polite.aiDocs
ACCOUNT

Beta feedback & session recording

What the beta records to improve the product, how sensitive fields are protected, and every way to tell us what worked and what got in your way.

polite.ai is in beta, and the most valuable thing you can give us is a normal working session. Rather than interrupting you with surveys, the beta observes how the product behaves around you — and gives you quiet, entirely optional ways to tell us how it felt.

What gets recorded

While the beta instrumentation is switched on, each signed-in session records:

  • What you did — meaningful product actions (saving an agent, running a test call, attaching a number), with their outcome, timing and any error.
  • How the session unfolded — a visual replay of the dashboard as you used it, so we can see hesitation, retries and dead ends we'd otherwise miss.
  • Configuration text — the instructions you type to the builder are usually the heart of a usability problem, so non-sensitive configuration text is captured as you write and revise it.
Good to know

Recording never gets in the way of the product: if telemetry fails, your action still completes — observation is never part of the critical path.

What never gets recorded

Every input field carries a recording classification, and the default for an unclassified field is shape only — we see that you typed, not what:

  • Passwords, secrets and API keys are excluded outright: the fields are blanked from replay and their values are never stored.
  • Identifiable values such as phone numbers are masked.
  • Free-form text that hasn't been deliberately classified is recorded as interaction shape (lengths and edits), not content.

A separate hygiene pass looks for things like email addresses, card-like and key-like strings that slip into captured text, and flags or redacts them.

Please still be kind to yourself: don't paste real confidential information into test configurations that don't need it.

Telling us how it went

  • Reactions — hovering a builder confirmation reveals four small icons: loved this, worked well, something wasn't right, seriously frustrating. One click records how that moment landed; Add a note is there if you want to say more. Confirmation toasts carry a quick 👍/👎 pair.
  • Two quick questions — at a natural stopping point (after a test, or returning to the dashboard) we may ask, at most once a day: what worked particularly well? and what got in your way? Both optional, never a pop-up.
  • Beta · Feedback — the standing control in the top bar (also under Help). Whatever you write arrives with its context attached — screen, recent actions, build — so you never have to explain where you were.

Silence is fine too. No reaction is read as no reaction, never as satisfaction.

Retention and access

Raw replay and captured configuration text are kept for 90 days, then deleted; what we learn from them (aggregate findings, counts, timings) may be kept longer. Access to raw session data is restricted to the polite.ai team for beta research, support and engineering diagnosis, and every replay view is logged.

If you'd like a session's recordings removed sooner, contact support@polite.ai with the approximate time and we'll purge them.

Last updated 2026-08-17