Blog
How to document your business processes without losing your mind

SOPs, workflows, onboarding guides: everyone knows they should exist. Nobody has time to write them. The answer is documentation that writes itself.
Every growing business reaches a point where the founder or manager thinks: "We need to document our processes." The invoicing workflow. The client onboarding steps. The quality assurance checklist. The hiring process. The escalation procedure. All the things that experienced team members carry in their heads and new team members have to learn through osmosis.
The thinking is correct. Documented processes reduce errors, speed up onboarding, make the business less dependent on specific people, and provide the foundation for improvement (you can't improve what you can't see). The gap between companies with documented processes and those without grows wider with every hire, because each new person either reads the documentation or consumes someone else's time learning verbally.
The problem is that writing process documentation is miserable work, and the documentation goes stale almost immediately.
Why process documentation fails
The writing is painful. The person who knows the process (usually the most experienced and busiest team member) has to stop doing productive work, switch to writing mode, and produce a clear, step-by-step document that makes sense to someone who has never done the task. This context switch is cognitively expensive, the task feels low-priority compared to actual work, and it takes far longer than anyone estimates.
The documentation is immediately incomplete. Processes have exceptions, edge cases, and judgment calls that the writer doesn't think to document because they handle them automatically. The written SOP covers the happy path. The first time a new hire encounters an exception, the SOP doesn't help and they come to ask the experienced person anyway.
The documentation goes stale instantly. Processes evolve continuously through small adjustments that nobody thinks to reflect in the documentation. The SOP was accurate when written. Three months later, two steps have changed, a tool has been swapped, and an approval step has been added. The document now describes a process that doesn't exist, and following it creates more problems than it solves.
Nobody maintains it. Process documentation has no natural owner. The person who wrote it has moved on to other priorities. The people who follow it don't feel responsible for updating it. The documentation drifts from reality until someone notices, which is usually when a new hire follows the outdated instructions and produces the wrong result.
The self-writing alternative
The alternative to writing process documentation manually is letting it write itself from the activity that defines the process.
Your team's processes are already expressed in their daily work: the Slack discussions about how to handle a client request, the meeting where the team walked through the onboarding steps, the email chain that documents the approval workflow, the checklist in the project management tool that captures the QA steps.
Self-writing documentation captures these expressions of the process and structures them into documentation. The Slack thread where the team discussed "here's how we handle refund requests" becomes a documented refund procedure. The meeting where the operations lead walked a new hire through the client onboarding steps becomes an onboarding guide. The recurring discussion about how to handle a specific edge case becomes a documented exception procedure.
The documentation stays current because it's generated from the ongoing activity. When the process changes (a new step is added, a tool is swapped, an approval is removed), the change is discussed in Slack or a meeting, and the documentation updates to reflect it. The staleness problem that kills manually written SOPs doesn't apply because the documentation is derived from reality rather than maintained alongside it.
What this looks like in practice
Client onboarding. Instead of writing an onboarding SOP, connect the sources where onboarding activity happens: the Slack channel where the team discusses new client setups, the meeting recordings where onboarding calls are held, the email correspondence with new clients. The self-writing system captures the process as it's performed and produces documentation that reflects how onboarding actually works, including the informal steps and exceptions that a manually written SOP would miss.
Sales process. The stages of a deal, the approval steps for discounts, the handoff process between sales and delivery: all captured from CRM activity, Slack discussions, and meeting recordings. The sales playbook writes itself from real deals rather than being authored by someone trying to describe the idealised version.
Engineering workflows. The deployment procedure, the code review standards, the incident response protocol: all captured from GitHub activity, engineering Slack channels, and post-mortem meetings. The engineering wiki stays current as practices evolve.
Operations and admin. The invoicing workflow, the expense approval process, the vendor management steps: captured from the discussions and communications where these processes are executed and refined.
Getting started
You don't need to document every process at once. Start with the process that causes the most pain when it's not documented, typically whatever your team asks about most frequently or whatever new hires take longest to learn.
Connect the sources where that process is discussed and executed. Let the self-writing system generate the initial documentation. Review and refine it (this is a fifteen-minute task rather than a two-day writing project). Then expand to the next process.
Over months, the documentation library grows alongside the team's activity. Each process is documented, current, and searchable. New hires can find the answer to "how does X work?" without asking, which means the experienced team members who currently answer that question multiple times per month get their time back.
Frequently asked questions
Won't the auto-generated documentation be too messy to use? The self-writing system produces structured, cited documentation rather than raw transcripts. The output is a summary of the process with steps, exceptions, and links to the source discussions. You review and refine rather than writing from scratch, which is a much smaller task.
What about processes that involve sensitive information? Access controls ensure that process documentation is visible only to the people who should see it. You can configure which sources are captured and which are excluded, and the privacy architecture maintains the same access boundaries as the underlying tools.
How do we handle processes that span multiple teams? Cross-team processes are captured from the channels where they're discussed across teams. The self-writing system doesn't respect team boundaries in its capture (unless configured to), which means cross-functional processes are documented more completely than they would be in a manual system where each team documents only their portion.
What if the process is informal and undocumented? That's the most common case, and it's where self-writing documentation has the most value. If the process is performed and discussed but never formally documented, the self-writing system captures it from the discussions and produces the first documentation. An informal process that's captured is more useful than a formal process that was written once and went stale.
Does this work for regulated industries? Documented processes are often a compliance requirement in regulated industries. Self-writing documentation provides auditable, cited process documentation that can be reviewed and approved through the organisation's compliance workflow. The documentation is more reliable than manually maintained SOPs because it reflects actual practice rather than intended practice.
Can we use this for ISO or quality management documentation? The self-writing documentation provides a strong foundation for quality management systems. The cited, structured output can be reviewed and approved through your existing quality processes. For formal certification, you'd review and approve the generated documentation as you would any manually written SOP, but the initial creation and ongoing maintenance are handled by the system.
What if our processes aren't consistent yet? The self-writing system captures processes as they're actually performed, which makes inconsistencies visible. If three people handle refunds differently, the documentation reflects all three approaches, which is the starting point for standardising. You can't standardise what you can't see, and the documentation makes the current state visible.
How do we handle process changes? Process changes are typically discussed in Slack or meetings before being implemented. The self-writing system captures these discussions and updates the documentation to reflect the new process. There's no separate "update the SOP" step because the documentation derives from the activity rather than being maintained alongside it.
Can we export the process documentation? Yes. The documentation is accessible through the API and exportable in standard formats. If you need to share process documentation with external auditors, clients, or partners, the content can be exported or published as a shareable page.
What about processes that involve external tools or physical steps? The documentation captures the discussion and decision-making around the process from your communication tools. Physical steps or steps in external tools that aren't connected won't be captured automatically but can be documented manually and integrated with the auto-generated content. The hybrid approach (auto-generated for what can be captured, manual for what can't) still saves significant time compared to fully manual documentation.
Related reading: Nobody reads the wiki, Your team keeps asking the same questions, How to build a company brain. Related pages: Self-writing docs, Docs that write themselves, Onboarding.
Other blog posts:

Notion tries to be everything and fails at most of it

Notion AI is an expensive add-on that should be built in

Notion is too complicated for normal people

What is AI knowledge management?

How to document your business processes without losing your mind

How to stop losing information at work

Your team keeps asking you the same questions. Here's how to stop it.

How to actually use AI in your business (not just ChatGPT)