Build an AI contract playbook so the model negotiates your way
Write your positions down once, in a form a model can actually apply.
In short
An AI contract playbook records your preferred, fallback, and walk-away position for each clause type, in language a model can apply directly.
- Start with one high-volume agreement type instead of covering everything at once.
- Write rules that are testable: a cap of at least one million, not "reasonable".
- Mark the clauses that must escalate, so the model flags instead of deciding.
Positions on paper, in machine-readable form
An AI contract playbook is the document that turns your judgment into something a model can apply. Most legal teams already hold these positions, just informally: everyone knows which liability cap is fine and which indemnity needs a partner. Writing them down for a model takes one extra discipline, which is that every rule has to be testable. 'Reasonable notice' means nothing to a model; 'at least 30 days written notice' can be checked against any draft. Cover each clause type with your preferred language, the fallbacks you will accept in order, and the terms you refuse, then mark which ones must escalate. Start with your highest-volume agreement and leave the rest of the library for later. It then feeds directly into AI contract redlining and AI contract review. It is a foundational piece of AI for legal.
How to build an AI contract playbook
One agreement type, testable rules, then the escalation markers.
- 1
Pick one high-volume agreement type
Your standard vendor or customer contract. One type done properly beats a library half-covered.
- 2
List the clauses you actually negotiate
Liability, indemnity, termination, payment, IP, data protection, notice. Skip the boilerplate nobody argues about.
- 3
Write three positions per clause
Preferred language, the fallbacks you accept in order, and the term you will not take. Rank the fallbacks explicitly.
- 4
Make every rule testable
'A cap of at least 1 million' can be checked. 'A reasonable cap' cannot. Numbers and named conditions, not adjectives.
- 5
Mark what must escalate
Flag the clauses where a person decides. The model should surface those instead of applying a fallback.
- 6
Update it from real outcomes
When a position is repeatedly conceded or refused, change the playbook. A stale playbook teaches the model last year standard.
Watch-outs
- Adjectives do not survive contact with a model. Write thresholds and named conditions instead.
- A playbook that covers everything badly is worse than one agreement type covered properly.
- Review it on a schedule. Regulation and commercial norms move, and the file will not tell you.
Common questions
What is an AI contract playbook?
An AI contract playbook is a written record of your negotiating positions in a form a model can apply: for each clause type, the preferred language, the fallbacks you accept in order, the term you refuse, and whether it must escalate to a person.
How is it different from a normal playbook?
The content is the same judgment; the discipline is that every rule has to be testable. A model cannot act on "reasonable notice" but can check "at least 30 days written notice" against any draft, so vague standards get rewritten as thresholds.
Where do I start?
Start with the one agreement type you sign most often, and only the clauses you actually negotiate. Covering a single high-volume contract properly delivers more than a thin pass across your whole template library.
How often should it be updated?
Whenever a position is repeatedly conceded or refused in real negotiations, and on a fixed review cycle besides. A playbook that has not moved in a year encodes last year market standard and will quietly push the model toward it.
See how you actually work with AI
The AI Skills Quiz scores your real habits, including whether you give AI standing rules to work from. Free, and it takes a couple of minutes.
Professionals from these companies build their AI skills with Candova
The future belongs to AI‑native companies.
Make sure yours is one.
We make your people AI-fluent so your company becomes AI-native, with Cando, a personal AI trainer for every employee. Start with Candova AI today.
Free to start · Productivity gain guarantee · Built for your real job