Getting started with AI: Practical examples

You can probably get a powerful model — tools like Claude, ChatGPT or Copilot — and give it a template and say, how should we do this?

That's how I would start.

The model will probably give you something very close to the solution. You can get to 60–70 per cent just like that, with the right data and a clean template.

Interestingly, a lot of construction teams don't always have templates. So the bonus of doing these AI workflows is that you can develop them, and also streamline or standardise your work as you go.

AI might actually be the first time we do standardisation in this industry. We always say we want to do standardisation, but how do you actually action that?

AI can be that persistent force that means we will finally do it.

I’d also start with a coding agent. Something like Claude Code or GitHub Copilot, which for me is one of the best tools for engineering right now. It’s around $30 a month, which is usually quite affordable.

It might be a challenge at first to convince your manager or colleagues that it will give you tens of thousands of value if you use it. But it will.

Any coding agent you pick, that would be my simple answer, because these are the best knowledge work tools in the world right now. You can do anything with a coding agent: run your proposals, make presentations, reply to your email, anything. That's where I would start. That's what I say, even internally, to people.

You should develop data generation processes that are correct at the source. We shouldn’t have a future of unstructured data. I think that would be a failure.

But we don’t need to wait and clean our past to do AI. Not at all.

Most of the workflows I do internally, they all start with: 'Here’s my SharePoint folder of the problem, this is what it is.' It’s all a bit structured and messy. But the models and the experts around them can solve it.

The difference in quality between a messy knowledge base and one transformed into a completely structured object is immense.

For example, if you're trying to show me all projects in the last decade in transportation where we had this problem, the exhaustive answer only exists on the structured side.

The other side only gives you a limited view: 'Oh, I found a few of them, here they are.'

I do expect some of the larger firms to transform in the next few years, or at least to organically create smaller counterparts. But smaller companies can move faster, they can test and experiment.

I often tongue-in-cheek say that this is actually the first digital transformation that we will see that actually works.

Think of everything that’s gone before it — BIM, Digital Twins, the Internet of Things, even computational design. They changed the landscape of where we try to solve problems, but in many ways we just shifted things from one place to the other without actually solving them.

There’s a big temptation to think it’s all about efficiency, and that's good. But you should also think about how it actually transforms completely what you're doing.

That might mean new solutions, or actually rebuilding what we have from the point of view of this technology. I do think AI is actually the transformer in technology that will change everything.