The AI you’re putting into your products and services has to earn back what it costs, and the people who have to use it decide whether it does. They’ll use it. They have to. The question is whether it helps them do the work or gets worked around, and that decides whether the money comes back.
When people choose a product and it doesn’t work, they leave. When they’re required to use it and it doesn’t work, they stay and find another way. “That’s the way we’ve always done it” is what a workaround sounds like from the inside, and it’s the first thing I hear when I sit down with the people doing the work.
Every workaround, every “let me try one more thing,” every apology is a person quietly stretching themselves across a hole in the process so the customer doesn’t fall in. The metrics look fine because human beings are the duct tape. “Nobody ever asked me” is what they say when someone finally does.
More surveys, more dashboards, another score. Those describe the problem and move nothing. “It’s good enough” is what gets said when nobody can say what good looks like, how they would measure it, or what to ask the people using the product to find out. I’d rather sit next to five of those people and watch them do the job.
The deck was good, and the first time anyone did their actual job with the thing was after launch, when fixing it cost the most. Proof is humbler than that. A few real people doing their real work while someone watches, before the money is spent.
The research says the same. Firms shown how other firms mapped AI into their work earned 1.9 times the revenue of firms left to find it themselves.
Kim, Kim and Koning, Mapping AI into Production. Harvard Business School working paper, March 2026. A randomized trial with 515 high-growth startups.
Bring up to six people and one of your biggest problems. Anyone can generate an AI concept in a minute now. Knowing which one is worth building takes the people who’ll have to live with it. You leave with one direction the team agrees on and a straight answer on whether to spend money on it. I share fees when we talk.
It started with my dad. He installed manufacturing computer systems for a living. We weren’t rich, but we had a personal computer at home when most of the neighborhood kids didn’t. It was hard to use, and I’ve been at it ever since.
Since then I’ve worked on the systems people have to use to do their jobs, mostly at the moment an old one is being replaced and how the work gets done is still open to change. I work on what has to change in and around the product so it works for the people using it and, more importantly, returns the investment.
One issue a week, written to be used. Every issue ships with the Fieldwork, something free and editable you run with your team the same day. It’s where I think out loud. Expect opinions.
Email me a sentence about the problem and I’ll come back with dates. Or just tell me what’s on your mind. Either works.
aaron@kennellydesign.co