Start With Better Questions

We’ve spent quite a lot of time talking about risk assessment.

But some of the most useful applications of AI in safety happen before you get anywhere near that point.

Think about the early stages of a project. You might know where you’re going, broadly what you’re planning to do and roughly how you intend to do it. But there may still be dozens of things you haven’t considered yet.

Do we need permissions? Are there unusual local rules? Does the location create access or emergency-response problems? Are we going to need specialist contractors or particular competencies? Is there something about what we’re doing that needs a much longer lead time than we’ve allowed? Is there an obvious clash between what we want to do, and where or when we’re planning to do it?

Those aren’t necessarily “risks” that go into a risk assessment. They’re questions that can change the planning. And this is an area where AI can be extremely useful. Give it enough context, and it can interrogate a project from several different angles, research unfamiliar issues, identify possible dependencies and draw your attention to things that may need resolving before detailed planning begins.

Crucially, AI doesn’t need to make the decision. Its value is in identifying things that may need further investigation – while you can still change the plan, get the right people involved or allow time to resolve them.

That’s a very different use of AI from asking it to produce a finished risk assessment. And it’s the thinking behind SCOUT, which we developed specifically for early-stage project planning.

SCOUT isn’t an AI risk assessment generator. Instead, we use AI selectively, where its particular strengths genuinely add value: interrogating the information you’ve provided, exploring a project from different angles, researching unfamiliar issues and identifying things that may need further attention.

That makes SCOUT useful at precisely the stage when plans are still flexible, options are still open and finding something important can save considerable time, cost and difficulty later. It doesn’t replace competent safety advice, and it isn’t trying to. Its job is to help you ask better questions before the answers become expensive.

And that’s really the wider lesson from this course. AI is extraordinarily good at producing answers. Using it well isn’t about handing it more responsibility simply because it can produce something impressive. It’s about deciding where its capabilities genuinely make the process better – and where human judgement needs to remain firmly in control.