Customer success, rebuilt as an agentic operation
One person ran support, billing, the help centre and founder analytics for a B2B SaaS — by turning the whole function into a routed, self-correcting workflow on Claude Code.
The whole function, one person
Extrovert is an AI LinkedIn engagement product for B2B sales teams, with a founder and four engineers. I ran support, billing, the help centre and founder-facing analytics as Head of Customer Success.
Rather than work the queue by hand, I rebuilt the function as an agentic operation on Claude Code: a router that decides what kind of question it is, an investigation skill wired directly into production, and a loop that turns every wrong answer into a rule so the same mistake does not come back.
The point was never automation for its own sake. It was that a support answer should be traced to the code path that produced it and confirmed against production, not assembled from a plausible story — and that one person should be able to hold that standard across a whole queue.
Six live sources, one setup document
The whole environment rebuilds on a fresh machine in under an hour, with a smoke test per integration.
One rule runs through all of it: the database is read-only and there is no send tool on the inbox. The agent produces SQL and drafts; I run them and I send them. Nothing reaches a customer or the database without a human in between.
How a ticket gets routed
The router's first and only real decision: does this answer depend on the state of this account?
A help centre with two audiences
Three AI agents answer on the company's behalf and all three read these articles. When an agent walks someone through a setup, the article is its prompt.
one wording
Every wrong answer becomes a rule
A post-mortem in a document is a diary entry. The same post-mortem written into the skill is a guardrail — it fires on the next case, without anyone remembering it.
190 investigations, 14 weeks
Each one ends in a verdict backed by production evidence and a code reference — two independent sources, or it does not ship.
to the 18th
Six questions, answered from the raw data
Not a dashboard reading. Each question fans out into parallel reads of production, and converges into something the founder can act on.
It’s been great working with Denis on automating our support function with agentic workflows. One person now covers what would normally need a team, and the bugs that reach engineering arrive already traced to the cause.
