Answer capsule
A coaching-system buyer should distinguish chat content from interaction metadata, earlier session data, derived insights, and coach sharing before describing an incognito control to participants.
What the source establishes
- BetterUp's current AI Privacy Notice says participation requires explicit opt-in by the individual and the sponsoring organization.
- The notice identifies collected AI-interaction data including the user identifier, initial message, input messages or prompts, outputs, and timestamps.
- It says optional incognito mode stops collection of future message content and responses for the remainder of the session and erases those interactions when the chat session closes, while BetterUp can still know an interaction occurred; the mode may not be available in all AI experiences.
- The notice also describes use of personal data for tailored experiences and optional sharing of specific insights or chat content with a coach at the individual's choice, so a buyer needs to map what remains outside the incognito content boundary.
Incognito is a scoped control
BetterUp's current AI Privacy Notice describes individual and sponsoring-organization opt-in, the data collected in AI interactions, and an optional incognito mode that may not be available in all AI experiences. It says the mode stops collection of future message content and responses for the rest of the session and erases those interactions when the chat session closes, while BetterUp can still know that an interaction occurred. That is more specific evidence than a generic privacy promise. It is also a scoped control, not proof that every related datum disappears. A buyer should avoid telling participants that a session is invisible, anonymous, unrecorded, or wholly excluded from downstream use unless current contracts, configuration, and testing support each of those separate claims.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Map content and residual data separately
The data map should distinguish messages and responses from user identity, timestamps, session and device events, opt-in state, safety and support records, feedback, usage measures, pre-incognito content, and any derived recommendation or profile signal. It should state exactly when incognito begins, what happens when the participant leaves or closes the session, whether a transcript can be recovered, and which operational or legal retention rules apply. The buyer should also identify whether prior personal data can still shape the interaction and whether any insight can be shared with a coach at the member's direction. A clear diagram or table is preferable to one blanket label because each data class can have a different owner and lifecycle.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Test the participant experience
Using non-sensitive test data, administrators and participant representatives should verify organization opt-in, member choice, incognito availability and its visible indicator, the point at which content collection stops, session closure, history, export, deletion, coach-sharing controls, opt-out or revocation, and support escalation. The test should compare participant, coach, administrator, provider-support, and downstream-system views. It should also check what usage or interaction evidence remains after content is erased. Any mismatch between the notice, contract, configuration, and observed behavior should remain an evidence gap and block stronger participant claims. A feature name or interface message cannot resolve that conflict by itself.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Give the participant meaningful control
Before use, the participant should receive plain-language notice about what incognito covers, what may remain, when the choice takes effect, and how to obtain help or exercise available data rights. The organization should not use participation, opt-in, interaction metadata, inferred themes, or coaching content for employment decisions unless a separately authorized and lawful process makes that use explicit. Coaching, privacy, security, workforce, legal, and procurement owners should approve the boundary; the vendor and administrator do not replace participant agency or employer accountability. If residual data cannot be explained or the control cannot be reproduced, sensitive coaching topics should stay outside the AI system.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Decision test
Ask whether the source changes the decision itself, the evidence required, the implementation sequence, or only the language used to describe an existing capability. Record which claims are directly supported, which are provider statements, which require an independent test, and which remain unknown. A source-linked review should make uncertainty easier to see, not bury it inside a blended score.
Questions to take into review
The publication supports research and executive decision preparation. It does not provide legal, financial, accounting, employment, clinical, cybersecurity, investment, procurement, or implementation advice.