How to review supplier contract terms before signature with an AI agent

Before a supplier contract is signed, finance wants to know what it costs over its life and what could go wrong with the money. The agent reads the draft against your contracting playbook, flags payment terms, price increases, auto-renewals, minimum commitments, liability limits and termination costs, and proposes redlines. Procurement, legal and finance decide; a person with signing authority signs.

Why should finance review supplier contracts, and who owns the decision?

Procurement negotiates and legal protects, but many supplier contracts commit money in ways neither checks closely: payment in advance, annual increases tied to an index with no cap, minimum volumes, automatic renewal for another term unless notice is given months ahead, early termination fees, currency risk. These terms decide the true cost and are hard to change after signature.

Procurement owns the negotiation. Legal owns the contract terms and redlines. Finance owns the commercial and accounting view: whether the cost fits the budget, whether payment terms meet policy, whether the contract creates a commitment or an asset that must be accounted for. The signatory owns the decision to sign, under the delegation of authority. The general bounds are in AI agents for finance. Whether an agent could sign a contract itself is a separate question with a clear answer: not in a well-run business.

What does the agent prepare?

1. Agent, depth 1: reads the draft and related documents. The supplier's paper or your template, order forms, pricing schedules and service levels.
2. Agent, depth 1: compares with the playbook. Your contracting playbook sets preferred, acceptable and unacceptable positions for each term.
3. Agent, depth 2: summarises the money. Total contract value over the full term including renewals, payment schedule, price change mechanisms with a worked example, minimum commitments, termination costs at each point, and currency.
4. Agent, depth 2: flags risks. Terms outside the playbook, uncapped increases, auto-renewal with long notice, payment in advance without protection, liability caps below what the supplier could cost you, missing data protection terms where personal data is involved.
5. Agent, depth 2: proposes redlines. Using your playbook's fallback language, marked as proposals.
6. Person, procurement: decides commercial positions. What to push back on, what to accept.
7. Person, legal: approves redlines and final terms.
8. Person, finance: confirms budget and accounting. Including whether the contract creates a lease or other item finance must record.
9. Person, signatory: signs.

What do the named people approve?

Procurement approves commercial positions; legal approves the words; finance approves the budget fit and accounting treatment; the signatory approves the commitment. The agent's output is an annotated draft and a one-page money summary. It never sends redlines to the supplier, because what you put to a counterparty in negotiation is a position you may be held to.

Playbook version. Evidence: the playbook version on each review.

Money summary attached to approval. The signatory sees lifetime cost, not just the first-year price. Evidence: the summary in the approval record.

Redlines approved before sending. Evidence: legal's approval on the version sent.

Renewal terms recorded. Notice periods and dates go into the contract register at signature. Evidence: the register entry.

What does a flagged term look like?

Take an illustrative three-year software subscription at 60,000 a year. The agent finds an annual increase tied to an inflation index with no cap, automatic renewal for further three-year terms unless notice is given 180 days before expiry, payment annually in advance, and a liability cap equal to one month's fees. The money summary shows a six-year cost under two assumed index paths, the notice deadline as a date, and the gap between the cap and the cost of replacing the service. Procurement negotiates a cap on increases and a shorter notice period; legal accepts the rest.

Where does it usually go wrong?

Reviewing the first-year price only. The real cost is in renewals and increases. The money summary must cover the full possible term.

Auto-renewal missed after signature. The notice date passes unnoticed and the contract renews. Feed every notice date into a register and track contract renewal dates.

Supplier paper not compared. The agent reviews against your template when the supplier sent its own. Compare terms by meaning, not by clause number.

Different questions for sales contracts. On the sales side the concern is revenue timing, covered in reviewing sales contracts for revenue recognition terms. Do not reuse a sales review for a purchase.

Data protection terms skipped. Where the supplier will process personal data for you, the contract needs terms covering it, and a data protection impact assessment may be needed under GDPR Article 35 or its UK equivalent if the processing is high risk. The agent should flag the gap and route it to your data protection lead.

Playbook out of date. The agent applies positions legal no longer holds. Review the playbook periodically and version it.

What should you measure?

• Contracts reviewed before signature, against all contracts signed.
• Flags accepted, negotiated away or waived.
• Unplanned renewals after signature.
• Lifetime cost against first-year price for new contracts.

Where Pryme fits

In Pryme, comparing drafts with the playbook, summarising cost and proposing redlines are separate capabilities, each certified on your own past contracts and the positions legal took on them. The playbook arrives as a signed source from legal and the supplier's draft as external, so a proposal always shows which side of the table it came from. Sending and signing sit outside the agent's role scope, and every review is filed.