AI & Technical question

Design the first version of Harvey’s cross-system legal research experience for an enterprise customer using systems like iManage, SharePoint, and internal knowledge bases. Which user workflow would you support first, and how would you handle permissions, source freshness, citations or provenance, and failure modes so legal professionals trust the output?

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

Whether you can design a trustworthy enterprise legal research experience that handles permissions, freshness, and provenance correctly from day one.

Pick the first workflow: due diligence or precedent research pulling from a specific matter's document set across iManage, SharePoint, and internal knowledge bases, since it is high-frequency and has clear, bounded scope.
Design permissions as a hard constraint, not a feature: results must respect each source system's existing access controls, with no cross-matter leakage, verified before any content is surfaced.
Handle source freshness by surfacing the retrieval date and flagging when a cited document may be outdated relative to the matter's current state.
Require citations and provenance on every claim, showing the exact source document and location, since legal professionals will not trust or use ungrounded answers.
Define failure modes explicitly: when permissions block a relevant document, tell the user a restricted result exists rather than silently omitting it or (worse) revealing its content.
Pilot with a single practice group and a bounded document set to validate permission handling and citation accuracy before expanding to broader cross-system scope.

Picks a bounded first workflow instead of trying to cover all legal research use cases at once, showing real MVP judgment.
Treats permissions as a precondition, not a feature to add later, appropriate for a legal product where data leakage has serious consequences.
Requires citations and provenance as non-negotiable, directly matching how legal professionals evaluate trust in research output.
Names the specific edge case of a permission-restricted but relevant document and proposes a safe way to acknowledge it without leaking content.

Treating permissions as an implementation detail rather than a core design constraint.
No citation or provenance requirement, which legal professionals will reject immediately.
Scoping too broadly across all systems and workflows at once instead of piloting narrowly first.

How would you handle a document that exists in two systems with conflicting versions?
What would you do if citation accuracy is high but permission enforcement has a false negative?
How would you expand from the first practice group to the rest of the firm?