Case study in progress
Rate Automation, customer zero
First live run: pending
This page exists so you can see what has actually been built before anybody claims a result. There are no numbers on it yet, and there will not be any until a real agent has completed a real task against a live connector.
What Rate Automation is
Rate Automation is a platform for Alberta contractors, run by the same two people who own Counter. It keeps a business's own documents and history searchable with citations, tracks incoming leads, and holds follow-ups until a human approves them. It is customer zero because the founders control its API, which means the connector could be built and broken and rebuilt without anybody else's release schedule in the way.
What exists today
- Three tools are exposed. Ask a question of the business's own memory and get an answer with citations; list the leads that are still open; propose a follow-up to one of them. The first two only read. The third is a write, and it does not run until a human approves it.
- The pre-flight harness was built and run against test fixtures. It injects the failures agents actually produce: a dropped response that makes the agent retry, a session that expires part way through, an error where a result was expected, and hostile text planted in a field the agent will read back.
- The approval path, the idempotency keys and the structured errors are the same code every other Counter tenant gets. Nothing here is a special case.
What does not exist yet
- A live run. No production assistant has completed a task against this connector in front of a customer.
- Numbers. No call counts, no conversion rate, no time saved, no uptime figure. We have run the harness against fixtures, which tells us the code behaves; it tells us nothing about demand.
- Quotes. Nobody has been asked for one, because there is nothing yet to have an opinion about.
What gets published here when it does
The first live run, with the trace: which assistant called, what it asked for, what the harness had already caught before that day, and what the owner approved or refused. If the run goes badly that will be on this page too. A case study that only ever reports good news is an advertisement, and you can tell the difference.