Gains:
- Ability to systematically identify high-risk items such as liability, compensation, termination and confidentiality
- Ability to classify items as accept/negotiate/reject with a playbook containing corporate standard positions
- Ability to question the protective clauses that should be in the contract but are missing
Not all clauses of a contract are equal. While the agreement is on its way, no one looks at the articles; But when a dispute, delay or damage arises, everything boils down to a few critical items: who will pay how much compensation, where is liability limited, how to terminate the contract, how long confidentiality lasts. Artificial intelligence (AI) is very fast at systematically finding these high-risk substances and putting them in front of you. But the real power comes when you combine AI with your organization's playbook. In this unit, we will first learn where to look for risk and then how to have AI classify it according to your corporate standards.
Let's clarify the terms. Playbook is the written version of the "acceptable limits" and "preferred text" that the institution predetermines for each typical contract clause; It ensures that everyone defends the same line in negotiation. Limitation of liability is a clause that caps the maximum compensation to be paid by a party. Indemnity/indemnity is the commitment of one party to protect the other against certain damages and to compensate for the resulting damage. Termination is the conditions for terminating the contract before its expiration date. Force majeure is the suspension of obligations due to extraordinary events (earthquake, war, epidemic) beyond the control of the parties.
Substances Where Risk Concentrates
In commercial contracts, risks are concentrated in certain clauses. If you give the AI this map, it won't waste its attention:
- Limit of liability and exclusions. Is there a ceiling, how much, what damages does it cover, which situations are exempt from the ceiling?
- Indemnity commitments. Who indemnifies whom and against what demands; Is the scope one-way or reciprocal?
- Dissolution. Termination conditions with and without just cause, notice periods, post-termination obligations.
- Privacy and data. Duration, scope, personal data processing, return/destruction obligation.
- Intellectual property. Who owns the rights, scope of license, third party infringement.
- Payment and penalty requirement. Delay interest, penalty clause rate, right of offset.
- Applicable law and dispute resolution. Competent court or arbitration, applicable law.
- Transfer, change, unilateral rights. Unbalanced powers granted to one party.
Your role: a legal expert who assesses contract risks.Task: Identify high-risk clauses in the following contract.Table for each determination:| Risk clause | Finding | Why is it risky | Article number and quote | Risk level |Focus: limit of liability, compensation, termination, confidentiality, intellectual property, penal clause, applicable law, unilateral powers. Mark the risk level as high/medium/low. Write a determination without the article number and short quote; If you cannot find a topic, write "no relevant article".
Caution: The AI bases its general training knowledge when deeming a risk "high" or "low"; It is not aware of your sector, position (buyer or seller) and risk appetite. Always re-evaluate the risk level label with your own context.
Classification with Playbook
Raw risk assessment is useful but does not make decisions. The real value is in giving your organization's standard positions to AI and having each item evaluated "along our lines." If you embed your playbook in the prompt, the AI becomes your negotiation assistant rather than a general commentator.
Below is our institution's contract playbook. Evaluate the contract I gave you according to this playbook. For each relevant clause:- Playbook rule- Current status in the contract (with quote)- Decision: ACCEPT / NEGOTIATION / REJECTION- Proposed amendment text if negotiation is required<playbook>1) Liability ceiling: must not exceed the price paid in the last 12 months. Unlimited liability is accepted only in cases of intent/gross negligence and breach of confidentiality.2) Compensation: must be mutual; Unilateral compensation is rejected.3) Termination: we should be given the right to terminate without reason for at least 30 days.4) Payment: delay penalty should not exceed 2% per month.5) Applicable law: Turkish law and Istanbul courts are preferred.</playbook><sozlesme>[contract text]</sozlesme>
What is not written in the contract is as much a risk as what is written in it. It is a necessary but missing protective agent, a silent trap.
In this contract, list the safeguards that are expected to be in such a [supply/service] agreement but are NOT included in the text. For each: why is it necessary, what risk does its absence create, short proposed clause text. Examples: data processing annex, data return after termination, right to audit, subcontractor approval, insurance obligation.
