Overview
Search Criteria are the engine of PlaybookOS (PbOS).
They tell the AI what to look for inside a contract and determine when and how a topic’s logic should activate. While Topics define the issues that matter, Search Criteria define the signals that matter — the words, phrases, or contextual patterns that indicate a clause is relevant.
In short: no matter how well your topic is written, it only works if the AI knows what to find.
Criteria
TERMS & CONNECTORS (rare) - Westlaw syntax (e.g., "liability /5 limit!" OR "consequential /3 damage*")
EDITED (rare) - Used to detect incoming redlines against a standard template Operations:MUST (most frequently used) - Always included in search results
MAY - Optional search terms that broaden results (e.g., obligation for CI to be marked or identified)
MUST NOT - Excludes or filters out unwanted results
What Are Search Criteria?
Each Topic in a PbOS playbook includes a set of one or more Search Criteria. These criteria teach the AI how to identify the corresponding clauses during analysis.
A single criterion includes:
- Operation → The logic type (MUST, MAY, or MUST NOT).
- Type → How the AI should interpret it (PROMPT, EDITED, or TAGS).
- String Value → The actual instruction or search phrase.
- Scope → Where to search (always set to EDITED for contract review).
While the schema allows multiple types, PbOS authors should rely on PROMPT as the primary method, since it uses natural language to tell the AI what to find — keeping playbooks easy to read and maintain.
Why Search Criteria Matter
🧭 They Direct the AI’s Focus
The Playbook can’t act without knowing what to look for.
Search Criteria give the AI its compass — the rules that define when a topic’s Outcomes should trigger.
Example:
PROMPT: “Find any clause that limits a party’s liability.”
With that simple instruction, PbOS scans the document, identifies potential matches, and flags them under the “Limitation of Liability” topic. From there, the defined Outcomes take over to guide what happens next.
Without clear criteria, the AI might:
- Miss relevant clauses (false negatives), or
- Surface unrelated content (false positives).
The precision of your criteria directly affects both accuracy and reviewer confidence.
🧠 They Translate Human Expertise Into Machine Logic
Search Criteria act as the bridge between human reasoning and AI execution.
Instead of coding complex Boolean queries, PbOS lets authors use plain language prompts that mirror how a lawyer or contract manager thinks.
For example, instead of writing:
indemn! & hasredline()
You can now write:
“Identify any indemnification clauses that have been modified.”
This natural-language model allows non-technical users to encode expertise directly into PbOS, turning policy knowledge into reusable automation.
🎯 They Control Precision and Breadth
Each criterion includes an operation that defines how the AI combines multiple conditions:
| Operation | Meaning | Example |
| MUST | Clause must meet this rule | “Must contain limitation of liability language” |
| MAY | Clause may meet one of these rules (broadens search) | “May include terms like ‘cap’, ‘maximum’, or ‘limit’” |
| MUST NOT | Clause must exclude this | “Must not include references to consequential damages” |
{insert table}
Using these operations strategically helps balance coverage vs. precision:
Use MUST to anchor your primary search. Add MAY to catch variations in wording. Add MUST NOT to eliminate noise or false positives.
⚡ They Drive Automation at Scale
Search Criteria are what allow PbOS to scale.
Each criterion can trigger logic across hundreds of contracts automatically — consistently identifying, categorizing, and routing clauses for review.
Because PbOS criteria are flexible, teams can:
- Run consistent clause detection across all agreements.
- Avoid missing risky provisions hidden in nonstandard language.
- Power both human review and AI-driven markup generation.
The result: faster, smarter, and more reliable first-pass contract analysis.
Best Practices
Start with PROMPT
Use clear natural language. Example:
“Locate clauses describing how and when payments must be made.”
Keep it single-purpose
Each criterion should focus on one intent. Don’t overload with multiple legal concepts.
Limit MUST criteria
The schema supports one primary MUST rule — use it as your anchor.
Add MAY only for variations
Broaden gently, not wildly. Example:
“MAY include words like ‘fee’, ‘invoice’, or ‘compensation’.”
Use MUST NOT to cut noise
Exclude confusing overlaps. Example:
“MUST NOT include references to marketing approvals.”
Test and refine
Run a few sample contracts to validate that PbOS highlights what you expect. Adjust the prompt wording to improve relevance.
Common Pitfalls
| Pitfall | Why It’s a Problem | How to Fix It |
| Using complex Boolean logic | PbOS already interprets natural language—manual syntax adds confusion. | Stick to conversational prompts. |
| Stacking multiple MUSTs | Creates overly strict conditions that may miss valid clauses. | Combine related terms into one PROMPT. |
| Skipping negative logic | Leads to irrelevant matches or false positives. | Add “MUST NOT” to exclude unrelated contexts. |
| Being too short or vague | The AI can’t infer intent from a single keyword. | Use descriptive phrases or examples in your prompt. |
Example: Search Criteria in Action
Let’s say you’re building a topic for Limitation of Liability.
You might tell the AI to “find any clause that limits a party’s liability or caps damages” — then add “exclude clauses that only mention confidentiality.”
This balance of what to find and what to ignore helps DocJuris surface the right clauses while filtering out false positives, making your playbook faster and more accurate.
Key Takeaway
Search Criteria are where the intelligence of your playbook begins. They translate your organization’s expertise into machine-readable instructions that power every downstream decision — from detection and task generation to AI-assisted editing.
Write them as if you’re explaining the clause to a colleague — clear, specific, and focused on intent. When your criteria are strong, the rest of your playbook becomes effortless.