Question 1: What problem, exactly?

Most AI conversations start with the solution. The vendor explains what their software does. The senior leadership team nods, sees the use case, and signs.

The better starting point is the problem. Not "improve productivity": that's too abstract to be useful. The specific kind: "Our quote turnaround is 11 days. Competitors are sending in 36 hours. We're losing tenders to slower decision speed." Or: "Our QS spends 14 hours a week chasing missing documents. That's £20k a year of senior time on filing."

If you can't write the problem in one sentence with a number attached, you don't know what you're buying.

Question 2: Who's going to use it?

The other half of failed AI deployments comes from this question being skipped. Software arrives, but no role in the business is responsible for using it.

Every AI tool needs a specific person who'll fold it into their week. Sometimes that's the QS. Sometimes the bid manager. Sometimes the most senior admin. The point is naming the person before signing, not after.

A useful test: if you can't name three people who'll use the tool in week one, and roughly which tasks it replaces, you're not buying a system. You're buying an option to buy a system.

The third question (the trap)

Vendors usually want you to ask a third question: "What can your software do?" Resist this.

Their list of features will always sound impressive. It tells you nothing about whether their software solves a problem you actually have, or whether anyone in your business will use it.

Buy software because it solves a problem someone in your business already has. Not because it has impressive features you might use someday.

How to know you've answered both

Before signing, write down three things: the specific problem (with a number), the specific person who'll use it, and what their week looks like differently after week four.

If any of those is fuzzy, don't buy yet. Either the diagnosis isn't sharp enough, or this isn't the right tool.

That's not over-engineering. It's the minimum honest standard for putting money into AI in a construction or built-environment business.