A playbook is your best way of doing a piece of work, written down so anyone can repeat it: the steps, the information needed at each one, the tools used, and the checks that keep quality up. Playbooking is the work of capturing those plays, and in an AI-enabled company, writing them so that both people and AI agents can run them.
Most companies already have good ways of working. They just live in a few people's heads. When those people are busy or leave, quality drops. When AI enters the picture, the problem gets bigger: an AI tool can only follow a process that has been made explicit.
Who it is for
Playbooking fits companies that:
- Depend on one or two experienced people for work others cannot do the same way.
- Use AI tools inconsistently, with everyone writing their own prompts and getting different results.
- Are onboarding new staff and want them productive faster.
- Plan to build AI agents and need the process defined first.
- Want a standard for when AI output must be checked by a person.
What you get
- Mapped workflows. Each chosen workflow drawn step by step with the people who do it, including the exceptions and the decisions that need judgment.
- Written plays. Clear instructions for each step: what goes in, what comes out, who or what does it, and how to check it.
- Prompt and tool standards. Tested prompts, the tools approved for each task, and what data may and may not be shared with them.
- Review rules. Where a person must approve AI output before it reaches a customer or a system of record.
- A living format your team owns. Playbooks kept where your team already works, with an owner for each and a simple way to update them.
How it works
1. Listen
We sit with the people who do the work and watch how it is really done, which is often different from how it is documented. We collect the examples, templates and messages that show what good looks like.
2. Shape
We write the play, test it with someone who did not write it, and refine it until it produces the same quality result. Where AI can do a step, we write the prompt and the check for it and test them against real cases.
3. Build
We put the playbooks where your team will use them, set up any shared prompts or assistants, and train the people who will run them.
4. Sustain
We agree who owns each playbook and how changes are made, and we review them after a few weeks of real use. A playbook that nobody updates goes stale quickly.
How it connects to agents
A good playbook is the specification for an AI agent. Once a workflow is written down with clear inputs, outputs, tools and approval points, building a custom agent to run it becomes far faster and safer. Many clients start with playbooks for the workflows they plan to automate, and use the same plays to train staff on the work that stays with people.
How long it takes
A single workflow can be mapped and written in a matter of days. A set of playbooks for a team or a department takes a few weeks. We agree the list of workflows and a schedule at the start.
How pricing works
We start with a free conversation about the work you want to capture. We then quote a defined scope, usually a named set of workflows, so the price is known before work begins.
Common questions
How is a playbook different from a standard operating procedure?
It is similar, but written to be used, not filed. A playbook is short, shows examples of good output, and says exactly where AI may help and where a person decides.
Do playbooks work if we do not use AI yet?
Yes. Clear, shared ways of working help any team. They also make adopting AI later much easier.
Who keeps them up to date?
Your team. Each playbook has a named owner, and we set up a simple routine for reviewing and changing them.
Related reading: how small businesses can use custom AI agents and what AI consulting usually includes. Playbooks often come out of an AI transformation roadmap.