What the System Actually Does, Measured.
Autonix has no client case studies yet, because it is new. What it has is a test programme — run against deliberately difficult enquiries, with the failures published alongside the wins.
Everything on this page measures the lead-response system. Website projects are scoped and discussed separately, on the Websites page.
How the Benchmark Is Run
A single friendly test proves nothing. The programme is designed so that the numbers mean something even when they are unflattering.
Every Number We Publish
Including the two that most vendors would quietly leave out.
| Metric | Why it matters | Run 01 |
|---|---|---|
| Median first-response time | The core promise, at the typical case. | Pending |
| 90th-percentile first-response time | The honest number — what your slowest tenth of customers actually experience. | Pending |
| Booked with no human involvement | The headline outcome. | Pending |
| Escalated to a human, split by reason | Escalation is a feature. Correct escalation is a win, and the split proves the rules fire as designed. | Pending |
| Escalated when it shouldn’t have been | A real failure mode — wasted team time. | Pending |
| Handled when it should have escalated | The dangerous failure mode. Published prominently, because it is the one that costs you a customer. | Pending |
| Correctly said “I don’t know” | Proves it declines rather than inventing an answer. | Pending |
| Booking accuracy | Right slot, right calendar, no double-booking. Where automation usually breaks. | Pending |
| CRM record completeness | Backs the “every lead tracked” claim with a number. | Pending |
| Recovery success on an induced failure | When we break the calendar on purpose: was the lead preserved and the team alerted? | Pending |
What the Testers Are Told to Try
The easy cases are included so the numbers stay representative. The highlighted ones are where systems like this usually fall over.
How We Report It
- 01Methodology is stated before any number, every time — including sample size, who the testers were, and what they were told.
- 02Failure rates get the same prominence as success rates. If a number is bad, it appears at full size, not in a footnote.
- 03Every run is dated and versioned. Old runs stay published, so improvement is visible and nothing is quietly deleted.
- 04Simulated data is labelled simulated, on the page and in the page’s structured data. It is never presented as a client result.
- 05No review or rating markup until real, attributable client reviews exist.
Client Case Studies
Want to Be the First Real Number on This Page?
The 30-day pilot runs on your real enquiries, heavily discounted, in exchange for an honest testimonial — good or bad — once it has been live long enough to mean something.