The Fastest Way to Get What's In Your Head Onto Paper
Say you write down how to open your café in the morning.
Unlock, alarm off, lights on. Turn on the coffee machine and let it come up to pressure. Check the fridges. Bring out the pastries. Set the floor. Count the till. Unlock the door at seven.
That is an accurate procedure. You could hand it to someone tomorrow.
Now watch what happens when you do. On day one they ring you because the coffee machine is making a noise. On day three they ring because a pastry delivery is short and they don't know whether to call the supplier or just work around it. On day five they open with a fridge sitting at nine degrees because nobody told them what the number should be.
None of that is in the document. All of it is in your head.
The problem isn't detail. It's the exceptions.
When you describe a process, you naturally describe the version where everything works. That's the version you can see in your mind.
But most of what you actually know is about what to do when something is off. The fridge is warm. The delivery is late. The regular customer wants something you don't normally do. A staff member calls in sick twenty minutes before open.
You handle those without thinking. You've made those calls hundreds of times, and they no longer feel like decisions - they feel obvious. That's exactly why they never get written down. You can't document a decision you don't notice yourself making.
To a franchisee, none of it is obvious. They will hit the same situations in their first month, and they'll either ring you or guess. Both are expensive.
Let the AI ask the questions instead
I've written before about using AI to draft operations manual content: narrate a process out loud, paste the transcript in, and let the tool structure it into a procedure. That works well and it gets you off the blank page.
But narration only ever produces what you thought to mention. And the whole problem is that you don't think to mention the parts you consider obvious.
So reverse it. Instead of talking at the AI, get it to question you.
Give it the context - what your business does, which process you're documenting - then paste in the procedure you've already written. Tell it to act as a new franchisee in their first month who has read this document and keeps running into situations it doesn't cover. Ask it to question you until it has enough to run the process without ringing you.
Then answer properly. Speak the answers if it's faster and paste them in.
What that actually looks like
Take the café opening procedure above. Here's the kind of exchange it produces.
You wrote "check the fridges". Check them for what? What number am I looking for, and what do I do if it’s wrong?
Which forces you to say: under four degrees, and if it’s above that you check whether the door was left ajar overnight, give it an hour, and recheck before you throw anything out - because nine times out of ten it’s a door, not a fault.
You wrote "let the machine come up to pressure". How long does that take, and how do I know it’s ready?
Which forces you to say: about twenty minutes, and you pull the first two or three shots and dump them before you serve anyone, because the early ones are always off.
What if the pastry delivery is short?
Which forces you to say: if it's under about a third, adjust the display and note it for the supplier. If it's more than that, ring them straight away, because they can usually still get a second run out before nine.
Notice what's happening. Every answer contains a number, a threshold, or a reason. None of that was in the original document, and you'd never have written it down unprompted - but all of it is knowledge someone needs to run your café without you.
Twenty minutes of this will usually get you more than an hour of talking.
Four questions worth using
If you'd rather run it yourself, these four do most of the work.
-
"What number am I looking for, and what do I do if it’s wrong?" Turns vague checks into real standards. "Check the fridges" becomes something a stranger can act on.
-
"How would I know if this had gone wrong?" Gets at the quality control you do by eye. You can look at a finished job and know instantly whether it's right. Naming what you're actually looking at is hard, and it's worth more than almost anything else you'll capture.
-
"Why is it done this way and not the obvious way?" Most of your processes solved a problem years ago. Write down the problem, or a franchisee will "improve" the process straight back into it.
-
"At what point should I ring you?" Escalation never gets documented because you are the escalation. Set the trigger - how bad, how long - and a franchisee stops either panicking early or hiding a problem until it’s expensive.
Two things this won't do
It won’t fix vague answers. If you tell it "check the fridges are fine", you’ll get a clean, professional-looking document that’s just as useless as the one you started with. What comes out depends entirely on how specifically you answer.
And it won't tell you whether your way is right. It captures how you do things. It can't tell you whether the same approach will work in a different suburb with different rent, different staff, and different customers. That's a separate question - and it's the one that decides whether franchising will actually work for you.
Getting your knowledge on paper makes it transferable. It doesn't make your business replicable. Plenty of people find that out the expensive way.
Start with one process this week
Not the whole business. One process.
Pick whatever people interrupt you about most. Run the questioning, and have the AI write up the procedure with the reasons included, not just the steps.
Then test it properly: give it to someone who doesn't already know the answers and watch where they stop.
They will stop somewhere. That's the point. Every place they stop is something still living only in your head - and now you know precisely what to go and capture next.
Do that with five or six processes and you'll have a clear, unsentimental picture of how much of your business currently depends on you being reachable by phone.
Worth knowing either way. Essential if you're thinking about franchising.
If you want to do this properly across your whole business, that is what the Franchise-Ready Accelerator is built for. Over 90 days we work one-on-one to get your operation documented, assess whether it can genuinely be replicated, and design a franchise model that works commercially for you and for a franchisee.
Document Business Processes Using AI
-
1
Give AI context
Tell the AI what your business does and which process you are documenting.
-
2
Provide existing procedure
Paste in any procedure you've already written for the process.
-
3
Instruct AI to ask questions
Tell the AI to act as a new franchisee in their first month who has read the document and keeps running into situations it doesn't cover. Ask it to question you until it has enough information to run the process without needing to contact you.
-
4
Answer properly
Provide thorough answers to the AI's questions. Speak the answers if it's faster, then paste them in. Focus on including numbers, thresholds, and reasons behind each step or decision.
-
5
Test and refine
Give the AI-generated procedure to someone who doesn't already know the answers and observe where they stop or have questions. Use these stopping points to identify and capture remaining undocumented knowledge.
Keep Reading
Related Articles
Funding Your Franchise Growth: What Start-Up Franchisors Need to Know
If you're a business owner preparing to franchise, there's something you need to understand early: funding your franchise growth is not like funding a regular business expansion.
Understanding Franchise Fees: What to Charge and Why
A comprehensive breakdown of franchise fees, royalties, and how to structure them for mutual success.
The Mindset Shift Every Franchisor Must Make: Operator to Business Builder
One of the biggest surprises for new franchisors is just how much their role changes once they decide to franchise. Running one business and running a franchise network are two very different things. ...