Weak Prompt / Strong Prompt
Weak prompt:Is this contract risky?
Result: A vague answer like "Some items may require attention" that doesn't tell you which item to look at or what to do. It does not turn into a decision.
Powerful prompt: [risk item map + table format + item number/citation requirement + institution's playbook + ACCEPT/NEGOTIATE/REJECT decision + missing item analysis]
The result: An actionable output that labels each item according to your institution's line, proposes negotiation text, and shows missing protections.
Anatomy of a Playbook Line
component
Description
example
Subject
Which item?
Liability ceiling
ideal position
what we want most
50% of the annual fee
acceptance limit
Can't go beyond
1 times the annual fee
red line
Never accepted
unlimited liability
fallback text
Alternative substance to be suggested
ready paragraph
justification
Why this limit?
fuse capacity
Three Mini Cases
Case 1 — The unlimited liability trap. A software company was about to sign the service contract sent by its customer. The AI risk scan caught and cited that in the liability clause the cap was not only lifted for “data breach” but was also left unlimited for “any consequential damage”. According to Playbook, this was a red line. Liability in the negotiation was limited to 1 times the annual price; The company has closed a gap risk that could potentially reach 6 times its annual turnover.
Case 2 — One-way compensation. A supply contract had a structure in which only the buyer compensated the seller, and not the other way around. AI marked this as "NEGOTIATION" according to the "compensation must be mutual" playbook rule and suggested mutual compensation text. The human lawyer refined the proposal; As a result, third-party intellectual property claims were also imposed on the other party. The suggestion came from AI, the decision came from the lawyer.
Case 3 — Incomplete right to control. A compliance team discovered through AI's missing clause analysis that a "right to audit" clause was missing from a contract with a service provider handling sensitive data. Of the portfolio of 30 contracts, 22 had the same deficiency. The team added a standard audit clause and made it mandatory for renewals; in a regulatory review, “can you audit the supplier?” They could now answer yes to the question.
Common mistakes
- Risk screening without a playbook. If the corporate line is not given, AI makes general comments; does not support your decision.
- Blindly trusting AI's risk label. The "high/low" tag doesn't know your position; Weigh again with your own context.
- Just looking at written items. Missing preservatives are the quietest risks; Conduct a separate missing item analysis.
- Accepting the determination without quotation. If there is no article number and citation, the determination cannot be verified; It could be a hallucination.
- Using the negotiation text as is. The alternative article proposed by the AI is a draft; must be approved by a lawyer.
- Generalizing the playbook to a single contract. Acceptance limits may vary depending on the sector and counterparty; Keep the playbook updated.
In summary
The risk is concentrated in a few critical clauses of the contract. AI finds these items quickly; But the real power comes from combining it with your organization's playbook and placing each item on the "ACCEPT / NEGOTIATE / REJECT" line. Ask questions about missing items as well as written items, link each observation to a quote, and consider the negotiation texts suggested by the AI as an initial draft. AI signals risk and generates options; The competent professional decides what risk you will accept and how you will negotiate.
Application task
Write a short 5-line playbook for your organization (or example) (liability, compensation, termination, payment, authority). (1) Scan a contract with the risk table prompt. (2) Have the same contract classified as "ACCEPT/NEGOTIATE/REJECT" in your playbook. (3) Run missing item analysis. (4) Prepare the proposed text for at least 2 articles that are "NEGOTIATED", making notes for a lawyer to review.
checklist
- [ ] Have high-risk substances been identified by item number and citation?
- [ ] Has the institution's playbook been added to the prompt?
- [ ] Has each relevant item been classified as ACCEPT/NEGOTIATE/REJECT?
- [ ] Were the missing items (which should be there but are not) questioned?
- [ ] Have AI's risk labels been re-evaluated in your context?
- [ ] Have the proposed negotiation texts been marked as "legal approval required"?
- [ ] Have decisions and justifications been recorded in a traceable manner?