Turn Tribal Knowledge Into Written Processes With AI
There's a person in most small businesses — maybe it's you — who just knows how things work.
They know which supplier needs a heads-up before placing a big order. They know the three fields in the CRM that sales actually cares about. They know the exact sequence that keeps a client onboarding from going sideways. None of it is written down anywhere. It lives in their head, passed on through shoulder-taps and answered questions and "just watch me do it once."
That's tribal knowledge. And it's one of the most common growth bottlenecks there is.
When that person is unavailable — or when you need to hire someone new, or delegate a task you've always done yourself — the whole thing has to be reconstructed from scratch. Every time.
AI can't replace the person who knows the process. But it can take a rough, unpolished brain-dump from that person and turn it into a clean, usable document in minutes. It's the first step toward giving AI your business's own knowledge. Here's how.
Why Most Process Documentation Never Gets Done
Ask most business owners why their SOPs (Standard Operating Procedures) don't exist, and you'll hear some version of the same answer: "We know we should, but nobody has time to sit down and write them."
That's honest. Writing documentation is slow, feels low-priority next to actual work, and requires a specific kind of focused effort that busy operators rarely have. The result is that process docs either don't get written at all, or they get written once and immediately go stale.
The AI opportunity here isn't about getting smarter documentation. It's about removing the friction that stops documentation from happening in the first place.
What You Can Feed AI (And What You Get Back)
The core workflow is simple: give an AI tool a rough, unstructured account of how a process works, and ask it to turn that into a structured document.
"Rough" means actually rough. You don't need to write well. You don't need complete sentences. You can ramble. The messier the input, the more impressive the output feels — which is part of what makes this genuinely useful rather than just theoretically clever.
Here are the formats that work well as inputs:
Voice notes / transcripts. This is the fastest method for most people. Open your phone's voice memo app (or a transcription tool like Otter.ai), walk through the process out loud like you're explaining it to a new hire, and stop. Transcribe it — Otter, Whisper, or even Google's built-in transcription will do — then paste the transcript into your AI tool of choice. A five-minute voice note can become a usable SOP draft in under ten minutes total.
Slack or email threads. Already explained a process to someone in writing? That thread is a starting point. Paste it in and ask the AI to extract and structure the steps.
Screen recording transcripts. Record yourself doing the process on screen while narrating. Use a tool like Loom, which auto-generates a transcript, and feed that into the AI.
A rough bullet list. Even "step 1, step 2, step 3" notes — incomplete, out of order, however you jotted them down — work fine as raw material.
A Prompt That Actually Works
The quality of the output depends a lot on how clearly you ask. Here's a prompt structure that works consistently:
"Below is a rough description of how we [name the process]. Please turn this into a clean, step-by-step SOP that a new team member with no prior knowledge could follow. Format it with a brief overview, numbered steps, and any important notes or warnings at the end. Use plain, direct language.
[Paste your raw content here]"
If you want a checklist instead of a narrative SOP:
"Turn the following process description into a concise checklist a team member could use to verify they've completed each step correctly."
If you want a training document for onboarding:
"Using the process description below, write a short training guide for a new hire. Explain not just what to do, but why each step matters."
ChatGPT (GPT-4 or later), Claude, and Google Gemini all handle this kind of task well. You don't need a specialized tool for basic documentation work — any of the mainstream AI assistants will do.
An Example of the Input/Output Gap
Here's what this looks like in practice.
Raw input (spoken, then transcribed):
"So when we get a new client, basically I send them a welcome email — I've got a template somewhere — and then I add them to the project board, set up a folder in Drive, and make sure finance knows to send the contract. Oh, and someone has to schedule the kickoff call, usually that's me but it shouldn't be. We also add them to the weekly client email list. Sometimes people forget that part and then the client misses updates. The contract thing — finance needs about two days' notice or they get annoyed."
AI output (condensed):
New Client Onboarding SOP
Overview: This process ensures every new client is set up consistently from day one — with the right communication, tools, and internal handoffs in place before the kickoff call.
Steps:
- Send the welcome email using the saved template within 24 hours of contract signing.
- Create the client folder in Google Drive using the standard naming convention.
- Add the client to the project board with their name, start date, and assigned team member.
- Notify finance at least 48 hours before the client's start date so they can issue the contract. (Finance needs two days' notice minimum.)
- Add the client to the weekly client update email list. (This step is frequently missed — confirm it's done before moving forward.)
- Schedule the kickoff call and assign a team member to own it — this should not default to the business owner.
Notes: Steps 4 and 5 are the most commonly skipped. Consider adding a checklist confirmation before marking onboarding complete.
That output came from one paragraph of disorganized talking. It took about three minutes to produce and is immediately more useful than anything most small businesses have written down about this process.
The One Rule You Can't Skip
Here it is plainly: someone who actually knows the process has to review every AI-generated document before it goes into use.
AI doesn't know your business. It will organize and clarify what you gave it, but it can also fill gaps with reasonable-sounding guesses that aren't accurate for your situation. It might omit a step you mentioned but didn't emphasize. It might standardize a detail you handle case-by-case.
The document it produces is a draft — a very good draft, usually better structured than what most people would write themselves, but still a draft. The expert review step isn't optional. Think of it this way: AI handles the writing work, and you handle the accuracy check. That's still a significant time savings, but it requires human sign-off.
A practical approach: after the AI generates the draft, send it to the person who knows the process with a simple note — "Does this capture it? What's missing or wrong?" Most people find it much easier to edit an existing document than to write one from scratch, which is exactly why this works.
Building a Library Over Time
One-off documentation is useful. A library of documented processes is a business asset.
The businesses that get the most value from this approach don't do it all at once. They build the habit: whenever someone explains a process — in a meeting, on a call, in a training session — that explanation gets captured and turned into a document. Over six to twelve months, you end up with a reference library that makes onboarding faster, delegation cleaner, and the business meaningfully less dependent on any one person.
A few categories worth prioritizing first, because they create the most friction when undocumented:
- Client onboarding — the sequence from signed contract to first deliverable
- Recurring operational tasks — weekly reporting, invoicing runs, inventory checks
- Sales handoffs — what happens when a lead converts and needs to be transitioned to a delivery team
- Common troubleshooting — the ten questions new hires always ask in their first month
These aren't glamorous. But documented, they make the business easier to run and easier to grow.
What This Looks Like in Practice With Clients
This is the kind of foundational work we do early with businesses we work with — identifying where knowledge is bottlenecked in one or two people, and building a practical system to get it out of heads and into documents. The AI piece is just one part of it, but it's often where people feel the most immediate relief: the thing that felt like it would take weeks to write actually takes an afternoon.
Getting Started: A Simple First Step
Pick one process — just one — that lives almost entirely in your head or someone else's. Something you've explained verbally more than once. Set a ten-minute timer, open your phone's voice memo app, and walk through it out loud like you're explaining it to someone brand new.
Transcribe it, drop the transcript into ChatGPT or Claude with the prompt above, and see what comes back.
Then have whoever knows the process best read it and mark what's wrong.
That's it. That's the whole workflow. If the output is useful, you've proven the method for your business and you can start applying it systematically.
If this is the kind of operational thinking you want applied to your business — not just documentation, but figuring out where the real leverage points are — let's talk. A strategy call is a good place to start.