A 24/7 Customer Assistant With Live Network Lookup
Vehicle Services · 2026 · Customer Support Automation
Around 95% of conversations end without an operator. The assistant calls the client's own network database live, so a driver gets an answer and an address instead of a queue.
The challenge
A vehicle services company runs a national network of affiliated workshops. Its customers asked the same things all day — what the company does, how to get something done, who to contact, which rules apply — and every one of those questions landed on a person.
Then there was the question that could not wait. A vehicle stops at night. The driver needs one thing: where do I go, now. That information existed — who is affiliated, where they are, which one is closest — but it sat in a database no customer could reach, behind a line that opened at nine in the morning.
The company did not want a bot that guesses. It wanted a stated line between what software answers and what a person handles.
What we built
A public assistant on the client's website, running in the client's own cloud account.
- It answers. General questions are answered from the company's own documentation, which the client updates whenever it changes. The assistant reads the current version, not a copy from launch day.
- It does something. When a customer says where they are, the assistant calls the client's network database through their own API and returns the nearest affiliated workshops — from a network of 2,000 — with somewhere to drive to.
- It hands over on purpose. The topics that go to a person were chosen with the client before we built anything, and written into the assistant as a perimeter, not as a fallback.
- Operators watch it work from a dashboard: what was asked, what was closed, what was escalated and why.
The second point is the whole difference. A text generator can describe a network. This one queries it, live, at 3am — so a customer whose vehicle broke down gets an answer and a destination instead of an apology.
The dashboard is also how the perimeter moves. Escalated topics are reviewed with the client: the ones that turn out to be answerable get added to the assistant, the ones that need judgement stay with a person. The line is a decision, and it is revisited with evidence.
The results
Around 95% of conversations end without an operator, day and night. Volume runs at roughly 10-20 conversations a day and is still climbing as the assistant covers more of what customers actually ask.
The rest is not a failure rate. It is the perimeter: those conversations go to a person because that is what we decided should happen. In a market full of resolution rates nobody defines, saying which conversations the software does not touch is what makes the ones it closes believable.
| Number | What it measures |
|---|---|
| ~95% | of conversations closed without an operator |
| 2,000 | affiliated workshops the assistant can search |
| ~10-20 | conversations a day, and growing |
| Stated up front | which topics go to a person |
What you get
The assistant on your public site, connected to one of your own systems for the lookup that matters in your business — the nearest technician, the nearest branch, the nearest service centre, what is in stock nearby. The perimeter is agreed in writing before we build: which topics it closes, which go to a person. Process audit first, scope in writing, and 30 days of support. The same assistant also runs behind the client's partner login, where the answers are scoped to whoever is logged in.
Wondering how much of your inbound could be answered without an operator? Start with the free AI Audit, or book a call.