From that comes suggestions, taking that idea of, “I’ve seen this happening before, and I think I know what’s going to happen here.” Then the AI can say “you could intervene now. Which of these three options would you be most comfortable with?”
Then it starts the learning journey. The AI can think, “OK, the fleet manager always does that in this scenario.” So, next time, it says: “I’ve made this change. Hope that’s OK. Let me know if not.” You’ve then got the ‘noise’ being dealt with by this agentic community.
There’s a long way to go to make that real. Data quality is probably the most significant thing. If AI starts making decisions based on bad quality data, things can be very, very disappointing.
We’ve put data quality metrics in place in our system. We can lift the lid on those things that you just accept, because you pull all the data out of the system you don’t trust, you put it in a spreadsheet, you tweak a few numbers to make it feel better, and then you send the report out.
The data quality lens, for us, is really saying “if you improve in these areas, this will go away, this will get better.”
It’s like doing the garden, you’ve got to stay on top of it. In transport, like in many organisations, there are so many systems with so many different versions of the same data. You’ve got to choose which is your master, which is the one system that’s going to have your data, and make that the defined system in the organisation.
That way, all the other systems are just recipients of quality data, rather than having five systems forever arguing in the background as to which one’s the real version of the truth.
First, we show customers how we see their data, what quality we see, rating it from a level zero to level four. Level zero being a long way to go. Level four being agentic AI ready. It’s important that we help customers get to that level. Otherwise, the agentic things we do won’t help them.
What is the biggest way AI has improved operations for fleets so far?
Video telematics. Image recognition, object detection, that’s driven advanced driver assistance systems (ADAS) in cars and vans. That’s driven autonomous vehicles and has driven a lot of safety improvement.
Otherwise, you had a lot of people looking at videos going, “oh, what was that? I don’t know.” Having the ability to do classification, object detection really helps bring the safety agenda to the fore.
What are the risks of using AI in fleet operations?
The risk of it not being used is the first one. It’s like anything in life, if you choose not to use it, you will be left behind. You have to be curious, you have to be interested in what it can do.
Next on the list is the maturity of it. If you take it for granted, then you’re probably going to get sold down the river at some point, by some snake oil that isn’t really what it purports to be. Stick with it, don’t be closed to it, but don’t be immediately sold by what it’s saying or doing.
We tend to use more than one source. Look on Copilot, get Claude to have a look for you. Maybe you get Gemini to do something, and just make sure that you’re getting similar, if not the same outputs.
Fleet technology is a market with a huge number of players. What sets Microlise apart?
It is the fact that we run both portfolios, fleet and transport. We’ve got the context of the asset, its health, its utilisation, the way it’s driven. But we’ve also got the context of what it should be doing. Why is it there? Where is it going? When did it get there? Did it get there on time?
Our portfolio is about blending different strands of data, and then we create some unique opportunities around risk, maintenance, and on-time deliveries. We build our picture and we create these opportunities to say, “this risk occurred, but we knew that risk was going to occur, because these other things have happened, not necessarily with that driver or that location, or even with your fleet.”
We’ve got this cross-cutting visibility. We can anonymously use data from the underlying assets. We can see things that are going wrong in different places, and we can help customers contextualise performance.
Cybersecurity has been a hot topic for a while now, with a few high-profile incidents last year. How does Microlise protect customer data?
We generally fit a device to the vehicle, so we make sure that all our hardware is certified, approved and meets the industry standards for cybersecurity. There’s the Cyber Resilience Act that comes in next year, and the Radio Equipment Directive that was updated last year. For the latter, we had to go through a significant certification, so there’s a big piece about securing the asset and making sure that the connectivity from the asset is as bulletproof as we can possibly make it.
Our biggest annual spend increase year-on-year is probably cyber security. The teams, the data centres, the training. Of course, with AI, you’re in a position where your employees might be putting sensitive data into ChatGPT or Copilot. As well as training and education, we’ve got tooling that can make sure that doesn’t happen.
GDPR, data privacy, the EU Data Act that we support, the EU AI Act, that we are going to be supporting later in the year – we’ve got to stay on top of it all, work with the experts, and do everything we can to protect our customers’ data and our own.
It’s about protecting everybody involved in the journey, including the partners that we integrate with. It’s a big, big cost, but a very worthy one. You’re not in business if you can’t stay secure.
