A founder emailed me last month, thrilled. He'd asked ChatGPT to write his company's onboarding manual, and four minutes later he had twelve clean pages. "Look how much time this saved," he wrote. So I read it. It was well organized, grammatically flawless, and almost completely useless.
The real change isn't the one people celebrate
That manual is a good illustration of how AI is changing the way companies document work, because the change isn't what most people cheer about. The tool is fast, tireless, and confident, and none of that is the same as knowing how a business actually runs. What AI really changed is the economics. Generic execution just became free. Anyone can generate a competent-looking onboarding manual in four minutes now, which means the manual itself is worth almost nothing. The value moved entirely to the one thing a model can't produce: what your specific people actually know.
What AI can't know about your work
Here's where the founder's twelve pages fell apart. AI knows how the average company onboards a new hire. It has no idea how yours does, because that lives nowhere it can reach. The exception your best rep makes for a certain kind of customer. The vendor you never call on a Friday. The reason you check one number before you approve anything. That's the wisdom that makes your business yours, and it's sitting in a handful of people's heads, not on the public internet the model trained on.
An AI model was trained on that public internet. Your competitor's model was trained on the same internet. Whatever it can generate for you, it can generate for them. The one thing it can't hand your competitor is what your people know. So the more powerful these tools get, the more valuable your specific, captured knowledge becomes. That's the asset, and most companies are letting it walk out the door every time an expert retires.
Right now you don't own your wisdom. You're renting it. And every day you don't change that, the rent goes up.
Capture beats prompting, every time
Almost everyone gets the order backward. They start with AI. They prompt it to "write our sales process," accept whatever generic thing comes back, and then spend weeks trying to bend reality to match a document nobody in the building actually authored. That's how you end up managing to a fiction.
Do the opposite. Start with the human, and don't hand them a blank template either, because "go write up your process" produces three bored sentences from even your best person. Sit them down and interview them. Get them talking about how they really do the work, exceptions and all, and film it. That raw, specific, slightly messy capture is the gold. It's the part no prompt can reach, because it was never written down to be scraped in the first place.
This is what we call wisdom harvesting, and it's a skill, not a form to fill out. A good interviewer keeps asking the quiet "why" behind each step until the expert surfaces the judgment they didn't even know they were using. The exception. The tell. The reason step four exists at all. That's the material a model can never generate, because it only lives in the room. Get it on the record and you've captured the one thing your competitor's AI can't reproduce, no matter how good the prompt.
A wall of AI text still isn't training
Suppose you get past all that and end up with a genuinely good document. It's still just a document. And documentation stores knowledge. It doesn't transfer it. A library full of books doesn't teach anyone to read. People forget about 70 percent of new information within a day without reinforcement, so ten flawless pages get skimmed once and forgotten by Thursday.
What actually transfers is a video-first playbook: your expert captured on short clips, one task each, backed by checklists and quick knowledge checks, pushed to people in the exact moment they need it. Video carries the tone, the pace, and the "watch out for this" asides no document holds. That's the difference between storing knowledge and moving it into someone else's hands. It's also why we built PlaybookBuilder around capture and adoption rather than cranking out longer manuals faster.
This is a job, not a prompt
The last shift is the one nobody wants to hear. This is somebody's job, not a side task and definitely not a prompt. Companies that actually transfer knowledge give it to a dedicated owner, the person who interviews the experts, captures their wisdom, and drives each playbook until it's Followed By All. And they measure that, because the finish line isn't a full library, it's the same process running the same way on every shift without the founder in the room. It takes a person, a process, and a platform, in that order. We train that owner in-house through certification, because you want the capability to stay after the work is done, not leave with a consultant.
So yes, AI changed how companies document work. It made the writing cheap and the wisdom priceless. The winners aren't the ones generating longer manuals faster. They're the ones capturing what their best people know before it's gone. If you want the fuller version of why that matters, here's why we built this in the first place.
Jon LoDuca
founder
PlaybookBuilder
