Three questions to ask before any AI build

Before we build anything with AI, we ask three questions:

1. Why are we building this?
2. Who are we building it for?
3. What problem does it solve for them?

They sound basic. They're also the questions most AI projects skip.

On the Product Insider podcast, I put it this way to Dr. Nancy Li: if your product doesn't have a solid why, you're probably not going to have a solid product. A solid why makes a solid product. And these days, the AI needs to know that why too.

1. Why are we building this?

This is the question that gets skipped most often, because it feels like it's already been answered. Somebody decided to build the thing, so surely somebody knows why.

Usually, what they know is the trigger, not the reason. "A competitor launched an AI feature." "The board asked about our AI strategy." "We have budget left this quarter." Those are reasons a project started. They aren't reasons it should exist.

A real why points at an outcome for the business: keep the customers you have, cut the hours your team loses to a manual task, open a market you can't reach today. If you can't connect the build to an outcome like that, stop and find it before you go further.

2. Who are we building it for?

"Our customers" isn't an answer. "Small businesses" isn't either.

The who should be specific enough that you could picture one real person using the thing on a normal Tuesday. What job do they have? What are they trying to get done when they open it? What do they already use instead?

The who matters because it decides almost every trade-off later. Speed or depth. Simple or powerful. Automatic or reviewed by a person. When the who is vague, every one of those decisions turns into an argument, and every argument gets settled by whoever is loudest instead of by the user.

3. What problem does it solve for them?

This is where the first two questions meet. You know why the business wants it and who it's for. Now: what does that person stop struggling with once it exists?

I like to test this with a fill-in-the-blank sentence: "We're building ______ for ______ so they can stop ______." If the last blank is hard to fill in, that's the problem to solve first. I wrote about that test in why most AI projects fail before anyone writes code.

Answering this one honestly sometimes changes the whole project. One client, a health and wellness company, came to us asking for AI to bring in new customers. When we asked what problem their customers actually had, the answer was that they didn't feel heard when something went wrong. That turned an acquisition project into a churn project, and the first fix didn't need AI at all. The full story is in Stop starting with AI.

Why AI makes these questions more important, not less

There's a common misconception that AI works differently from the rest of your team, as if you could skip the briefing and it would figure things out.

It doesn't. AI is a capable but naive assistant. It builds exactly what you describe and misses everything you didn't. A good engineer might stop and ask who the feature is for. An AI agent will usually just build it, quickly and confidently, for whoever it imagines.

So you treat it like you'd treat anyone else on the project: you give it the why. When the why, the who and the problem are written down clearly, every person and every agent on the project works toward the same goal. That's how I direct the AI agents that build my own products, and it's the single biggest difference between AI projects that ship and ones that stall.

Which question is hardest for your team?

In my experience, teams find the third question hardest. The first two feel like strategy. The third one requires you to talk to the people you're building for and hear something you might not want to hear.

If you can answer all three for your next project, write them down and put them at the top of every brief, for people and for AI. If you can't, that's not a failure. It's the most valuable thing you'll learn before you spend any money.

Not sure how to answer all three for your idea? That's exactly what the free consultation is for. Book a free consultation and we'll work through them together, or start with the AI readiness quiz.