I’ve been wanting to write a post on lessons learned advising companies on scaling GitHub Copilot at enterprise companies, and how those lessons apply broadly to AI tools. But, Will Larson (author of An Elegant Puzzle), recently wrote an excellent piece on this very subject:

Facilitating AI adoption at Imprint

The content of this post is gold, and worth reading for anyone who has a hand in AI corporate programs. Matching my own experiences, these quotes stood out:

Measuring AI adoption is, like all measurement topics, fraught

Many organizations look for the silver bullet metric that tells them whether their AI adoption is successful or a failure. The reality is there’s lots of gray area. Ultimately, organizations need to decide what matters most to them, how to approximately measure success, and ensure stake holders are aligned internally.

And next, Larson’s guide for bolstering adoption by applying AI agents to a net new domains is great:

(product eng) find a workflow with a lot of challenges or potential impact (product eng) work closely with domain experts to get the first version working (platform eng) ensure that working solution is extensible by the team using it (both) monitor adoption as indicator of problem-solution fit, or lack thereof_

Applying agile principles to AI adoption is key. Finding a feature, product, or team, and then supplying them with AI tools, and finally measuring adoption of this cycle is a great approach to determining AI utility. It sounds simple, but it’s surprising how few companies do this well.