In the Digital Sales Room, sellers had a static report showing which content a buyer viewed and which they didn’t. It showed the data, but not what to do with it.

So I explored using AI to turn that report into a short summary: which content was viewed longest, who viewed it, and when. Then I added suggested next steps, like an automatically drafted follow-up message based on that buyer’s activity. I showed it to internal sellers. They liked the drafted message, but every one of them said they’d tweak it before sending. That told me the right model was AI proposing and the person deciding, not AI acting on its own.

It never shipped, but it shaped my approach to agents: be clear about what the AI handles, and keep the person deciding.