How AI processes your data

AI in the platform is fed from your project data. It is worth knowing exactly which data, because that is what determines what you can safely put in.

What each surface reads

Every page that uses AI discloses what that AI reads, in the page header, next to the feature itself. That disclosure is generated from the actual data flows rather than written by hand, so it does not drift away from the code.

Where it runs

Model inference runs on AWS infrastructure inside the platform's own accounts, in the region for your cell. Your project content is not used to train shared foundation models. Where the platform learns across deals, it does so on anonymised aggregate signal at population level, never on raw tenant content.

Consent

AI features are consent-gated at account level. An Owner or Admin grants it, and revoking it disables the AI-powered features, report composition, risk analysis, work plan generation and the rest, until it is granted again.

Retention and deletion

Transcripts are retained according to your account's settings. Deleting a transcript cascades: every update derived from it goes too, so there is no residual extracted content left behind. This matters more than it sounds, because the derived content is the part that would otherwise survive quietly.

Your rights, and those of people named in your record

You can export your account's data at any time, on any plan including a free trial, because that is a right rather than a feature. Where the record holds information about a named individual, there is a subject-access export and an erasure path for that person specifically.

The rule underneath all of it

The AI drafts and proposes; a person decides. There is no path by which model output becomes a project record without somebody accepting it.