Powering Up! How a Shared Foundation Model Is Improving Electrical Grid Management

Why It Matters

A sector‑wide AI model boosts grid reliability, cuts duplicate development costs, and strengthens UK energy security by keeping critical analytics in‑house.

Key Takeaways

• FoSMo processes over a billion grid images for condition assessment
• Shared anonymized dataset improves detection of rare defects across operators
• Ofgem funds model, enabling industry‑wide AI collaboration and cost reduction
• Governance separates asset condition analysis from sensitive network data
• Adoption hinges on internal champions and sector‑wide change management

Pulse Analysis

The electricity sector has long wrestled with labor‑intensive visual inspections, where engineers painstakingly review thousands of tower videos each year. Recent advances in computer‑vision convolutional networks have turned that bottleneck into an opportunity. Keen AI’s founder Amjad Karim leveraged these techniques as early as 2018, building a model that can automatically flag component wear and corrosion. By scaling to over a billion images, FoSMo delivers a unified view of asset health, allowing operators to move from intuition‑driven maintenance to data‑driven decisions that enhance safety and reduce downtime.

What sets FoSMo apart is its shared‑data architecture, funded by Ofgem’s Strategic Innovation Fund. Operators contribute anonymized photographs, creating a dataset orders of magnitude larger than any single utility could assemble. This collective intelligence dramatically improves the detection of rare defects—events that, while infrequent, can have catastrophic consequences. The model’s design deliberately separates condition analysis from any identifiable network information, preserving security while still delivering actionable insights. Moreover, the platform’s sovereign‑ready architecture means the UK can retain full control over the AI stack, even if the underlying compute runs on global cloud providers.

Adoption, however, hinges on more than technology. Utilities must cultivate internal champions, navigate complex procurement processes, and meet stringent reliability standards before trusting a shared model with critical assets. The success of FoSMo offers a template for other fragmented infrastructures—water, telecoms, and transport—where similar data silos impede AI progress. By demonstrating how pooled imagery can power digital twins and predictive maintenance without compromising competitive advantage, FoSMo paves the way for broader collaborative AI initiatives across essential services.