Stop Waiting for the Perfect Playbook

A lot of legal teams believe they need playbooks before AI can help them review contracts. It's a reasonable belief. It's also holding a lot of them back.

To be clear, we love playbooks. Every team that reviews the same agreement every week should have one. A well-built playbook is the most precise way to tell AI what you will and won't accept.

The problem is treating the playbook as a prerequisite.

Most Agreements Will Never Get a Playbook

Look at what your team reviewed last quarter. A few agreement types made up most of the volume. Those deserve playbooks.

Then there's the long tail. The consultancy agreement. The reseller deal. The vendor's own paper that looks nothing like your template. Each one shows up a few times a year.

Nobody is going to write a 40-rule playbook for an agreement they see three times a year. Nor should they. So the rules for those agreements stay where they've always been: in the heads of the lawyers who review them.

If AI only works once those rules are codified, it will never touch most of the agreements a legal team reviews.

Most of a Review Isn't Agreement-Specific

Playbook-first thinking misses something simple. Most of what a lawyer checks in an unfamiliar agreement has nothing to do with the agreement type.

Uncapped liability. Governing law. Auto-renewal. What survives termination. Whether your company's name and address are right. These apply to almost everything you review.

An experienced reviewer also notices the unexpected: a clause that doesn't belong, or a section that should be there and isn't. AI can do a version of that by comparing the agreement against the market and flagging what's missing or unusual.

A short list of universal rules plus a market check won't beat a tailored playbook on your highest-volume agreement. It doesn't have to. It takes the long tail from nothing to a consistent first pass.

The Lawyer Still Makes the Call

Without a detailed playbook, won't the AI make calls it shouldn't? When we rebuilt Risk Review, we left out an “apply all” button on purpose. Every issue comes back with a risk rating, an explanation and a proposed fix. The lawyer decides each one.

A playbook encodes judgment in advance. A lighter setup asks for it at the point of review. Both are legitimate. For an agreement you see three times a year, I'd rather spend my judgment on the agreement than on a document about the agreement.

When You Need a Playbook, Make It Cheap

Teams put playbooks off because of the blank page. So make the first version cheap: start from a standard playbook built on market data, or write down the rules you already follow and let the AI turn your notes into rules with fallbacks.

Bottom Line

Write down the handful of rules you apply to every agreement and start using AI on the long tail now. Then pick the one or two agreement types that make up most of your volume and build playbooks for those.

Waiting for a complete set of playbooks means waiting for something most teams will never finish. Don't wait.