust the first win enough to want more, and they are now doing your roadmap for you.

The third indicator sounds strange but it is the truest one. The AI becomes boring. It stops being a demo and becomes plumbing. Nobody talks about it in the all-hands anymore because it is just how the work gets done now. When I built an agent that took a workflow from about three hundred and twenty hours down to about one hour, the real proof was not the first reveal. It was the week three months later when the team forgot the old way existed. Invisible is the goal. Magic that needs a spotlight is still a magic trick.

Now the vanity signals to ignore. Announcements are not adoption. A press post, a town hall slide, a shiny internal wiki page. These measure intent, not behavior. Logos on a tooling map measure what you bought, not what people use. And the hours saved slide is the most seductive lie of all, because the number is usually a guess multiplied by headcount. If the hours were really saved, you would see them somewhere. You would see a queue get shorter or a deadline move up. If you cannot point to where the saved time went, it was never saved.

So measure it honestly. Pick three or four people who do the painful work and ask them one question every two weeks. Did you use it this week, and would you be annoyed if I took it away. That second half is the whole game. If losing the tool would annoy them, it is real. If they shrug, it is theater, no matter what the dashboard says. Track unprompted opens if you can, track how many new workflow requests come in from the team, and watch whether anyone has stopped talking about the AI because it just works. Three signals, checked every two weeks, will tell you more than any quarterly readout.

Theater is loud about the plan. Real transformation is loud about the receipts.

The objections you will hear, and what I say back

It will replace us. No. It replaces the part of the job nobody wanted in the first place. The taste, the judgment, the knowing-why, the standing behind the work. That stays with people. I believe AI will not replace designers, and I have put real money and my own time behind that belief by building a company on it.

If you want the full case, I made it in AI will not replace designers. The short version is that the bottleneck was never making more pictures. It was deciding which ones are right. That is human work, and AI makes more of it possible, not less.

We tried this and it did not stick. Almost always because it was rolled out as a tool, not a way of working. The login went out, the standards never got drawn, the wins stayed quiet, and the path of least resistance was still the old path. Fix those four things and it sticks. Skip them and it is theater every time.

Our work is too high-stakes to trust a machine. Good. Then do not trust the machine with the call. Trust it with the volume and keep the call with a person. That is exactly the Story Genie model. High stakes is an argument for a clear line, not for doing nothing.

I help teams work through these objections directly. That is what my AI design consulting is really about. Not picking software. Getting people to change how they work and stay changed.

What real transformation actually feels like

Here is the tell. In a theater transformation, the AI is a topic. It is in the meeting title and the quarterly update and the proud LinkedIn post. In a real one, the AI is boring. It is just how the work gets done now. Nobody mentions it because mentioning it would be like mentioning that you used a keyboard.

You get there by being relentless about adoption, not announcements. Start with the work people hate. Put the AI in the path of least resistance. Draw clear lines about trust. Make the early wins loud enough that the skeptics lean in on their own. Then do it again with the next workflow, and the next.

None of that fits cleanly on a slide. That is exactly why it works. The deck is the easy part. The team is the hard part, and the team is the whole point.

If you are trying to do this for real, start with the people and the path, not the platform. I wrote a companion piece on how to build a design team in the AI era that goes further on the structure. But the heart of it is simple. Stop staging the transformation. Go make one boring chore disappear, out loud, and let the rest follow.