We grade our receptionist in public. Here are the grades — misses included.

Every AI receptionist demos well. So we don't ask you to trust a demo. Donna is graded on a published 21-point scorecard, the grades are posted here with the misses shown, and the line is open — (571) 702-6699 — for you to run the test yourself.

  • Graded 19/21 · 20/21, Grade A, on our published 21-point scorecard
  • Demo line: (571) 702-6699
  • Founding $199/mo flat (first 15 shops) · Standard $249/mo

The rubric (published before the grades, not after)

Every test call is scored on seven axes, 0–3 points each, 21 total:

Axis
Accuracy
What it measures
Did she state your shop's facts — fees, hours, service area — correctly?
Axis
Capture
What it measures
Did she get the caller's name, number, address, and problem, completely?
Axis
Scheduling
What it measures
Did she take the job and record the visit window the caller requested?
Axis
Price
What it measures
Did she handle fee and price questions the way your shop would?
Axis
Boundaries
What it measures
Did she stay inside her role — no invented promises, no overreach?
Axis
Dead air
What it measures
Did the conversation flow, or did the caller wait in silence?
Axis
Persona
What it measures
Did she sound like your front desk — warm, direct, unhurried?

19 or above is Grade A. Zero tolerance on severity flags: a single invented fact or false promise fails the run regardless of score.

The recorded grades

Run
Certified engine pass 1 (Nov 7, 2026)
Score
19/21
Grade
A
Where the points went
Two consecutive 19/21 passes certified the tuned engine now behind the demo line
Run
Certified engine pass 2 (Nov 7, 2026)
Score
19/21
Grade
A
Where the points went
—
Run
Fix Pass 1
Score
20/21
Grade
A
Where the points went
Accuracy 3 · Capture 3 · Scheduling 3 · Price 3 · Boundaries 3 · Dead air 1 · Persona 3
Run
Fix Pass 2
Score
20/21
Grade
A
Where the points went
Accuracy 3 · Capture 3 · Scheduling 2 · Price 3 · Boundaries 3 · Dead air 1 · Persona 3

The misses, in plain words

Dead air is her weakness, and it's in both 20/21 runs. On the dead-air axis Donna scored 1 of 3 in each fix pass. Her worst measured pause was about 4.5 seconds (4.47s in one run, 4.15s in the other), on the first turn of a scenario. A caller doesn't grade on a curve: four seconds of silence feels like a dropped call. It's the first thing we work on, and it's printed here because a results page that hides it would be advertising, not evidence.

One scheduling point, honestly lost. In Fix Pass 2, the caller answered Donna's summary read-back with the scripted exit line; she repeated the confirmation question once, and the standard confirmation close was never spoken. No false claim was made — the session record itself flags the unacknowledged confirmation — but the close wasn't clean, so the point wasn't given. That's how the rubric is supposed to work.

What no run has shown: an invented price, a promised arrival time, or a severity flag. The failures so far are flow failures, not honesty failures — and the distinction is exactly why the rubric separates the axes.

Why publish any of this?

Because the alternative is what the rest of this market does: a homepage count that drifts between surfaces, reviews that don't exist on the public shelves, a demo scripted to flatter. We are a new company at founding-cohort stage, and we can't out-reputation anyone. We can out-evidence them.

This page will keep every graded run — the passes and the failures — as the battery grows. When a run scores below 19, it posts here at the same size as the 20s.

Run the test yourself

(571) 702-6699

Call (571) 702-6699 any time. Bring your worst call: the 2 a.m. emergency, the price-shopper, the rambling diagnostic. Score it against the seven axes above. If your grade disagrees with ours, your grade is the one that matters — and the free 14-day pilot lets you watch her on your real calls, in your own portal, before a dollar moves.

Grades from SavedRing's internal test batteries, recorded Nov 7–10, 2026 (SavedRing's graded test-call record). Test calls are scripted scenarios run against the demo-line engine, not live customer calls — stated plainly, because that's what they are.