Find the Gap Before the Solution

Yesterday a university friend messaged me about a research project. She had already worked out the problem and a solution, but what she really wanted to ask about was the agentic part. The more we talked, the more I felt she did not really know how agents work, or what they actually do.

And that is fair. This is all very new, very experimental, and nobody fully knows yet what can and cannot be done. It depends a lot on the person doing it. So I am not blaming anyone for the confusion. But it pushed me toward something more important, which is that we have to be careful about the idea first, long before we reach for a solution.

For a research project you need a clear research gap. Everyone wants to talk about the solution, and almost nobody can point at the gap, and that is exactly where the trouble starts. We tell ourselves the supervisor will accept the idea, but the supervisor is not looking for a solution. They are looking for the gap. A clear research gap, and one you can actually communicate. This is the gap, and here are the many ways we could close it.

An agentic solution is a solution, not the research case. Those are two different things. If you are building a new agentic architecture itself, that is different, because then you can go find the current agentic researchers, read their gaps, and fill one. That is real research. But wiring an agent onto a problem that already has five ordinary solutions is not a gap, it is just a tool choice.

A lot of people are struggling with exactly this right now, and I struggled with it too when I first tried to extract the research gap. But once it clicked, it really clicked. I could see the gap, I could see the novelty, and I could see what we could actually do about it. The agentic approach was only one of the options. I had more than five ways to solve the same problem without any agentic workflow at all. Find the gap first, and the solution stops being the hard part.