5 things I’ve learned working with AI, autonomous vehicles and smart road infrastructure

Over the past couple of years, I’ve had the opportunity to work on some super interesting projects at the edge of AI, transport systems, and infra. From running trials with autonomous trucks and roadside sensors, to integrating digital twins with vehicle data, to building perception pipelines using models like YOLO — I’ve seen firsthand how AI is starting to shape systems around us.

Some reflections I have in this space:

1. Safety-critical AI raises the bar — as it should.

When AI decisions affect real-world outcomes (especially on roads), the scrutiny rightly increases. It's not just about accuracy — it’s about being able to test, explain, and trust those decisions. Given the limited opportunities for real-world testing, esp. with large vehicles, the need for better simulation and controlled test environments is only growing.

2. Smart infrastructure is an underrated player in autonomy.

We often focus on the vehicle, but roads are getting smarter too. In our trials, we extended a truck’s perception range using roadside cameras and fed that data back to the vehicle in real time. When vehicles and infrastructure can "see" and communicate with each other, the system as a whole becomes more powerful — and safer.

3. Automation doesn't remove the need for human skill — it changes it.

There’s a thought out there that automation means humans can relax. In reality, it shifts what we need to focus on. Just like pilots now train more in simulators as automation takes over in the cockpit, we’ll need to actively maintain and hone skills, not let them atrophy (like astronauts running in zeroG). It’s something I think we’re only just beginning to understand.

4. AI is going to be everywhere — so we need to get the foundations right.

As systems become smart and connected, humans will start stepping out of the loop. That’s exciting, but means transparency is more critical — not just what a model decided, but how and why. Right now, we still don’t have great answers for tracing how data and model training lead to specific decisions. We’ll need to mature that fast.

5. We’re just getting started, and that’s a good thing.

AI still has plenty of rough edges — but its potential is enormous. What matters now is how we build: deliberately, ethically, and with the bigger picture in mind.

Bonus: AI is also changing everyday workflows — sometimes quietly, sometimes completely.

Not all AI comes wrapped in a big project. Some of the most useful tools simply make your day easier: summarising notes, finding patterns in data, or drafting content faster. Other times, it forces you to rethink your whole process from the ground up. Either way, I think we’ll all need to get better at spotting when AI is a helper — and when it’s a sign the whole system needs rethinking.

—

Keen to hear what others are seeing too:

What’s one way AI has already changed how you work — even in small ways?