Reviewing a contract

Task boundary

Reviewing a contract is assessed as a specific task and workflow, not as a general endorsement of AI for the surrounding job. AI may assist issue spotting only after the task is reframed away from legal decision ownership.

Assumptions

• The original task asks AI to determine whether contract terms are acceptable or what legal position to take.
• Errors may create material and difficult-to-reverse obligations.
• A qualified legal owner—not the model or general operator—must interpret and accept the terms.

Decision

• Disposition: Redesign
• Pattern: None for the task as currently framed
• Inherent risk: high

Rules that drive the decision

• High consequence plus non-delegable professional judgment blocks the task as framed.
• Supplying the contract as context improves grounding but does not transfer legal responsibility.
• The State Bar of California guidance describes professional responsibilities for lawyers using generative AI; it is not a universal legal rule, and the qualified owner must identify the obligations that apply.

Why alternatives were rejected

• A model cannot replace qualified legal responsibility.
• Autonomous acceptance is irreversible and high consequence.

Next safe step

• Separate clerical extraction or issue-spotting from legal interpretation and acceptance.
• Define any AI output as a non-authoritative draft tied to exact clauses, with uncertainty and omissions visible.
• Have qualified legal counsel set the scope, review every material item, and own the final decision before reassessing the narrower task.

Reassess the changed task

• Reassess only the newly bounded support task, with its reviewer qualifications, evidence, and acceptance checks specified.
• Stop again if the workflow recommends acceptance, negotiates terms, or acts without qualified legal ownership.