What Are We Trying to Make Better?

“Which option is best?” sounds like a straightforward question. Until someone asks what best means.

Lowest cost? Fastest result? Least disruption? Something that will still work a year from now?

Those could point to different answers. I’m interested in what happens when we ask AI to recommend something before we’ve worked out which of those things matters most.

Consider a hypothetical team choosing where to spend a limited budget. It has three possible projects: replace an unreliable tool, automate a repetitive task or build something new for customers. Each has a reasonable case behind it.

A request to rank those projects leaves an important choice open. Is the team trying to reduce the chance of disruption, give people time back or create something it can sell? Even a thoughtful comparison would need some basis for deciding between them.

That basis deserves attention before the ranking does.

Suppose the team’s immediate priority is reducing disruption, and it has six months and two people available. The question becomes more specific: which project could reduce that disruption within those limits, and what would the team have to give up to pursue it?

There is still plenty to investigate. How unreliable is the existing tool? Would replacing it introduce a different risk? Could a smaller repair be enough? The clearer question gives those details a purpose. They help test a choice rather than simply making the comparison longer.

It also leaves room to challenge the starting point. Perhaps replacing the tool looks attractive because everyone is tired of it, while the disruption actually comes from an unclear process. A useful answer would help expose that possibility.

That’s where I think the judgment sits: deciding what problem deserves attention, what improvement would count and which trade-offs are acceptable. AI can contribute to examining the options. The people responsible still have to decide whether the criteria fit the situation.

I’d want to carry that distinction into any recommendation. What is this answer helping us choose? Which assumptions does it depend on? What would make us change our minds?

The answer might still be uncertain. It might point toward a small next step rather than a complete solution. If it helps us see what needs to be decided, that could be enough to move the work forward.

Before asking which option is best, I’d want us to be able to explain what we’re trying to make better.