What a playbook actually is (and what it isn’t)

Ask five partners what their firm’s playbook is and you will get five answers. One will point to a folder of precedents. One will describe a style guide. One will say it lives in the head of a senior associate who has been doing the same kind of deal for eleven years. One will produce a forty-page PDF nobody has opened since 2021. One will say, honestly, that there isn’t one.

All five are describing something real, but only one of them is describing a playbook in the sense that matters when a model is going to read drafts against it. This guide is about that sense.

What a playbook is

A playbook is a structured statement of positions. For each clause type that matters to the practice, it answers three questions:

1. What do we want? The preferred position. For a limitation of liability in a services MSA, that might be a cap at twelve months’ fees, excluding consequential loss, with carve-outs for confidentiality, IP infringement and data protection.
2. What will we accept? The fallbacks, in order. Eighteen months’ fees. Then twenty-four. Then a cap at the contract value if the counterparty is a regulated entity and the term is under three years.
3. What will we never accept? The red lines. Uncapped liability. A cap that excludes the data protection indemnity.

That is the whole of it. Everything else, severity, explanation, who to escalate to, is metadata that hangs off a position.

A playbook written this way is readable by a junior associate on her first day and by a model on its first pass. Neither needs to infer what the firm thinks. It is written down.

What a playbook is not

It is not a precedent bank. A folder of signed agreements tells you what was agreed, not what was wanted. Every signed contract is a compromise, and the compromise is invisible unless someone wrote down where the firm started. Precedents are useful for drafting and for Recall; they are poor input for review because they contain the other side’s wins as well as yours.

It is not a style guide. “We write ‘shall’ not ‘will'” and “defined terms in bold” are house style. They matter, and Draft learns them, but they are not positions. A review that flags the use of “will” at High severity is wasting a partner’s attention.

It is not the senior associate’s memory. It might be derived from it, and the fastest way to write a first playbook is to interview that person for two hours. But a playbook that exists only in someone’s head cannot be applied consistently, cannot be audited and leaves with them.

How to write one that a model can read

In our experience across the firms that onboarded in 2024, the playbooks that worked had a few things in common. These translate directly into a checklist.

• One clause type per entry. Do not write “indemnities and liability” as one block. Indemnity scope, indemnity procedure, liability cap and liability exclusions are four entries.
• Positions as conditions, not prose. “Cap should be reasonable” is not a position. “Cap at or above 12 months’ fees” is. If the position depends on something, name the thing: deal value, term length, counterparty type, governing law.
• Fallbacks in order, each one explicit. The order is the negotiation path. If a fallback is only available with partner sign-off, say so in the entry; that becomes an escalation flag.
• Red lines stated as red lines. Not “we strongly prefer to avoid”. A red line is a flag at High severity with a note that there is no fallback. If there is in fact a fallback, it was not a red line.
• Severity set per entry. A missing fallback is not the same weight as a missing survival clause. See the Review 3.6 note on severity tiers for how the tiers resolve.
• Date and owner on every entry. Positions drift. An entry last touched in 2022 by someone who has left should be reviewed, not trusted.
• Keep it short. The best playbooks we have seen run to 60 to 120 entries for a full commercial practice. Beyond that, entries start to overlap and conflict, and the model, like the associate, has to guess which one wins.

Where to start if you have nothing

Do not try to write the whole thing. Take the last ten drafts your team sent back to a counterparty with redlines. For each redline, ask: what position were we enforcing? Write that position down as an entry. Ten drafts usually yield twenty to thirty entries, and they are the entries that actually come up.

Then take the last three deals where the firm conceded something it did not want to concede. Those are your fallbacks. Write them in the order they were offered.

Finally, ask the practice head for the three things the firm would walk away over. Those are your red lines. There are rarely more than five.

That is a working playbook. It will be incomplete, and the first month of review flags will tell you exactly where. Each flag that a reviewer dismisses as “we don’t care about that” is a position to loosen; each issue a reviewer catches that the model missed is an entry to add.

A note on conflicts between entries

Two entries will eventually contradict each other. A general entry says caps must be at least twelve months’ fees; a specific entry for software licences says six months is fine. Review resolves this by specificity: the entry with more conditions attached wins. The rule editor shows which entries overlap, but it cannot tell you which one you meant. Someone has to decide, and the decision should be recorded in the entry.

Write down what you want, what you will take and what you will not. Everything else is commentary.