What happens when AI starts handling the follow-through?

Tracking down invoices used to take Gavin Roy roughly two hours each month. Now AI agents gather them from his email and vendor portals and prepare his expense spreadsheet. (Sounds like a dream? See it for yourself.) It’s one example of how he’s handing off work that once required his attention.

At a recent Enrich AI Show & Tell, Gavin and fellow member Mona Chitnis demonstrated three projects they’re building, from an email assistant to a personal knowledge system and AI agents that help resolve customer issues.

Gavin’s email assistant started as an experiment in handling unwanted sales outreach. It now prioritizes messages, summarizes repetitive notifications, and connects related conversations. Mona tackled a different frustration: collecting useful information and losing track of it when she needs it. Her system brings together meeting notes, articles, and other sources to produce briefings shaped by her interests and priorities.

The experiments extend to engineering, too. At AWeber, Gavin’s team is building agents that investigate customer issues, develop fixes, and submit code for review. Around 80% of recent agent-generated pull requests were accepted, according to the results he shared.

On our blog, we take a closer look at all three projects, including how members are deciding what to delegate and where human approval still matters.

Upcoming Events

Want to hear more insights like this in real time? Join one of our upcoming events to continue the conversation with fellow Enrich members:

What is Enrich?

Enrich is a private community for ambitious, growth-minded leaders in tech. We host events, private dinners, and curated peer groups to help product and engineering leaders make better decisions, grow in their roles, and build meaningful relationships. Apply to join.