The AI Cake framework for Utilities

Executive Summary (tl;dr)

• Utilities are eager for AI but lack holistic solutions. Successful implementation requires following the "AI Cake" framework for Utilities, starting with a single asset view.
• The framework has three levels: Single View (data consolidation), Scoring (contextualizing data), and AI Applications (advanced insights and management).
• 90% of Utilities struggle by attempting level 3 before completing levels 1 and 2.

While Utilities across North America remain enthusiastic about AI adoption, comprehensive solutions to meet this need are still lacking. Many AI solutions exist but often overlook essential foundational tasks. There is a big push on generic customer service, billing systems and various other customer centric applications. While these are important use cases, there are far more pressing use cases for Utilities internally especially involving their asset management and reporting capabilities. Our team's firsthand experience in testing legacy software vendors with Utilities has consistently revealed the challenges associated with AI implementation.

A key to implementing AI is for them to work their way through the AI Cake framework for Utilities. We estimate ~90% of utilities attempting to implement AI do so before having a single asset view data table, and this is why they face so much confusion in implementing AI.

‍

Level 1 - Single view

In order to implement AI, a Utility requires having a single asset view. This singular data view can come in many names and forms. Some may refer to it by the front end "asset registry" they interact with. More technical operators may refer to an ETL (extract, transform, load) solution. Having this view requires that all your latest asset data is in one structure including supplementary data for Utilities: asset inspection reports as excel or PDF formats, LIDAR formats, asset images, asset videos and drone footage. Core data systems include: GIS, SCADA/ADMS, and Historian.  

‍Level 2 - Scoring

Scoring helps to interpolate data making sure the goal of data is clear. Having all the data in one place is great, but AI requires context to infer purpose. For Utilities, scoring typically involves many additional data layers but the critical inference is scoring assets. Please note that Utilities have various complexities and their own practices. Every Utility will utilize unique equations to score asset health. Many times scores are done from shared regional standards.

‍Level 3 - AI Applications (the ultimate goal!)

The final level of implementing AI is - the ability to run LLMs. Completing the first two levels of the AI Cake will unlock core AI functionality such as: managing assets and asset data outside of what is humanly possible. These applications provide you the ability to have inference and insights equivalent to an employee that has been working at a Utility since day 0 and knows all the ins and outs. These sorts of systems and applications can be looked at as basically super workers, working for your grid 24/7, don’t need food or sleep and never retire! They know everything that was and can help predict everything needed for success by connecting all data available.

We do believe there will emerge a level 4 of the framework when coherence into non-core operations, the managing of assets, is integrated. We do not yet speculate on the next level, as core AI matches the core responsibility of a Utility: manage a grid to get consumers electricity as efficiently and safely as possible.

‍