Short answer: An SOP tells one person how to do one task correctly. A workflow defines how work moves between people and systems. A playbook gives someone the context and judgement to handle a whole domain, including situations nobody wrote down. Small businesses usually need workflows first, SOPs second, and playbooks for the roles they intend to delegate.
An SOP is task-level and prescriptive. It covers a bounded, repeatable activity — raising an invoice, running the weekly backup, processing a refund — with numbered steps and a clear definition of done.
SOPs work well when the task is stable, the correct output is unambiguous, and the cost of variation is high. They work badly when the task requires judgement, because a prescriptive document forced onto a judgement-heavy activity produces either rigid nonsense or a document everyone ignores.
A workflow is system-level. It describes how a unit of work moves — the states it passes through, what triggers each transition, who owns each state, and what has to be true to move forward. It spans people and tools.
Workflows are what you build automation on top of. An SOP describes what a person does at one station; the workflow describes the production line. Businesses that document SOPs without defining workflows end up with a set of well-described tasks that still don't join up.
A playbook is role-level and contextual. It equips someone to own an area — client success, delivery, finance ops — by giving them the principles, the standard plays, the decision framework, and the escalation rules. It includes the SOPs and workflows but doesn't stop there.
The distinguishing feature is that a playbook prepares someone for situations it doesn't explicitly cover. That's the point. You can't write an SOP for every eventuality, so at some point you have to transmit judgement instead of instructions.
For a founder-led business under about thirty people, the sequence that works is:
Diagnose by symptom:
Automation implements a workflow. It cannot implement an SOP directly, because SOPs are written for humans, and it cannot implement a playbook at all, because playbooks encode judgement. This is a useful boundary: the workflow layer is the automatable layer. Steps that only appear in a playbook are, by definition, steps you've decided need a person.
Badly, yes. In practice the audiences differ — a task-doer, a system designer, and a role owner — and combining them produces something none of them reads.
Typically ten to twenty, covering the tasks that are frequent, consequential, or single-person dependencies. More than forty and maintenance becomes the problem.
No. A strong playbook for a single role is often ten to fifteen pages: principles, the standard plays, decision rules, escalation paths, and links to the relevant SOPs.
We hand over the full playbook at the end of every build — knowledge, not dependency. See how that works in practice.