Contents 🚀
Introduction 👋
A playbook in Juro is a set of rules that AI can follow when redlining incoming agreements.
Playbooks also guide humans through a process. These 'legacy playbooks' are a good starting place for setting up your AI contract review playbook in Juro.
|
Structuring your playbook 🏗️
Your playbook should consist of an identification section and rules for reviewing uploaded contracts.
The identification section 🕵️♂️
The identification section contains the context the AI will need to match relevant documents to this particular playbook. Each rule should relate to a component of the document you'd like the AI to check and potentially correct during the review process.
As much as possible, your playbook should read clearly from top to bottom: avoid reliance on formatting to convey meaning as much as possible (italicized text, etc.). Instead, try to use:
|
|
💡 TIP: If your playbook meets accessibility goals in line with the above guidance, the Agent is more like to adhere to its guidance. Microsoft Word has a 'Check Accessibility' function that works well for this.
|
Setting up your rules ®️
Your rules should contain the following elements:
1. Identification
|
Provide clear information that the AI can use to identify where compliance with a contract or a particular rule should be checked and, if necessary, corrected.
|
2. Logic
|
Provide unambiguous, non-contradictory logic for how the AI should apply the rule in question.
For example, should the language match your preferred position exactly, or are there objective steps the AI can follow to check playbook compliance?
|
3. Context
|
Provide clear context so the AI can perform the task you've outlined: even a seemingly obvious context must be given.
|
4. Examples
|
Whether you're looking for exact wording, specific components, or general principles, you should provide examples of what a good result would look like in different situations.
|
|
Example rules 🌉
Below is an example of a rule, 'NDA must be mutual'.
Start of Rule: "NDAs must be mutual"
NDAs must be mutual - Rule Identification:
To apply this rule, the AI should identify any clause(s) or provisions within the contract that pertain to: Definition of Confidential Information, Obligations of confidentiality, Exceptions to confidentiality, Term or duration of confidentiality, Disclosing and receiving parties.
The rule applies specifically to non-disclosure agreements (NDAs) and should trigger review when the NDA outlines confidentiality obligations for only one party (i.e., a one-way NDA). Look for asymmetrical phrasing such as: “Recipient shall protect…” “The Receiving Party agrees…” “Disclosing Party’s confidential information…”
Also identify structural asymmetry, e.g., the agreement defines only one “Disclosing Party” and one “Receiving Party.”
NDAs must be mutual - Rule Logic:
The AI must confirm that the NDA contains mutual confidentiality obligations. This means that: 1) Both parties are defined as disclosing and receiving parties. 2) The obligations, exceptions, and duration apply equally to both parties. 3) There is no clause that provides protection for only one party’s confidential information.
If the NDA is mutual, the clause(s) will: Use plural or symmetrical definitions like “Each Party” or “Both Parties”, Include language such as “each party agrees…” or “both parties shall…”, Define Confidential Information for both [YOUR COMPANY] and the counterparty
If the NDA is one-way, the clause(s) will:
Define obligations solely for the Receiving Party, Define Confidential Information from the Disclosing Party only, Omit protections for [YOUR COMPANY] confidential data
In such a case, the clause must be flagged and rejected.
NDAs must be mutual - rule context:
Perspective: The AI is reviewing the NDA on behalf of [YOUR COMPANy], who is a party to the agreement and typically the vendor in the contractual context.
Business rationale: [YOUR COMPANY] confidential information—including [X, Y, Z]—must be protected by the NDA. Signing a one-way NDA exposes [YOUR COMPANY] to risk if the prospect shares its information without restriction.
The AI should consider that prospects typically share information about their business, processes, or legal/commercial structures, and that mutual NDAs provide a fair and balanced framework for this exchange.
NDAs must be mutual - rule examples:
Examples of what is acceptable (i.e what is Mutual):
“Each party agrees that it shall not disclose any Confidential Information received from the other party…”
“The obligations of confidentiality in this Agreement apply equally to both parties.”
“Both [YOUR COMPANY] and [Prospect Name] may disclose confidential information under this Agreement and each agrees to treat such information with the same degree of care…”
Examples of what is unacceptable (i.e. One-way):
“The Receiving Party shall protect the Disclosing Party’s Confidential Information and use it only for the Purpose.”
“Confidential Information means any information disclosed by the Disclosing Party to the Receiving Party…”
“The Recipient agrees not to disclose or use the Discloser’s Confidential Information…”
In all “Unacceptable” cases above, the obligation is not mutual, and the NDA fails to protect [YOUR COMPANY]’s confidential information.
End of Rule: "NDAs must be mutual"
Uploading your playbook ⬆️
You should write your playbook in a text editor before exporting it as a DOCX file and uploading it to your Juro playbook library. To upload a playbook (Org Admins only):
1. Open the Legal Agent from the left sidebar (AI contract review).
|
|
|
|
Alternatively, you can drag and drop documents from your desktop into the upload portal.
|
|
4. Enter your playbook's name in the PLAYBOOK TYPE field and click Upload.
|
|
After a few minutes, your playbook will load into Juro, where it'll be available for you to review your documents against. You can update the playbook any time by clicking Upload new version.
If you don’t have a playbook, Juro provides an example NDA playbook accessible via the blue hyperlink in this section. This can be used as a base to build out/adjust your guidance according to your company's guardrails and positions.
Testing your playbook 🎺
Testing your playbook before rolling it out to your main Juro environment is important to identify where improvements can be made to better ensure accuracy and adherence, by the AI Agent, to your guidance.
Below are some tests we use that you can also run to highlight areas where your playbook could be refined to give you confidence in the AI-powered contract review before going live.
Playbook selection tests 🔶
For multiple playbooks in Juro, you should run a few test documents through the Agent or the AI Review in Word to ensure that the correct one is being selected, e.g. if you have an NDA playbook and a Supplier playbook.
It can also be a good idea to upload a few documents that don't match either playbook to check that unsupported document types are being rejected.
Rule application tests 🔷
To test that your rules are applying correctly, upload some test documents you know, for a particular rule, are not playbook compliant.
Review the outputs carefully and check for the following:
If the rule was triggered.
That the amendment makes the document compliant with the playbook.
That the justification given is accurate and helpful to potential end users.
You should also perform the same test for documents you know are fully compliant with the playbook, in order to test for false positives.
Improving your playbooks ⏩
Iterating your playbook as your tests progress, and also after you start using Juro's contract review tools on your live documents will only enhance the responses generated by the AI Assistant, and how it performs more generally.
To learn more about getting the most from our AI functionality more widely, you can read some of our dedicated guidance.
💁♀️ As always, our Support Team is happy to help you with anything further if needed. Start a chat with us right here by clicking the Intercom button in the bottom-right-hand corner of this page.
Alternatively, you can email your query to support@juro.com 🚀
|