EasyAsk

At Precision Nutrition, where I led product marketing, our support team kept answering the same questions about our nutrition coaching education products. Zendesk and templates helped, but chat was staffed for eight hours a day. Outside those hours, someone considering a purchase had to leave an email address and wait. The website was open. The answers had office hours.

We’d been experimenting with AI across the team, and this seemed like a useful job for it. The catch was getting it to use our product information instead of making things up. Confident nonsense wouldn’t have improved customer support, even if it arrived immediately.

Give it the answers first

I learned about retrieval-augmented generation, or RAG. The idea’s straightforward: look up the relevant information and give it to the AI before it answers. I used our product details and support responses, then tested hundreds of questions and worked on the cases where it still invented details.

That became EasyAsk. Visitors can ask a question on the spot. If the information doesn’t cover it, the assistant can pass the question to the human team rather than guess.

Precision Nutrition still uses it across its website, even after I left to join Toyo in April 2026. It’s doing the job I built it for, which is a satisfying thing to be able to say about a project.