Contract playbooks are the unsung heroes of legal teams. These carefully crafted documents can streamline negotiations and help to navigate legal risk with business goals. They also empower legal teams to make more informed decisions and handle complex legal issues in a consistent manner. 

What is an “AI-Ready” playbook?

AI-ready playbooks take traditional playbooks to the next level. Unlike static documents that require lawyers to interpret and apply their principles manually, AI-ready playbooks integrate your team’s expertise into a workflow that allows AI to analyze contracts, apply tailored risk profiles and generate bespoke redlines. Whether you are starting from scratch or updating existing playbooks in Word, Google Docs or a legal AI tool, the tips below will help you supercharge your legal team.

Why should I bother?

Handing traditional playbooks to AI is like giving a tourist with a phrasebook a legal contract to review; they might recognize a few words, but the nuance will be lost on them. While AI can produce outputs that look human, it doesn’t think like one. And no matter how shiny or advanced your AI tool is, off-the-shelf models rely on generic positions that will not necessarily match your business’ unique risk tolerance, goals or preferred positions.

This is where AI-ready playbooks come in. By tailoring your playbook for AI, you’re essentially giving it a custom cheat sheet, complete with context, actionable steps and specific instructions, so it can work smarter, not harder. Yes, there’s some initial effort involved, but the payback is likely to be transformative. Tools like Wordsmith AI can do a first pass of contract reviews, applying your tailored guidance to mark up documents. The result? Less time spent on repetitive tasks and more time for your team to focus on strategic, high-value work. Playing the long game with AI-ready playbooks isn’t just smart; it’s a game-changer.

Top tips on how to build an "AI-ready" playbook 

1. Format

The good news is that you do not need to scrap your existing playbook and start from scratch. In fact, it is likely that your existing format broken down in the table format (issue, clause, preferred position, rationale, compromise position, escalation etc.) will be helpful as tables, standardized headings and subheadings make it easy for AI tools to parse and retrieve specific guidance. However, it is likely that you will need to include additional information as a result of the tips set out below.

2. Modular design; breaking down complexity

When lawyers use a playbook, they naturally apply their legal expertise before deciding whether a contract needs amending. For example, something simple like “If it’s a unilateral NDA...” requires a human to first determine whether the NDA is unilateral or mutual, a step that feels intuitive to us but is anything but for AI.

AI often stumbles with “single-shot” prompts, where it’s expected to make logical leaps or assumptions without clear guidance (check out this blog for a deeper dive into why this happens). That’s why an AI-ready playbook should break down issues into small, atomic steps. Each decision point must be atomic, crystal clear and actionable, leaving no room for ambiguity or missteps.

By way of example, a traditional playbook may include a few different concepts:

"Given the scale at which we are growing and the fact that HR recruit independently of the customer-facing parts of the business, our starting position: push back on the inclusion of any non-solicit provisions in the first instance. If the counterparty can justify that they need it to give them comfort to deal with us openly, it should be limited to 6 months (maximum 12 months), be limited to identifiable senior employees (above Director level); and include carve outs for bona fide general offers or responses to ads. Push back on anything that is not possible to manage / control or any hair trigger."

If you handed this extract directly to AI for contract review, it would likely stumble. To help AI succeed, you need to break the task into the smaller questions you instinctively check off before reaching an answer. An AI-ready playbook should guide this process by breaking down the considerations into clear, step-by-step questions, such as:

"Modular questions to be answered:

1. Does the agreement contain any non-solicit provisions?

2. How long are restrictions?

3. Are the restrictions only applicable to senior people of Director level and above?

4. Are there any exceptions or general carve outs?

5. Have we already pushed back on the inclusion of a non-solicit in its entirety?

Preferred position: no non-solicit provisions.

Context: given the scale at which we are growing and the fact that HR recruit independently of the customer-facing parts of the business, our starting position

Compromise position: if we receive pushback, any non-solicit should be limited to 6 months in length, apply only to Director or above level employees and should include a carve out for any bona fide general offer or a response to an advert."

