How might we design the onboarding and ongoing consent experience of a personal AI agent so that users genuinely understand the privacy-utility tradeoff — and make an informed, confident choice rather than a rushed or anxious one?

Personal AI agents — assistants that manage calendars, messages, health data, location, and behavioural patterns on a user's behalf — become meaningfully more useful the more data they can access. This creates a structural tension: the agent's value to the user is directly proportional to the data the user is willing to share, but the decision to share is shaped by trust, perceived control, uncertainty about consequences, and the framing of the request itself. Most current AI agent onboarding handles this poorly: permissions are requested in bulk, benefits are overstated in general terms, and the actual data flows remain opaque. Users either accept everything without reading or reject everything out of caution — neither outcome serves them well.

The team will research how users currently experience data permission requests in AI contexts, identify the psychological and design factors that shape their decisions, and prototype alternative onboarding and consent flows that present the tradeoff honestly and clearly. The work is service design at its core, informed by behavioural psychology and UX research. ## Goals - User research synthesis: how do different user segments currently reason about sharing personal data with AI agents? - At least two contrasting consent/onboarding design concepts, tested with representative users - A framework of design principles for transparent data-for-value communication in AI agent contexts - Prototype (interactive or documented) of a preferred onboarding flow

As agentic AI moves from early adopter to mainstream, the quality of the data conversation will determine both user trust and the actual utility of the agents. Companies deploying personal AI agents — across health, productivity, finance, and consumer services — need validated design approaches. This topic is commercially relevant to any organisation building or integrating personal AI agent capabilities.