What “Supervised Use” Means (and Why It Matters)
In legal practice, supervision is not optional. AI can speed up work, but it can also amplify mistakes if no one is clearly accountable.
The supervision principle (plain‑English)
Supervision means: a qualified human reviews the AI‑assisted work, verifies key claims, and takes responsibility for the final output.
For non‑attorney staff, the goal is to make attorney review easy and defensible—by documenting what you did and what you verified.
A workflow you can reuse
{{UPLOAD_ASSET:supervision_workflow.png}}
Key idea: The workflow is designed to prevent two failures: (1) using the wrong tool with the wrong data, and (2) trusting unverified output.
Attorney vs. staff responsibilities
| Step | Staff role | Attorney role |
|---|---|---|
| Task intake | Clarify scope; identify sensitive facts | Approve scope; set risk tier |
| AI run | Use approved tool; log prompts/outputs | Set expectations; approve templates |
| Verification | Spot‑check facts; flag uncertainty | Independent verification for high‑risk tasks |
| Release | Never send client‑facing work without sign‑off | Own final judgment; communicate to client/court |
Activity: build an escalation rule
Write one sentence your team can adopt, e.g.:
Escalation rule: If AI output affects legal strategy, privilege, client communication, or court filings, escalate to the supervising attorney before the draft leaves the team.