3. Be specific

Providing specific instructions is really important to avoid AI making assumptions or jumping to conclusions. In short, treat AI like a five year-old so if you want something to be expressly stated in an agreement, you need to say that! Similarly, to avoid assumptions being made, spell out how certain scenarios should be addressed. For example, if you want to make sure late payment interest is not more than 2%, there are certain scenarios you may need to pre-empt:

"Late payment interest rate should not be more than 2%. Bear in mind that: 

(i) if a contract refers to the late payment of interest being the lower of: (i) [a fixed percentage]; and (ii) the maximum rate permitted by law, consider the fixed percentage and flag that (ii) may be higher than 2%; and

(ii) a percentage over and above a bank’s base rate could be more than 2% in aggregate even if the percentage specified is less than 2%."

4. Context

Provide context as you would to a less experienced team member for why specific clauses or fallback positions are preferred: 

"There should not be a mutual cap on liability. As the customer, our main obligation is simply to pay the fees, but the supplier has a range of obligations and therefore its liability should not be capped at the same level."

In the example above, it might be obvious to a lawyer that a liability cap should not be mutual when one party's role is limited to making payments while the other party has extensive obligations. However, this understanding relies on contextual or tacit knowledge about how the service is being delivered; information that is not included in traditional playbooks. While this distinction might seem trivial, it must be explicitly articulated for AI to interpret and apply it correctly.

5. Tell AI what not to do or how to think about something

Consider whether any of your playbook instructions, if taken literally, could result in inadvertent changes being made (that you would not need to call out ordinarily). If they could, be explicit about changes it should not make:

"If the playbook states that the liability should be at least two times the annual fees, you should make it clear that this does not apply to liability for indemnities you would expect to be uncapped (e.g. data protection and/or confidentiality). 

If you want to ensure no power of attorney is granted in an NDA, you should state that the agreement can still be signed under a power of attorney."

6. Highlight conditionality or dependencies

Contracts are interconnected and clauses should not be looked at in isolation. If you tweak one clause, it might impact others so you should specify these: 

"If an NDA is mutual, we can accept less stringent confidentiality obligations."

"If all liability is capped at 1x fees, the scope of the warranties is less critical."

Explicitly outline any dependencies in your playbook to help both AI consider the broader implications of individual changes.

7. Include examples

Provide examples of acceptable and unacceptable language for AI to reference.

Unacceptable: “For a period of 3 years after the date of this agreement, the Supplier shall not directly or indirectly solicit, engage, employ (whether paid or unpaid) or offer to employ, any employee, director or consultant of the Customer.”

Acceptable: “For a period of 6 months after the date of this agreement, the Supplier shall not, without the Customer’s prior written consent, directly or indirectly solicit, engage, employ (whether paid or unpaid) or offer to employ, any Restricted Person of the Customer unless it is an offer made to any person who contacts the Supplier solely on their own initiative or in response to a bona fide general employment advertisement.”

8. Test and iterate

By now, you might be asking, “How much do I really need to spell out for AI?” Here’s the hard truth: you could follow all the advice above perfectly, and AI might still get it wrong. There’s no magic bullet. Sometimes, AI’s insights will amaze you; other times, you’ll wonder if that five-year-old might have done a better job. 

The only way to strike the right balance and ensure consistent excellence is to test and refine. This step is absolutely critical. Once you have built or updated your playbooks, put them to the test with your legal AI tool using a variety of agreements. Analyse the outputs carefully - where the AI stumbles, you know you need to refine further. Break down concepts, add clarity and provide more context or assumptions where needed. Although it is the final step, testing is one of the most important ones.

Final thoughts

Creating AI-ready playbooks isn’t just about adopting the latest tech; it’s about setting your legal team up for future success. In the era of AI, a static, traditional playbook simply won’t cut it. To truly harness the power of AI, your playbooks must evolve to provide structured, tailored guidance that amplifies what AI tools can deliver.

At Wordsmith AI, we’re not just building tools; we are transforming how legal teams work. Our platform and the Wordsmith Academy are designed to help in-house teams confidently navigate the transition to AI while transforming their playbooks into bespoke, AI-ready workflows.

Feeling overwhelmed? Don’t worry, you’re not in this alone! Yes, there’s a lot to tackle, but that’s where we come in. At Wordsmith AI, we specialise in doing the heavy lifting for you - refining your existing approach or crafting bespoke solutions that align with your needs. After all, you’ve got better things to do than babysit your AI! 

Ready to get started? 

• Sign up to Wordsmith for free to explore how we can help future-proof your legal processes and save your team time.

• Sign up to Wordsmith Academy to skill up and learn more about the ways in which AI can help you with day-to-day legal work.