AI Build Lab

You've done the training. Something still isn't working.

Here's the Hard or Wrong test, the one my keynote audiences use on their careers: when something keeps not working, ask whether it's hard, which means keep going, or wrong, which means stop and change the approach. Now apply it to your AI training. If you've run the workshops twice and behavior still hasn't changed, it isn't hard. It's wrong.

You know the feeling. The workshop evaluations come back positive. People said they liked it. Then two weeks pass and nothing is different. The same three people are driving everything. Everyone else is waiting to be told what to do next.

Here's the part most training providers won't say out loud: the content is rarely the problem. The format is. Most corporate AI training asks people to sit, watch, and remember. That works for awareness. It does not produce behavior change.

Gartner found that without immediate practice, up to 70 percent of learned skills are never applied on the job. Companies spent $37 billion on AI in 2025, triple the year before, yet only 28 percent of employees know how to use the tools their company already paid for. If your team is part of that 72 percent, the issue isn't budget. It's behavior.

This is not a software training.

There is no enterprise software in this room. Nobody is getting certified. Nobody is sitting through slides about prompt engineering.

Every person builds something useful during the session. A personal app. A packing list for travel. A reminder system for the kids. A schedule that actually protects their time. They choose it. They build it. They walk out with it on their phone.

We start personal on purpose. Personal stakes remove professional anxiety. Nobody is worried about building the wrong thing for their company when the first prompt is about what they never want to forget on vacation. Once someone has built something for themselves, the professional application becomes obvious. They make that transfer in the room without being told to.

The app is not the point. The point is what happens when someone who has spent twenty years consuming technology looks at a screen and realizes they just created something. They laugh. They show the person next to them. They start asking what else they could build. That move, from user to builder, does not reverse.

Every build passes the test first.

Before teams build, they run the Hard or Wrong test on what they're about to build. Is this tool solving a real problem, or an interesting one? Is the manual process hard, or wrong?

Every build lands in one of three categories: lead generation, cost reduction, or revenue increase. Not because categories are magic, but because a tool tied to an outcome survives contact with Monday. A tool built for fun doesn't.

And the session teaches the skill most programs skip entirely: when not to trust the model. BCG and Harvard found employees using AI outside its capability were 19 percent less likely to produce correct results than people who didn't use AI at all. Knowing when to override the output is the difference between a team that uses AI and a team that gets used by it.

Terry has designed and led build sessions and corporate trainings for teams at L'Oreal, General Assembly, and EY, and teaches at NYU.

• You've done the workshops and you're still waiting for something to change.
• You're rolling out AI tools and watching your people ignore them.
• You have a few self-starters and a lot of people waiting to be told what to do.
• You're planning an offsite and you want people to leave different, not just informed.
• You want your team thinking like builders instead of waiting like passengers.