A year is a long time in a start-up. Twelve months ago, BiofuelAi was some very promising research, a small group of people who believed it could make a difference, and a new company with no customers, no commercial product, and a lot still to prove.

Today, we have a working platform, live industrial deployments, a growing team, international interest, and a £1 million Manchester Prize on the wall. Not bad for a first season.

Anyone who knows me will tell you I lean on sporting analogies, rugby especially, and building BiofuelAi has felt like a team sport in exactly that way: periods when everything clicks, others when we’re defending our own try line, and the odd occasion when the ball bounces somewhere nobody expected. That’s the nature of building something new, and here’s what the year taught us.

BiofuelAi grew out of the University of Surrey’s AI4AD research programme, which brought together artificial intelligence, process engineering, systems biology and life-cycle assessment.

The challenge was genuinely interesting. Anaerobic digestion plants are complex biological systems: feedstocks change, biology changes, operating conditions change, and decisions made today may not show their full impact for days. The research combined mechanistic models (our understanding of how an AD plant should behave) with machine learning that could learn from how a plant actually behaves. That hybrid approach became the foundation of BiofuelAi.

There’s a considerable gap between good science and a good commercial product, though, and our first year has largely been about closing it.

Models behave nicely in research environments. Real biogas plants are less accommodating. Data has gaps, sensors fail, feedstocks change, and every plant is different. Customers, quite reasonably, aren’t interested in how clever your model is if it doesn’t help them run their plant.

Getting our technology onto real sites has been one of our most important milestones this year, and it has changed our thinking. We used to talk a lot about digital twins and predictive models. We are still building them, but increasingly we start with a simpler question:

What decision are we helping the operator make?

What should I feed, how much, what’s likely to happen next, and which choice makes more money, produces more gas, or reduces risk? That’s where AI becomes useful rather than just interesting.

Another lesson from this year: AI isn’t the answer to everything. There’s an enormous amount of knowledge within the biogas industry, and experienced operators know their plants in ways no algorithm can replace, nor should it try to.

A good coach doesn’t run onto the pitch and play the game himself. They use information, experience and analysis to help the people on the pitch make better decisions, and we increasingly see BiofuelAi in the same way. Our models can process huge amounts of information, simulate scenarios, predict outcomes and quantify uncertainty, but the operator remains part of the decision. It’s about combining human experience with machine intelligence, not replacing one with the other.

Eventually every technology business has to move beyond interesting science and demonstrate commercial value, and our early results here are encouraging.

First-year pilots have shown revenue increases of 6-10% and profit improvements of 7-13% per site, alongside a reported 28% reduction in carbon emissions.

Those numbers matter, but the bigger point is what they tell us about the opportunity. There’s significant additional renewable energy potentially available from the UK’s existing AD infrastructure, and we don’t necessarily need to build new plants to produce more renewable gas: sometimes we just need to help the plants we already have perform better. Across an industry, relatively small improvements add up to something significant.

In June, BiofuelAi was announced as the winner of the £1 million Manchester Prize, the UK Government’s flagship AI innovation challenge. I won’t pretend this was just another day at the office.

Winning was an extraordinary endorsement of the science, our team, and, importantly, the problem we’re trying to solve. It also reinforced something I feel strongly about: AI needs to do useful things. Can it produce more renewable energy, reduce waste, improve profitability, help an operator make a better call on Monday morning? If the answer is yes, we’re getting somewhere.

The other big shift this year has been moving from a research-led group towards a technology company. Our team has grown across modelling, AI, software engineering, product development, customer success and business operations, and we’ve built the infrastructure and processes needed to deploy BiofuelAi repeatedly rather than treating every new plant as a research project.

A rugby team doesn’t become successful by producing one brilliant performance. It has to turn up the following Saturday and do it again, and that’s the challenge for Year Two: repeatability and scale. More plants, more data, more evidence, faster deployment, more automation, more customers, and increasingly, international markets. We’ve put some points on the board. Now we need to build the score.

No team achieves anything worthwhile because of one person, and BiofuelAi’s first year has very much been a team effort.

Thank you to our BiofuelAi team; to Michael and our colleagues at the University of Surrey who created the science this company is built on; to SETSquared Surrey; to our board, advisers, investors and industry partners; to the Manchester Prize team and UK Government; and to everyone across the biogas industry who’s offered advice, challenged our thinking or opened a door.

A particular thank you to our early customers and plant operators. You trusted a very young company with your data, shared years of operational experience with us, and have been remarkably generous with your feedback. BiofuelAi is a better product because of it. And finally, our families. Start-ups have an unfortunate habit of eating into evenings, weekends and holidays, and their support is very much part of this story too.

We’ve achieved more in Year One than I expected when we started, but we’re still in the opening stages of the game.

There’s plenty left to play for. Bring on Year Two.

Alan Beesley is CEO and co-founder of BiofuelAi. He combines commercial leadership with hands-on experience in anaerobic digestion and biological process optimisation. His focus is turning complex modelling and AI into practical tools that help operators make better decisions and deliver stronger performance from real-world assets.