Product design question

Design the first-run onboarding for Glean in a large enterprise so a new employee quickly understands the value of Search, the AI Assistant, and agents without violating trust, permissions, or governance expectations. Walk through the user flow, the key UI moments, and what information you would and would not show at each step.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Enterprise onboarding design skill, balancing fast value demonstration against permissions and governance constraints.

Map the new employee's likely first need, probably finding a document or policy, and start onboarding there instead of a generic tour.
Sequence the flow: Search first since it needs the least trust and shows value fastest, then the AI Assistant, then agents.
Design each UI moment to show a real, permissioned result for that employee rather than a canned demo.
Decide what to withhold at each step, such as full agent capabilities or cross-team data, until permissions are established.
Add lightweight, contextual trust signals, like source citations on results, instead of an upfront governance lecture.
Test the flow against a low-access employee persona to confirm nothing over-promises visibility they lack.

Leads with Search because it has the lowest trust bar and most concretely shows value on day one.
Shows only content the specific employee is permissioned for, using their own account, not a demo environment.
Explains answers are grounded in real permissioned sources at the moment of first use, not in an intro nobody reads.
Delays showing agent capabilities until the user has context, avoiding an overwhelming first impression.

Front-loading a governance explainer that nobody reads before they see any value.
Using demo data that does not reflect the employee's actual access, creating a mismatch later.

How would this differ for a highly regulated customer versus a fast-moving startup?
What would you cut if you only had one screen to prove value?
How do you measure whether onboarding actually built trust?