Education, then SOPs, then automation
Almost every team starts by automating something. That is the third step, and starting there is why it never sticks.
“The automation comes at the very, very end.” I say that in every session now. It is the part nobody expects, and the part that makes the rest work.
Stage two has a page of its own, with the prompts that do the writing: The SOP Generator →
Education
Teach people what it can and cannot do
Before anything gets built, get everyone comfortable with the thing itself.
A plain conversation about what AI is good at, and where it falls over.
- Who on the team has never actually opened it?
- What do they quietly think it will do to their job?
- What is one small thing each of them could try this week?
You could skip this and go straight to stage two. I would not. People are intimidated by it, and a generic conversation is what eases them into the water.
SOPs
Write down how the work actually happens
An SOP is the write-up of how you do a task, clear enough that someone else could run it.
One per workflow, written before AI touches any of it.
- Which job stalls completely if you are off for two weeks?
- What does everybody know and nobody has written down?
- Who do people go to, instead of going to the document?
This is also how you find out where your time goes. No separate exercise needed. Write the workflows down and the repetition is sitting there on the page.
The work worth automating is the work you stopped noticing.
Automation
Last, and only on a process you can already see
Now pick one thing. Not six.
Build it properly, finish it, then spend the time it gives you back on the next one.
- Which single job costs the most time every week?
- What sets it off: an email arriving, a meeting ending, a payment?
- When it goes wrong, who finds out, and how?
A person stays in the loop. I built one for a music school: an enquiry arrives, a reply is drafted with the bits it cannot know left blank, and the owner fills those in and sends. Unpaid, built to prove the idea.
How it goes with a real team
Four moves, in this order. The tools I take to each one are in the skills toolbox.
- 1
Talk to them about AI first. What it does, what it does not, and no build anywhere in sight.
- 2
Write one SOP per workflow. Messy notes are fine. A prompt turns them into the document.
- 3
Read the SOPs back cold. Where is time leaking, and what could a machine take over tomorrow?
- 4
Automate one thing and finish it before you start the next. Never six at once.
Stage two is what makes stage three worth buying. A task that eats five hours a week, times everyone who does it, is the whole prize. You can only put a number on it once the SOP exists.
What stage two is worth on its own
Cover
when someone is out- Somebody else can run the job.
- No frantic phone call on a sick day.
Onboarding
the first week- A new starter reads it instead of asking.
- Training stops eating someone’s afternoon.
What is in heads
before it walks out- Specialists retire and take the how with them.
- Written down, it stays in the business.
Automate nothing at all and the business still keeps every one of those. One client I teach gave it the right name: knowledge management.
- Education: what the tool can and cannot do, before anything is built.
- SOPs: how the work really happens, one document per workflow.
- Automation: one at a time, finished, with a person still in the loop.
- The order: stage two pays for itself even if stage three never happens.
This week: write one SOP. Pick the job you would most dread handing over.
Do stage two now, with the prompts that write it for you: The SOP Generator → Or see what other people hand over first: Use cases →
Do it with someone. Everything here is free and you can run it alone. If you would rather work through it with me, one to one, book a free 30-minute chat. No pitch.