AI how-to guides Risk, compliance and governance
How to review contracts with AI
Load your contracts into an AI system that extracts clauses, obligations and key dates, compares them with your standard positions and flags what deviates. Lawyers review the flagged items, decide and negotiate. Start with one contract family and a short list of clause types, validate on real agreements, then widen.
What AI does
- Finds and classifies clauses (liability, termination, renewal, data protection, payment) across any contract type.
- Compares each clause with your templates and playbook and flags deviations, missing protections and unusual terms.
- Extracts obligations, renewal windows and deadlines into a register with owners and alerts.
- Compares contracts across the portfolio to spot inconsistencies, such as different liability caps with the same counterparty.
- Answers questions about the portfolio citing the exact clause and document it relied on.
What stays with people
- Legal judgement: whether a deviation is acceptable and what to do about it.
- Negotiation and the relationship with the counterparty.
- Defining the playbook: standard positions, fallbacks and red lines.
- Signing off before anything is sent or signed.
How to put it in place
-
Pick one contract family and its priority clauses
Choose a type with volume and a clear standard, such as NDAs, supplier agreements or customer terms. List the clauses and dates that matter most and the questions the business keeps asking about them.
-
Write the playbook down
For each priority clause, record your standard position, acceptable fallbacks and red lines, with example wording. This is what the AI will compare against, so vague rules give vague flags.
-
Ingest the contracts with access rules
Gather the agreements from shared drives or your CLM, including amendments and side letters, and fix scanned files that cannot be read. Keep the same access permissions the documents have today.
-
Validate with counsel on real agreements
Have lawyers check the extractions and flags on a sample they know well, including recent renewals and disputes. Adjust clause definitions and the playbook until misses and false alarms are acceptable.
-
Put it into the review flow and track dates
Route flagged contracts to the right reviewer with the clause, the source and the reason. Turn on alerts for renewals and obligations, and keep the register updated as new amendments arrive.
What you need
- Access to the contracts and their amendments, ideally from a CLM or a single repository.
- Readable documents: text-based PDFs or scans of enough quality to extract text.
- Your templates, standard positions and red lines for the priority clauses.
- Lawyers with time to validate the first results and own the playbook.
- A secure environment with access control, since contracts are confidential.
When it is not worth it
- A handful of bespoke, high-value contracts that lawyers will read in full anyway.
- When there is no standard position to compare against and nobody will define one.
- Large backlogs of poor-quality scans that cannot be read without costly digitisation.
Risks and how to control them
- Missed or misread clauses: validate on known contracts, show the source text for every flag and keep a lawyer reviewing the outcome.
- Invented answers about the portfolio: ground every answer in the contract text and require a citation to clause and document.
- Confidentiality and personal data in contracts: use an environment you control, respect existing permissions and apply data protection rules.
- Stale results as amendments arrive: re-process changes automatically and show the date each extraction was made.
- Treating flags as legal advice: the tool supports review, the legal decision stays with qualified people.
How Thinkia does it
We index your repositories, counterparties and priority clause types with retention and access rules, then encode standard positions, deviations and red-flag language against your templates. Counsel validates extractions and answers on real renewals and disputes before wider self-service.
Flagged items go to the right reviewer with context, priority and source location, and every extraction and flag is logged with its source and confidence. Insights then feed negotiation, procurement and risk workflows as contracts turn over.
The Thinkia solution behind this guide
- AI contract intelligenceYour legal team should be negotiating, not reading boilerplate.
- AI document intelligenceTurn unstructured documents into structured decisions.
Thinkia products involved
Frequently asked questions
Can AI review a contract on its own?
It can read it, extract the key terms and flag what deviates from your standard. Deciding whether a term is acceptable, and negotiating it, is still a job for a lawyer.
How accurate is AI at finding clauses?
It depends on the clause type, document quality and how clear your definitions are. That is why you validate on contracts your lawyers already know before relying on it.
Is it safe to upload contracts to an AI tool?
Only to an environment with enterprise terms, access control and no training on your data. Contracts are confidential and often contain personal data, so treat the tool like any other system that holds them.
Does it work with scanned PDFs?
Yes, if the scans are good enough to extract text. Poor scans need a digitisation step first, and that can change the effort of the project.
What is the difference from a CLM?
A CLM stores contracts and manages their workflow. AI reads what is inside them, so it complements a CLM by filling in clauses, obligations and dates that were never captured.
Keep exploring
Related guides
- How to manage company documents with RAG
- How to use AI in procurement
- How to monitor regulatory change with AI
Related decisions
- LLM evals vs human review: how should you assure the quality of enterprise AI?
- RAG vs fine-tuning: which one does your enterprise use case need?