The AI Tool Isn’t the Bottleneck. The Playbook Is.

I’ve been talking to hundreds of General Counsels about how they’re using AI. The pattern is pretty clear: almost everyone is experimenting with the tools. Some are getting real leverage. Others are getting generic mush. The difference is usually not the model or the vendor (they’re mostly using Claude’s Opus model at this point), but whether they have a playbook.

That lesson is not limited to legal. It applies to almost every function. Sales teams need playbooks for qualification, objection handling, follow-ups, pricing approvals, and handoffs. Customer success teams need playbooks for onboarding, renewals, escalations, churn risk, and executive reviews. Finance teams need playbooks for reporting, budget reviews, variance, investor updates, and approval workflows. Founders need playbooks for hiring, fundraising, customer calls, product decisions, and internal communication.

AI is only as useful as the operating context you give it. If you ask it to write a follow-up email, review a contract, summarize a customer call, or analyze a budget, it will usually give you something plausible and generic. Sometimes that is good enough. But if you want it to sound like your company, reflect your standards, and make decisions the way your best people would, you need to give it the underlying logic.

That is what a playbook does. It captures what exceptional looks like. In legal, that might mean fallback positions and escalation rules. In sales, it might mean your real qualification criteria. In customer success, it might mean the difference between a saveable account and a bad-fit customer. In finance, it might mean how you explain misses without sounding defensive.

The point is simple: generic input produces generic output. Better prompts help, but prompts are not enough. The teams getting the most from AI are not just “good at AI.” They have done the work of making their judgment legible.

The AI tool is not the bottleneck. The playbook is.