Contract review is one of the highest-value, most time-intensive parts of legal work. First-pass review of a standard commercial agreement — checking against playbook positions, flagging non-standard clauses, summarising key terms — typically takes two to four hours for a junior lawyer and one to two hours for a senior. AI reduces first-pass time to twenty to thirty minutes. For a team handling fifty contracts a month, that's a structural change.
What AI is actually good at in contract review
AI performs well on pattern matching: identifying clauses that deviate from a standard position, comparing against a redline playbook, summarising non-standard positions across multiple documents. It also performs well on extraction tasks: pulling defined terms, key dates, and obligations from dense legalese, and formatting them into a usable summary.
What it's not good at
AI performs poorly on judgment tasks: deciding whether a deviation from a standard clause is acceptable given the specific commercial context, weighing legal risk against business need, or advising on strategy in a negotiation. This distinction matters because the value of a lawyer isn't in reading the contract — it's in knowing what to do about what they find.
The workflow that holds up in practice
- Run the contract through an AI review tool to flag non-standard clauses and produce an issues summary
- Review the AI summary against your playbook — this takes 15 to 20 minutes instead of two hours
- Apply judgment to each flagged issue: accept, push back, or escalate
- Draft the redline using AI assistance, then review before sending
- Document AI use per your firm's or company's AI governance policy
On risk and governance
The legal profession's instinct to be cautious about AI is appropriate — not because the technology isn't useful, but because the stakes of getting it wrong are high. Teams with clear AI use policies — addressing confidentiality, review requirements, and disclosure — use AI more confidently, not less. Governance enables adoption; the absence of it prevents it.
The competitive divide in legal is increasingly between those who use AI to recover time for judgment — and those who still spend that time on tasks a model could do in minutes.