Negotiation
How to Use AI to Screen a Contract Before You Negotiate
June 21, 2026
You cannot negotiate a contract well until you know what is in it and what is wrong with it. The problem is that the clauses that matter most, the liability cap, the price-adjustment mechanism, the step-in rights, the auto-renewal, tend to hide in dense language a dozen pages apart, and reading for them line by line is slow enough that most buyers do it partially and hope. AI is genuinely strong here. It reads the whole document in seconds and gives you a structured, risk-oriented first pass, so you walk into the internal review and then the negotiation knowing exactly where the problems are.
One boundary before the workflow. This is a buyer's risk screen, not legal advice, and it is not a substitute for it. A second boundary matters just as much: a supplier contract is confidential business information, so before you paste one into any AI tool, follow your organisation's data rules. Prefer an approved, private enterprise instance over a public consumer chatbot, redact party names, pricing, and identifiers where you can, and never upload a counterparty's contract to a tool that may retain it or train on it. Treat the output as a fast way to find the clauses that deserve attention, verify anything it asserts against the actual text, and send the flagged clauses to Legal. Held to that, the three steps below do a large amount of the preparation for you.
Step one: the fast screen
Start with a first pass that tells you what kind of contract you are holding and where the obvious risk sits.
You are an experienced contract-review assistant for procurement. Analyse the following contract from the perspective of a commercial buyer. The goal is a fast first-pass screen to prepare an internal review. Give me, in structured bullet points: the contract type and its typical risk areas; any missing or unusual standard clauses; any liability or risk shifted onto the buyer; the clauses that must be reviewed with Legal; and five specific questions procurement should ask internally. This is a risk-oriented buyer's perspective, not legal advice.
In under a minute this turns forty pages into a short list of what to look at and what to ask, which is often enough to decide whether the contract needs a light touch or a serious negotiation.
Step two: compare against your own position
A screen tells you what the contract says. The more valuable question is how far it sits from what you are willing to accept. If you have a reference or framework agreement with your preferred and fallback positions, put the two side by side.
You are a structured contract-comparison assistant for procurement. Document A is the supplier's contract to be reviewed. Document B is our buyer reference agreement, with our preferred and fallback positions marked. For each critical clause area, payment terms, delivery and Incoterms, late delivery and penalties, liability and its cap, termination, force majeure, IP and indemnity, security of supply and step-in rights, price changes, and export and compliance, compare Document A against Document B and classify it as "matches preferred", "matches fallback", "below fallback", or "no-go from the buyer's side". For each, give the risk impact in one or two sentences, the likely negotiation direction, and a priority from one to five. At the end, list the top five negotiation levers and an overall risk rating of low, moderate, high, or critical.
This is the step that turns a review into a negotiation plan. You come out of it with each clause labelled, a clear view of your no-go items, and a ranked list of the levers worth spending your credibility on.
Step three: a structured, repeatable analysis
For teams handling many contracts, or anyone who wants a consistent record rather than a one-off read, ask for output in a fixed structure.
Produce a structured, repeatable analysis. Give me an executive summary of no more than 250 words, then a risk register where each entry has the clause area, a short risk description, a severity and likelihood score from one to five, a negotiation priority from one to five, and whether it requires legal review. Finish with an overall risk score and the three biggest risk drivers. Mark anything you cannot determine from the text as "not clearly identifiable", and never present an assumption as a fact.
A fixed structure is what makes the analysis reusable: it reads the same way across every contract, it can be compared over time, and it can be handed to a colleague or fed into a contract-management system without reformatting.
The rules that keep it safe
The screen is only trustworthy if you hold it to a few rules, and they are worth stating to the model explicitly: no binding legal advice, no assumptions presented as facts, and anything the text does not clearly support flagged as uncertain rather than filled in. An AI that invents a clause you do not have, or misreads a liability cap, is worse than no screen at all, so the discipline of verifying against the source is not optional.
Where the screen hands off to the room
Done well, this gives you something precise: the risky clauses named, classified against your position, ranked by priority, with your negotiation levers and no-go lines already drawn. That is most of the analytical preparation for a contract negotiation, and it is exactly the work AI should be doing.
It is not the negotiation. The supplier will defend the liability cap, justify the price-adjustment clause, explain why the step-in rights are standard, and push back the moment you ask to move any of it. Your ranked list of levers does not win a single one of those exchanges. That happens in the conversation, under the same real-time pressure this series keeps returning to, and it is where a well-screened contract is either renegotiated or quietly signed as-is. Voice2Evolve is where you rehearse that part: you take the levers the screen handed you and practise arguing them out loud against a counterpart who defends the clause, questions your reasoning, and holds the line, so the concessions you identified on paper are ones you can actually win in the room. Let AI find what to negotiate, then rehearse until you can negotiate it.
Negotiation · Read
Where Principled Negotiation Stops Working
Fisher and Ury's framework transformed how people think about negotiation. It also assumes conditions — mutual interest in agreement, separable people and problem, discoverable objective criteria — that adversarial procurement situations often do not have.
When Anchoring Runs Wrong
The Ackermann method assumes you can walk away. In a negotiation where you cannot, the anchor you open with becomes a debt that has to be repaid — to the supplier, and sometimes to your own organisation.
Why AI Negotiation Practice Has to Be Spoken, Not Typed
Typing to a bot and speaking to one are not the same exercise. Voice removes the edit buffer that text quietly gives you, and that single difference is what makes spoken practice transfer to a real negotiation.
Price-Adjustment Clauses: Caps, Floors and Collars
In a multi-year deal, the price-adjustment clause decides how much the price climbs over time. It reads like a formula, which is exactly why buyers accept it, and why years of margin erosion hide inside it.
Train the moment, not the theory.
Voice2Evolve puts you in the scenario repeatedly until your reaction under pressure is no longer panic.