AI in Transport

The UK’s Department for Transport invited me to attend the Transport AI 2025 event, and here are my 7 key takeaways:

1) Data is the key - but data is poor

Transport data is unstandardised, siloed within teams, held by different suppliers, and may require a commercial license to use, among other issues. There was a consensus that the data is not sufficient to provide the insights the industry desires.

2) Main barriers: data and skills

The main barriers are data, data, data, and then a lack of AI analysis skills within pubic transport bodies. The rest is relatively easy to solve.

3) What is AI in transportation, anyway?

That question wasn't answered, and many solutions presented were either using LLM (ChatGPT, etc.) on top of existing solutions or presenting algorithms/machine learning (at best) as AI. Few were AI Agents (what is).

4) Work from the customer backward!

There was a lot of talk about data, governance, and what technology enables - but very little discussion about what transport bodies actually need and are willing to pay for, i.e., the solution to a customer pain point. For now, the buzz around the benefits of AI overshadows customer needs.

5) Big money in infrastructure?

Spending on road maintenance is in the billions £ annually. Every 0.1% saving is meaningful. But nobody was sure how AI could solve that challenge, and "AI won't go and fix potholes."

6) Show me the money!

To create a thriving AI startup ecosystem, funding is needed. To create pilot projects that show the ROI to local councils (future clients), who cannot afford AI in its current form, funding is needed. Funding needs to come from a government innovation pot. Heidi Alexander that’s on you.

7) We should start with good enough.

Perfect is our enemy. We should start with what we have and improve as we go.