AI & Society
Building AI Products People Actually Need
Adding AI to a product is easy. Making it something people would miss is the hard part.
There is a pattern I keep seeing right now. A product team adds an AI feature. A little sparkle icon appears. The launch gets announced. And six months later, almost nobody uses it.
The feature was not bad. It just was not needed.
Start where product always starts
I have spent most of my career in product, from UGO to Receptiv to BrightLine to Talisai to Kbean. The fundamentals have not changed once. Who is this for? What are they trying to do? Where do they get stuck? What would they pay for, in money or attention, to get unstuck?
AI does not change those questions. It changes some of the answers, because some problems that were too expensive to solve are now cheap. That is exciting. It is also a trap, because cheap solutions tempt you to stop asking whether the problem was real.
Watch people, not demos
AI demos are impressive. They are also staged. The real test is watching an actual user try to get something done, with messy inputs, on a bad day, in a hurry.
That is where you find out whether the output is good enough, whether people trust it, whether they understand what it did, and whether it saves them time or just moves the work to checking.
Be honest about limitations
Every AI feature is wrong sometimes. Good products are designed around that fact. They show uncertainty. They make it easy to correct the system. They do not pretend to be more reliable than they are.
Users forgive mistakes. They do not forgive being misled about how often mistakes happen.
Measure the right thing
Usage of the AI feature is not the goal. The goal is whatever the user came to do. Did they finish faster? Did they make a better decision? Did they come back? If the AI feature is popular but the outcome did not improve, you built a toy.
[Jaime: one example from your product career, ideally Talisai or Inssa, where research changed what you built.]
The quiet test
Here is a question worth asking about any AI feature. If we removed it tomorrow, who would complain, and what would they say?
If you cannot answer that clearly, you have more discovery to do.
I advise teams on AI product strategy. If you are trying to separate what is useful from what is just new, let's talk.
