Design an AI assistant for knowledge workers (search, writing, automation).

### Signal to interviewer

I can design cross-functional AI that compounds productivity while preserving trust and execution quality.

### Clarify

I would clarify user roles, tool ecosystem, top repetitive tasks, and risk tolerance for automated actions.

### Approach

Create a knowledge work operating layer unifying grounded search, guided writing, and approval-aware automation flows.

### Metrics & instrumentation

Primary metric: time reduced on repetitive low-leverage tasks. Secondary metrics: search success, writing acceptance, and automation completion reliability. Guardrails: hallucinated content incidents, misfired automations, and user override spikes.

### Tradeoffs

More automation boosts throughput but raises verification burden. More review controls increase trust but can reduce speed gains.

### Risks & mitigations

Risk: incorrect generated claims; mitigate with citations. Risk: workflow misfires; mitigate with staged approvals. Risk: fragmented context across tools; mitigate with shared workspace memory.

### Example

For operations teams, AI answers policy questions with sources, drafts stakeholder updates, and routes approvals for recurring weekly reports.

### 90-second version

Design the assistant as a unified productivity layer for search, writing, and automation. Optimize reclaimed focus time while keeping verification and control explicit.