Anonymity isn't a promise. It's the architecture.
Anyone can run an AI conversation. Few can promise, and then enforce in code, that what a person says is never tied back to them. That second half is the real product.
We store paraphrases, not their words. We do not attach identity to answers. And reports never show slices smaller than five.
What we keep, and what we throw away.
- ·An AI paraphrase of each answer
- ·A sentiment read on the paraphrase
- ·Aggregate themes across everyone
- ·The verbatim transcript, it is discarded after extraction
- ·A name, email, or token on any answer row
- ·Anything that joins a response back to a person
Four rules a competitor cannot bolt on afterwards.
The AI paraphrase is stored. The verbatim transcript is discarded after extraction, not archived.
No user, no email, no token that can be joined back from a response. The link is severed by design.
No theme, slice, or sample is ever shown below five contributions. Thin groups stay hidden.
Wherever a narrow group could leak a headcount in the display, the number is clamped.
This is organisational listening, not a safeguarding product.
If someone mentions distress, Allie responds with care and points to support, using the organisation's own resources where the campaign provides them. There is no notification that identifies the person and no escalation that names them. The anonymity promise holds even then, because the moment it does not, the product is worthless.
The question every buyer asks
No. We store paraphrases rather than words, we attach no identity to answers, and reports never show a slice smaller than five.
No. Named private shows who completed the conversation, never what they said. That completion reference is never joined to the answer text.
Everything runs in the London region, so data stays in the UK from end to end, in a schema isolated from any other system.
Honesty, made safe.
Ask the thing people will only say when they know they cannot be traced.