E.ON is cutting smart meter costs with an AI-powered remote diagnostic tool. The German energy giant teamed up with AWS to ditch field visits for error checks.
Instead of sending engineers to homes, customers just record a 7-second video of their smart meter LEDs using the E.ON app. Amazon Textract scans the video frames, detecting the LED states and blink patterns. The system matches these pulses to known error codes—no manual reading needed.
The result? About 135,000 fewer site visits a year, £10 million saved, and 84% diagnostic accuracy in real-world tests. 
Here’s how it works:
- The video is split into frames; only those with active LEDs move forward.
- Textract reads text labels on meters (“SW,” “WAN,” etc.) to find LEDs’ locations.
- The system counts LED blinks over 7 seconds, categorizing pulse frequency as Off, Low, Medium, or High.
- It then pulls the corresponding error code and sends a natural language explanation to customers instantly.
This tech gets errors right, even catching some that engineers missed initially. E.ON hopes to expand the tool across devices and add more computer vision AI down the line.
The move slashes the £20 million annual cost of dispatching engineers and boosts customer convenience. Testing also shows it supports E.ON’s 95% smart meter connectivity goal.
Sam Charlton, Product Manager at E.ON, said the tech tackles "entrenched issues often ignored" by using existing tech smartly.
Tanrajbir Takher, AWS Data Scientist, leads generative AI efforts behind the scenes, while Applied Science Manager Satyam Saxena and AI Strategist Tom Chester guide the implementation.
E.ON’s remote diagnostic rollout is a clear win for AI efficiency in energy management.
More info: E.ON
