A few years ago, when clients asked us to add AI to their product, what they usually meant was a search bar that used embeddings, or a field that auto-generated a summary. These are fine features. But they are not what is happening in the best products right now.
What is happening now is that the AI does not just respond, it acts. It takes data from one place, makes a decision, and produces an output without a person clicking anything. That is what people mean by agentic AI, and it is a completely different thing to build.
We have built several of these in production. Here is what we actually learned.
The difference is not technical. It is architectural.
The mistake most teams make is treating AI as a layer they add after the product is built. The product exists, it works, and then someone says can we add AI to this part. Sometimes that works. But often the data model was not designed for it, the user flows do not make sense with it, and the result is something that feels bolted on — because it was.
When AI is going to be a core part of how a product works, you need to design for it from week one. Where does the data come from? What decisions will the agent make? What happens when it is wrong — who reviews it, how do they override it? These questions need answers before you write a line of code.
The products where AI feels natural are the ones where it was planned from the beginning, not added at the end. The difference is visible the moment you use them.
A real example from a healthcare client
One of our clients was a healthcare company with six hospitals. Their scheduling team was spending around 14 hours every week just resolving conflicts — who is available, which room, which regulation applies, which request has priority.
