Balancing the grid with an AI energy platform

The problem

Renewable generation is volatile and demand is shifting (EVs, heat pumps). Balancing supply and demand across millions of distributed assets in real time is beyond static, rule-based systems.

The AI approach

Kraken forecasts generation and demand, then orchestrates a large fleet of distributed assets — EV chargers, home batteries, heat pumps — as a virtual power plant, automatically shifting load to match grid conditions. A GPT-style tool (“Magic Ink”) also drafts customer communications.

Evidence it works

Kraken reports forecasting solar output to ~88% accuracy and wind to ~82% accuracy 24 hours ahead, orchestrating over 2GW of domestic assets, and saving consumers ~£150M/year; it is licensed to 35+ utilities globally. (Kraken was spun out of Octopus in 2024 partly to manage conflicts of interest with the utilities it serves.)

What “good” looks like

Measurable curtailment reduction and peak-shaving, lower bills, and stable grid services — with forecasting accuracy tracked and improving.

Feasibility & cost shape

A platform-scale capability; most organisations would license/partner rather than build, but integration into existing systems and data is substantial.