Agents that close the loop: lessons from 60% auto-resolution
What it took to move an AI support suite from demo to resolving the majority of tickets on its own.
Read articleMeasure before you optimise
We log tokens and cost per feature. Usually a few flows account for most of the bill.
Route by difficulty
Simple requests go to small, fast models; only hard ones reach the largest. Quality stays, cost drops.
Cache aggressively
Repeated context — system prompts, shared documents — is cached so you pay for it once, not every call.
Have a question about this, or a project where it applies? Write to us — Slava reads every reply.
