The promise of artificial intelligence in manufacturing is enormous — but so is the responsibility that comes with deploying it. As AI systems move from research labs into real production environments, the question shifts from "Can AI do this?" to "Can we trust AI to do this reliably and safely?"
We're proud to share that our case study on battery electrode thickness prediction using AI models and advanced metrology has been published in the High Value Manufacturing (HVM) Catapult's landmark report, Trustworthy AI Adoption. The report tackles one of the most pressing challenges facing UK industry today: how to adopt AI in a way that is not only effective, but dependable, transparent, and safe.
Why Battery Manufacturing Needs AI — and Why AI Needs Guardrails
Battery electrode manufacturing is a complex, multi-variable process. Small deviations in coating thickness, porosity, or material composition can have outsized effects on cell performance, safety, and lifespan. Traditional quality control methods, while valuable, often catch defects only after the fact. AI changes this equation by enabling predictive insights — identifying problems before they manifest and optimising process parameters in real time.
But prediction alone isn't enough. In an industry where the consequences of failure can range from reduced vehicle range to thermal runaway events, the AI systems guiding manufacturing decisions must themselves be rigorously validated. A model that works brilliantly under one set of conditions but fails silently under another is worse than no model at all.
This is exactly the problem our case study addresses.
What Our Case Study Demonstrates
Our work combines AI-driven prediction models with advanced metrology — the science of measurement — to characterise battery electrodes with greater accuracy and earlier in the production cycle. By fusing data from multiple measurement sources with machine learning, we can predict electrode characteristics that would traditionally require destructive testing or lengthy analysis.
Critically, our approach doesn't treat the AI model as a black box. We focus on understanding the boundaries of model reliability, quantifying prediction uncertainty, and establishing clear criteria for when the model's outputs should and should not be trusted. This is what moves AI from a promising experiment to an industrial-grade tool.
The Bigger Picture: Trustworthy AI Across UK Manufacturing
The HVM Catapult report brings together real-world examples from across six Catapult centres, each demonstrating how AI technologies are being applied in high-value manufacturing contexts. Together, these case studies make a compelling argument for something the industry urgently needs: a structured, regulated framework for testing and validating AI solutions before they go into operation.
This isn't about slowing down AI adoption. It's about accelerating it responsibly. Manufacturers need confidence that the AI tools they deploy will perform as expected — not just on day one, but consistently over time as conditions, materials, and processes evolve. The report highlights the technical mechanisms required to achieve this, from robust testing protocols to ongoing monitoring and assurance.
What Comes Next
This case study is just the beginning. We're actively developing AI assurance and reliable operation solutions, with a particular focus on battery manufacturing. As the UK and the world ramp up battery production to meet electrification targets, the need for trustworthy, validated AI in this space will only intensify.
The manufacturers who get this right — who pair powerful AI with rigorous assurance — won't just build better batteries. They'll build the trust that underpins the entire transition to a sustainable energy future.
Stay tuned. There's much more to share on this front, and we're excited about what's ahead.