insights

We let an AI agent brief every new support ticket. Here’s what changed.

Every new support ticket at an enterprise support org we work with now gets an AI briefing before an engineer opens it. Months in, here’s what we didn’t expect.

The briefing reads the whole thread, the customer’s history and similar solved tickets, then writes a short head start: what’s going on, what was tried before, and where to look first.

We expected the win to be faster first responses. What mattered more was:

  • Context. Engineers stopped spending the start of every ticket working out who the customer is.
  • Consistency. A new hire and a five-year veteran start from the same picture.
  • Honesty. When the agent isn’t sure, it says so, and that turned out to be the feature people trust most.

What didn’t work at first was long briefings. Nobody reads 800 words. We cut it to what you need in the first two minutes.

The lesson: an AI agent in a team succeeds on the boring things, like format, trust and fitting the workflow, not on how clever the model is.

All insights