Today I ran a masterclass on AI automation.
And the most important moment had nothing to do with AI.
The session was full. People had come prepared. Notebooks open. Questions ready. You could feel the energy in the room - that particular kind that comes when people believe they are about to learn something that will change everything.
And then the questions started.
"Which AI tool should I be learning right now?"
"Should I use ChatGPT or Claude?"
"Can AI reply to my customer messages automatically?"
"Is there an AI that can replace my receptionist?"
Good questions. Honest questions. The kind people ask when they've heard enough about AI to be excited but haven't yet had someone sit with them and help them think it through properly.
I answered them carefully.
But about forty minutes in, something shifted in me.
I noticed that almost nobody…. in a room full of business owners… was asking about their business.
They were asking about software.
The questions were about tools. About platforms. About which AI was smarter, cheaper, faster. About what other companies were using. About what they might be missing.
Nobody asked: "What is the actual process I'm trying to improve?"
Nobody asked: "Where exactly is my business losing time?"
Nobody asked: "What would have to be true before automation would actually help me?"
That silence was louder than any question in the room.
And it became the only thing I wanted to write about today.
The biggest AI mistake isn't choosing the wrong tool. It's automating the wrong thing.
Here is something I wish someone had told me early in my journey with automation:
"The fastest way to create chaos is to automate it." |
Most businesses approach AI the way someone might buy new furniture for a cluttered room. The room is disorganised. The processes are tangled. But the instinct is to add something new rather than to first understand what's already there.
The result is what I call automated confusion.
Every broken step now happens faster. Every inconsistency now happens at scale. Every error that used to affect one customer now reaches fifty.
Automation doesn't fix broken processes.
It exposes them.
If three people on your team answer the same customer question three different ways today, and you automate that reply tomorrow… you haven't solved the inconsistency. You've locked it in. You've made the inconsistency the official policy.
You can't automate clarity into a process that never had it.
This is the step most AI conversations skip. They start at the tool. They should start at the work.
What is actually happening in this process, step by step? Where does it break? Where does it depend on one person's memory? Where does it produce different results depending on who is doing it?
Those questions come before software. Always.
The client who wanted automation. And needed something else first.
I was working with a client who ran a service business. Growing — genuinely growing — but the growth was making everything harder, not easier.
She came to me with a clear request: automate her customer support. Too many messages, too few hours, too much of the owner's time disappearing into replies that should have been handled before they reached her desk.
Before I mentioned a single tool, I asked one question:
"If three different team members answered the same customer message tomorrow morning, would all three responses say the same thing?" |
She thought about it for longer than I expected.
"No. Probably not."
"Why not?"
"Because we've never written down how we handle things. Everyone just figures it out as they go."
That answer told me everything.
The problem wasn't the lack of an AI tool. The problem was that the process they wanted to automate didn't exist yet. It existed as habits. As individual interpretations. As institutional knowledge scattered across the team with no home.
If we had automated that - built an AI system before defining what a good reply looked like — we would have automated the inconsistency.
We would have made the problem permanent.
Instead, we spent two sessions doing something unglamorous but essential.
We defined what good looked like. We agreed on response standards. We built something that anyone…. or any system…. could follow reliably.
Then we built the automation.
The difference wasn't the tool. It was the thinking that came before it.
The Automation Readiness Framework - 5 tests every process should pass.
Before you automate anything, run it through these five tests. If it fails any one of them, the process isn't ready. The thinking is.
1 | Is it repetitive? Why it matters: Automation only creates value when it replaces something you do again and again. Repetition turns automation from a one-time project into a compounding investment. Example: Answering the same ten customer questions every week is repetitive. Handling a unique client situation requires a human. |
2 | Is it documented? Why it matters: If the process lives only in someone's head, it cannot be automated. Automation requires written instructions. If those don't exist, the first task is writing the manual. Example: An onboarding checklist on paper can be automated. A process the founder improvises each time cannot. |
3 | Is it predictable? Why it matters: Automation handles the expected well. It handles the unexpected poorly. If the process requires different judgment every time, it needs human involvement at the decision points. Example: Sending a payment reminder after seven days is predictable. Deciding whether to waive a late fee for a long-standing client is not. |
4 | Does it create measurable value? Why it matters: Not every repetitive task is worth automating. The cost to build and maintain a system must be justified by what it saves or protects. If you can't measure what changes, you can't know if it worked. Example: Automating lead follow-up to recover Kes 40,000 per month in missed conversions creates measurable value. Automating a task that saves ten minutes a month probably does not. |
5 | Would you trust someone else to do it exactly the same way? Why it matters: If you wouldn't hand this process to a new team member with written instructions and trust them to produce the same result — the process isn't ready to run without you. Fix that first. Example: If a new hire could follow your written onboarding steps and deliver a result you'd be proud of, it's ready. If they'd need to call you every five minutes, it isn't. |
What today's masterclass reminded me.
My favourite moment from today wasn't presenting.
It was listening.
There is something that happens in a room full of business owners when they stop performing confidence and start asking real questions. The energy shifts. The questions become more honest. And what people actually say reveals far more than the question they think they're asking.
What today revealed was this:
People aren't overwhelmed by AI. They're overwhelmed by the operations underneath it.
The processes that were never designed. The workflows that grew by accident. The team that never had a playbook because the founder was always there to fill in the gaps.
AI didn't create those problems.
But AI is making them impossible to keep ignoring.
A few moments from today - because the best learning often happens not in the slides, but in the conversations between them.
YOUR ACTION PLAN
The 15-Minute Automation Audit - do this before opening any AI tool.
Fifteen minutes. Before any tool. Before any platform. Just you and honest attention to how your business actually runs.
STEP 1 Tomorrow morning, before you open ChatGPT or any other tool...
Open a blank document and write down everything you do before lunch. Not just the important things. Everything. The small, routine, automatic things are the ones that matter most here. |
STEP 2 Highlight every repetitive task.
Look for anything you do more than twice a week. Responding to similar messages. Updating the same records. Sending the same type of email. Chasing the same type of response. These are your candidates. |
STEP 3 Circle the tasks that follow exactly the same process every time.
Some repetitive tasks change depending on context… those stay human for now. The ones that follow the same steps in the same order every time... Those are your automation opportunities. |
STEP 4 For each circled task, ask yourself:
"Could someone else do this exactly the same way if I wrote out the instructions?"
Yes: the process is automatable. No: write the instructions first. That is the real starting point. |
STEP 5 If the answer is yes...
You have found your next automation opportunity.
Not because of a tool. Not because of software. Because you spent fifteen minutes looking at how your business actually works. That is where every meaningful automation begins. |
5 THINGS TO REMEMBER
Key takeaways from this edition.
→ Automation starts with understanding the work - not choosing the software.
→ Automating a broken process doesn't fix it. It makes the broken version permanent.
→ A process isn't ready to automate until someone else could follow written instructions and produce the same result.
→ The most valuable thing AI can do for your business is make visible the work you've been tolerating for years.
→ The businesses that benefit most from automation aren't the ones with the best tools. They're the ones who did the thinking before the tools.
"Here is something to get clear on…." |
AI can only help you do what you already understand.
It can only automate what you've already defined.
It can only scale what is already worth scaling.
AI isn't becoming valuable because it's getting smarter.
It's becoming valuable because businesses are finally paying atten
Lucy Njuguna · AI Automation Specialist, Nairobi, Kenya
Subscribe at automate-with-ai-lucy.beehiiv.com · Follow on LinkedIn · Join Automate With AI on Facebook
If this changed how you think about automation, share it with one person who needs to read it.