Writing Effective Playbook Positions

The quality of your playbook depends on how well you write each position. Good positions lead to accurate AI reviews and actionable recommendations. Poor positions lead to confusion and missed issues.

This guide shows you how to write positions that work.

Every effective position answers three questions:

1. What should the AI look for? (The subject matter)
2. What’s acceptable vs. not acceptable? (Your standard)
3. What should the reviewer do if there’s an issue? (The action)

If your position doesn’t clearly answer all three, the AI will struggle to give useful results.

Check the limitation of liability clause.

This fails all three questions:

• What to look for? Vaguely defined.
• What’s acceptable? Unknown.
• What to do? Not specified.

Limitation of Liability

Our liability should be capped at the fees paid in the 12 months preceding a claim. Mutual caps are preferred.

Acceptable:

• Caps up to 24 months of fees
• Separate caps for different damage types, as long as direct damages are capped

Not acceptable:

• Unlimited liability for direct damages
• One-sided caps where only we are capped

If non-compliant: Flag for negotiation. Propose our standard mutual cap language. If counterparty insists on unlimited liability, escalate to Legal Director.

This works because the AI knows exactly what to evaluate and the reviewer knows exactly what to do.

The guidance field is the core of each position. Here’s how to structure it:

State what you want clearly and specifically:

Avoid subjective words like “reasonable,” “standard,” “appropriate,” or “acceptable” without defining what they mean.

Your preferred position won’t always be achievable. Tell the AI what compromises are okay:

Preferred: 12 months of fees

Acceptable with approval: Up to 24 months of fees (requires business owner sign-off)

Acceptable for enterprise deals: Up to 36 months for deals over $500K annual value

This helps the AI understand that “non-compliant” isn’t black and white—there are degrees.

Be explicit about what’s never acceptable:

Never acceptable:

• Unlimited liability for direct damages
• Indemnification for counterparty’s negligence
• Governing law of jurisdictions where we have no operations

Red lines should be truly non-negotiable. If you frequently grant exceptions to a “red line,” it’s not actually a red line—revise your guidance.

Tell the reviewer what to do when something doesn’t comply:

If cap exceeds 24 months: Negotiate to reduce. Propose our standard mutual cap.

If counterparty rejects: Escalate to Legal Director before accepting.

If clause is missing entirely: Add our standard liability clause.

Clear actions prevent reviewers from getting stuck.

Use when you have specific numeric limits:

Payment terms must be Net 30 or longer.

• Net 45 or Net 60: Acceptable
• Net 15: Flag for negotiation—propose Net 30
• Net 7 or payment on receipt: Escalate to Finance Director

Use when a clause must or must not exist:

The contract must include a data processing addendum (DPA) or equivalent data protection terms.

• If present and covers GDPR requirements: Compliant
• If present but incomplete: Flag gaps for negotiation
• If missing entirely: Non-compliant—add our standard DPA

Use when balance matters:

Confidentiality obligations should be mutual.

• Mutual obligations: Compliant
• One-sided (only we are bound): Non-compliant—propose mutual obligations
• One-sided (only counterparty is bound): Compliant (favors us)

Use when approval levels vary by risk:

Insurance requirements:

• Up to $1M per occurrence: Acceptable
• $1M–$2M: Acceptable with Legal review
• $2M–$5M: Requires Director approval
• Over $5M: Requires VP approval and risk assessment

Keywords are search terms that help the AI locate relevant clauses. Good keywords improve accuracy; bad keywords cause missed clauses or false positives.

Include:

• The common name of the clause type (“limitation of liability”)
• Variations in phrasing (“liability cap,” “cap on liability”)
• Key phrases that appear within these clauses (“shall not exceed,” “aggregate liability”)

For a Limitation of Liability position:

• “limitation of liability”
• “limit of liability”
• “liability cap”
• “cap on liability”
• “total liability shall not exceed”
• “maximum aggregate liability”

Example of negative keyword: If your “Limitation of Liability” position keeps matching insurance clauses (which also mention “liability”), add “insurance liability” as a negative keyword.

Run test reviews to check:

• Does the AI find the clause you expect?
• Does it find too many irrelevant clauses?
• Does it miss clauses that use different wording?

Adjust keywords based on results.

Fallbacks are pre-approved clause language. When the AI flags a non-compliant clause, reviewers can propose your fallback instead of drafting from scratch.

Title clearly: Name each fallback so reviewers understand when to use it.

• “Standard Mutual Liability Cap - 12 Months”
• “Alternative Liability Cap - 24 Months (Requires Approval)”
• “Minimum Acceptable - Cap at Total Fees”

Include complete language: Fallbacks should be ready to insert. Don’t use placeholders like “[PARTY NAME]“—the AI will adapt terminology when generating.

Order by preference: Put your preferred fallback first, acceptable alternatives second, minimum acceptable last.

For complex positions, include several options:

Fallback 1: Preferred

The total aggregate liability of either party… shall not exceed the fees paid in the twelve (12) months preceding the event…

Fallback 2: Acceptable Compromise

The total aggregate liability of either party… shall not exceed the fees paid in the twenty-four (24) months preceding the event…

Fallback 3: Final Position

The total aggregate liability of either party… shall not exceed the total fees paid or payable under this Agreement.

This gives reviewers negotiation flexibility while keeping them within approved bounds.

Before deploying a playbook, test each position:

1. Run reviews on several real contracts
2. For each position, check:
   • Did the AI find the right clause?
   • Is the compliance assessment correct?
   • Does the guidance make sense in context?

Ask someone unfamiliar with the playbook to review a flagged issue:

• Do they understand what’s wrong?
• Do they know what to do next?
• Can they explain it to a counterparty?

If not, simplify your guidance.

Test with contracts that are:

• Missing the clause entirely
• Written in unusual language
• From different industries or jurisdictions

See how the AI handles variations.

Positions need maintenance as your standards evolve:

• Negotiation outcomes consistently differ from guidance
• New risks emerge (new regulations, new business models)
• Reviewers frequently override the same assessment
• Feedback indicates guidance is unclear

1. Create a new version of the playbook (never edit a published version directly)
2. Revise the position
3. Test against recent contracts
4. Publish when confident

Add notes explaining why you changed a position. Future editors will thank you.

Before publishing a position, verify:

• Guidance states the preferred position clearly
• Acceptable alternatives are defined
• Red lines are explicit
• Actions are specified for each scenario
• Keywords cover common phrasings
• At least one fallback is provided
• Tested on real contracts
• Another person has reviewed for clarity

• Creating Your First Playbook — Build a complete playbook
• What is a Playbook? — Understand the concepts
• Quick Start: Reviewing Contracts in Word — See positions in action