AI & trust
One call, two views. What your manager doesn't see isn't hidden from them — it does not belong in their view at all.
What Mara sees
What her manager sees
Try it
An average of two people protects nobody. On a small team it can be worked back to whose number it is. Drag the slider and see what would reach the manager.
You don't count yourself: you know your own number and could subtract it back out.
With 3 people, 2 count towards it. Too few — a number like that can be worked back to how one person is doing.
The difference
If someone here adds a score to the manager's view, the change doesn't ship. Not because a person catches it by hand — the tests fail. This is what that looks like:
// manager-contract.ts daysSinceActive: number | null + callScore: number // “managers only” ✓ no-person-profiling.test.ts ✓ intended-purpose-copy.test.ts × employer-firewall-behaviour.test.ts > each member exposes EXACTLY the declared contract keys A field appeared on the manager-visible member row. If it says anything about how the person is doing, it does not belong on this object at all. Tests 1 failed
Reconstructed from our own test run. We checked every one of these by deliberately putting the bug back. A test that has never gone red proves nothing.
The fair objection
You do. You keep everything you need to manage. What's gone is what you'd need to grade.
Three flags that fall straight out of the numbers: quiet for a while, no practice this period, or practising without making calls. Enough to start a conversation. Not a grade.
Averages and skill gaps at team level, once enough people are in the mix. Plan training with it, without grading anyone.
Mara has her analysis. Ask her about it. She decides what she shares with you. What someone shows you willingly beats anything analysed behind their back.
What this costs us
Small teams see less. Under five people we show no average at all. That is a real limitation. We could lower the threshold — but then the number could be worked back to a single person again.
No leaderboard. If you want your people put in order, we're the wrong product. Better said now than after signing.
We removed features that went down well. One sorted people into five personality types. It has been gone since August 2026 — including the labels already stored.
For employee representatives
The manager sees activity numbers: calls made, practice done, when someone was last active. No score. No leaderboard. No target comparison for any individual. That these numbers aren't evidence for decisions about pay, promotion or discipline is stated inside the product — right above the numbers.
No. We read the text of the call and when each word was said. The timings give us pace; the text gives us word choice. We don't judge how someone sounds, and we don't infer feelings.
Yes. The technical claims on this page are covered by tests that rerun on every change. We'll hand you the list — including the times a test was itself faulty. Those are in there, with dates.
Reviewing a call honestly means naming what went badly. People only do that when the review doesn't land on their manager's desk.