This article should answer:

Playbooks allow you to review legal documents using AI combined with your organisation’s own review rules.

Instead of relying only on general AI knowledge, Playbooks applies explicitly defined internal rules to your document. It analyses clauses and paragraphs individually, cross-checks them against your requirements, and returns structured feedback with references to the relevant sections.

The result: faster reviews that reflect how your legal team actually works.

Large Language Models (LLMs) like GPT are trained on publicly available information. While they understand legal language well, they do not know:

Playbooks solves this by explicitly defining your rules and (optionally) asking deal-specific questions before running the review. This ensures the AI applies your standards — not generic assumptions.

For structured manual reviews without AI, consider using Checklists instead.

You can create playbooks in several ways:

Create a blank category and manually add rules.

Use GenAI to turn existing text (e.g. internal playbooks or client guidelines) into rules. Simply select text in a document and let ClauseBuddy generate a first version.

LawVu Draft provides ready-made playbook samples (Word documents). These are a practical starting point, especially if your team doesn’t already have structured rules.

These templates are intentionally imperfect - they may include too many or even conflicting rules. The goal is to review them internally, remove what’s not relevant, and refine what remains. This process often reveals differences in understanding within teams.

Once you’ve prepared a document, you can convert it into a review category.

For best results, format your Word document as follows:

ClauseBuddy will automatically convert this into structured rules.

If your document wasn’t based on a template, you can still import it if it uses proper outline levels (e.g. Heading styles). Higher levels become groups; the lowest level becomes individual requirements.

To preserve ordering, consider adding numbering (e.g. 01, 02, 03).

You can add multiple languages, but this is optional. LLMs can handle multilingual input, though translations can help avoid ambiguity in legal concepts.

LLMs perform better with clear, focused instructions. Overly long or complex rules can reduce accuracy. Think of the LLM like a junior colleague—clear key points often work better than extensive detail.

You can store additional internal guidance in the Context field. This is shown to users but not sent to the LLM.

Include information the LLM may not know (e.g. recent developments, internal policies, niche legal areas). You can omit widely known concepts or standard legal knowledge unless precision is required.

Within a category, you can add different types of elements:

A requirement defines what should be checked in a document.

You can enhance requirements with:

Conditions control when a requirement is relevant.

They can include questions for the end-user (e.g. deal size, product type). The answers are passed to the LLM and help determine whether the rule applies.

Conditions can be reused across multiple requirements, which avoids duplication and keeps rules consistent.

Actions guide the user on what to do, regardless of whether a requirement is met. For example:

You can provide multiple actions per requirement (e.g. strict vs. flexible fallback options).

Insights give general background information to the LLM (e.g. explaining your product or internal practices). They help avoid repeating the same explanation across multiple requirements.

This feature checks for exact text, using traditional (non-LLM) matching.

Use it when wording must be precise (e.g. mandatory legal clauses). Keep in mind:

For multiple clauses, create separate rules rather than one long match.

You can define different levels of compliance, such as:

This allows more nuanced feedback during review, especially when multiple acceptable positions exist.

Once your category contains multiple rules, you can group them into review sets.

Review sets allow you to:

This is useful when not all rules apply to every deal. For example, a low-value contract may require fewer checks than a high-risk, high-value agreement.

If no review set is used, all rules in the category are applied by default.

AI-based document review is still evolving.

Playbooks should be considered a structured first-pass review that helps you:

You should not blindly rely on the output, regardless of how detailed your rules are. Final legal assessment remains the responsibility of a qualified professional.

The quality of results will continue to improve as AI models evolve.