01 Define
Name the finished output
Record the trigger, contract type, represented side, governing law, expected output, people involved and exception route. A broad instruction to review the contract is not a stable test.
AI contract review where your lawyer makes every call.
AI contract review is software that suggests where a contract departs from a set position. At a law firm, a lawyer then confirms, corrects or rejects each suggestion.
Operational guidance for law firms in England and Wales. No contract upload and no legal advice.
Measure the review before evaluating the AI.
Contract-review technology becomes easier to judge when the firm can describe the incoming work, the approved position, the exception route and the qualified review needed to reach a usable result.
The operating sequence
Review is more than reading clauses.
The work begins when an agreement arrives and ends only when the appropriate lawyer has accepted the result, dealt with exceptions and retained the evidence behind the decision.
02 Measure
Count completed work
Measure the base pass, correction time, escalation and qualified approval. Timing an initial AI response alone ignores the work needed to make that response usable.
03 Decide
Compare the four routes
Fix the current process, buy an existing product, run a small supervised pilot or leave the workflow alone. Compare each route against the same baseline and governance needs.
Playbook readiness
Give the reviewer a position they can actually apply.
A playbook is not a folder of precedents. It needs enough structure to tell a reviewer what matters, which position applies and when the issue must move to a person.
Define the scope
- contract type and represented side
- jurisdiction and law-as-of date
- approved positions and fallback language
- materiality and escalation thresholds
- source and approval history
Test the exceptions
- unusual structures and missing clauses
- conflicting provisions and cross-references
- counterparty wording outside the examples
- issues requiring commercial judgment
- evidence the reviewer must retain
Who decides each finding?
In AI contract review, the software suggests and a lawyer at the firm decides. Each finding should show the clause it relies on and the firm position it was checked against. A named lawyer should then confirm, correct or reject it, and the firm should keep a record of who decided and when. A second AI model checking the first does not replace that step. As the Law Society's practice note of 21 October 2026 puts it, "Relying on AI to verify AI outputs will not satisfy your professional obligations if there are inaccuracies." The SRA's warning notice of 17 August 2026 says, "The use of AI does not diminish or transfer your professional responsibilities." Margo is being built to record who in your firm confirmed or corrected each finding.
Evidence before a pilot
A representative test should be capable of failing.
Use made-up, properly redacted or approved historical material that preserves the variation of the real task. Put the expected issue, source evidence and reviewer standard beside every example.
Compare like with like
Measure completed and supervised output under the current process and the proposed route. Include correction, source checking and escalation in both totals.
Agree the stop condition
Stop for unsupported material output, prohibited data use, unacceptable correction time, missing evidence, review-capacity failure or operating cost outside the agreed boundary.
Keep a named owner
One person owns the decision. Lawyers doing the work, the qualified reviewer, sponsor, risk, data and IT roles contribute without dissolving accountability.
Independent decision route
Margo Legal starts without a product answer.
Margo Legal recommends no product by default. The tools, available at no charge, help your firm map one workflow, set a baseline and compare four routes: fix the process, buy a tool, run a small supervised pilot or leave it alone. Buying a product is only one possible outcome.
Score one workflowSeparate product route
Margo is a controlled-development preview.
Margo is being built to check contracts against your firm's own positions, with a lawyer at your firm confirming, correcting or rejecting every finding. Many AI tools can now mark up a contract against a playbook. Margo is being built to record who in your firm confirmed or corrected each finding. Each review will be set to the side you act for and checked against your firm's positions as at a date you state. Margo is in development and is not authorised for confidential live client matters. Margo Legal is not a law firm and does not give legal advice.
Inspect Margo's statusStart with the current work
Build the baseline before choosing the route.
Use the calculator for a first estimate, then freeze the five measures and score the workflow against recurrence, ownership, approval, data and review capacity.
Put the answer to work
What should we record before changing a document-review workflow?
Trace one document from request to final record and isolate delay, rework and legal-review points. Map facts, precedent, drafting, review and approval as one route. Record the context, permission and source checks rather than relying on fluent output.
These optional PDF worksheets are available without an email address or client data. They are operational aids, not legal advice.
- Modernising Legal Document Workflows (PDF): A practical map for efficiency, accuracy and controlled change.
- Context-Aware AI for Lawyers (PDF): A boundary guide for source material, instructions, retrieval and review.
- Legal Document Drafting Automation Guide (PDF): A vendor-neutral route for templates, clause logic and review.
Build the contract-workflow baseline, or browse all 63 ungated guides.
Questions about AI contract review
Can AI review a contract?
AI can suggest where a contract departs from a firm's positions and point to the clause behind each suggestion. It cannot take responsibility for the review. A lawyer at the firm still checks each suggestion against the source and decides whether to confirm, correct or reject it.
How should a lawyer check AI contract review findings?
Open the source clause beside each finding, compare it with the firm's written position for the side it acts for, and record whether the finding is confirmed, corrected or rejected. The Law Society's practice note of 21 October 2026 says relying on AI to verify AI outputs will not satisfy professional obligations if there are inaccuracies.
Who is responsible if AI contract review gets something wrong?
The firm and its lawyers remain responsible, whichever tool produced the first pass. The SRA's warning notice of 17 August 2026 says, "The use of AI does not diminish or transfer your professional responsibilities."
What is a contract review playbook?
A contract review playbook is a firm's written set of positions for one contract type and the side it acts for, with preferred wording, acceptable fallbacks and the point at which an issue goes to a named person. AI contract review checks a contract against that playbook, so the playbook sets the standard.
Can ChatGPT review a contract?
A general chat tool can comment on contract text, but it does not know your firm's positions unless you supply them, and what happens to the text depends on the tool's terms and settings. Our guide, Can ChatGPT review a contract?, sets out the checks.
Is Margo available for live client contracts?
Margo is in controlled development and has been tested only with synthetic material. It is not authorised for confidential live client matters. Live use depends on hosting, an AI provider, security testing and contracts that are not yet in place.
Two-minute self-check