How Ad-MOTO Uses AI to Make Advertising Data Accessible
Earlier this month, I had the privilege of speaking at the AI Innovation event at Exeter Science Park, sharing how we at Exe Squared have been using artificial intelligence to solve real problems in the advertising technology space — specifically through our work with Ad-MOTO.
For those unfamiliar, Ad-MOTO operates a fleet of electric scooters fitted with digital advertising screens across London. But these bikes do far more than display adverts. They collect environmental data — pollution levels, temperature, crowd movement patterns — and our job has been to build the systems that turn that raw data into actionable insights for brands like Pepsi, Channel 4, and Paramount TV.
What We Actually Built
The challenge wasn’t building an AI demo that looks impressive in a pitch deck. It was building production systems that process millions of data points daily and deliver results clients can trust.
Our analytics platform uses AI to generate natural language reporting — campaign managers can ask questions in plain English and get answers drawn from complex multi-dimensional datasets. We built heatmap visualisations that combine GPS tracking data with audience measurement to show exactly where and when ads are being seen. And we developed anomaly detection that flags unusual patterns in real time, whether that’s a bike going off-route or an unexpected drop in impressions.
The Honest Bits
I was candid with the audience about what works and what doesn’t. The natural language reporting took three iterations to get right — the first version was technically impressive but gave answers that were subtly wrong in ways that were hard to spot. That’s worse than no AI at all. We had to build a verification layer that cross-checks AI-generated insights against raw data before presenting them to users.
I also talked about the temptation to use AI everywhere once you start. We evaluated AI for route optimisation and concluded that a well-designed algorithm actually outperformed the ML approach for our specific use case. Knowing when not to use AI is just as important as knowing when to use it.
Key Takeaways
The response from the audience was encouraging. The questions weren’t about the technology itself — they were about practical implementation: How do you validate AI output? How do you handle edge cases? What’s the actual cost of running these models in production? These are the questions that matter when you move past the hype.
Thanks to Exeter Science Park for hosting, and to everyone who attended. Events like these are what makes the South West tech community genuinely valuable — practitioners sharing real experience, not vendors selling dreams.
If you’re exploring how AI could work within your business operations, we’re always happy to have a conversation. The first step is usually working out whether AI is actually the right tool for your specific problem — and sometimes the honest answer is that it isn’t.
