Every founder I talk to opens with some version of the same sentence: I don't have time to learn AI. They say it the way you'd decline a gym membership. I get it, and I also think they're aiming at the wrong target.
They're picturing a curriculum. Twelve hours of tutorials, a prompt engineering course, maybe a certificate at the end. That is not how anyone I've ever trained actually learned anything, and it's certainly not how you get to your first working backend automation.
The faster route, the one I've watched work over and over, is this: take a job you already do every single week and make the machine do it with you. The learning happens on the way to a finished thing, not before it. That's the whole point.
The time was never the problem. The syllabus was.
I've been teaching people software since 1998, more than 90,000 of them at this point. The ones who learned were never the ones with more free hours. They were the ones whose learning was bolted onto work they had to finish anyway.
Somebody staring at 40 unanswered inquiry emails learns Zapier on a Tuesday afternoon. Somebody with a clean inbox and a bookmark to a course learns it never. Same person, same brain, different stake in the outcome.
A marketing studio owner I worked with spent a full weekend on tutorials. By Sunday night he could describe an agent architecture in detail. On Monday he was still personally copying every new lead from his contact form into a spreadsheet. He had learned about AI for consultants and agencies. He had not automated one minute of his own agency.
That gap is the whole problem. The general curriculum has no skin in the game, so your brain files it as optional. Your actual Tuesday afternoon does not feel optional.
Pick the job you already hate, not the one that would impress you
Before you touch a tool, choose the workflow. This is the decision that determines whether the sprint works.
Four criteria. It happens at least once a week. It follows the same sequence of steps almost every time. You are the one doing it, personally, with your own hands. And it is boring enough that you zone out somewhere in the middle.
For service businesses, the candidates are usually pretty obvious:
- An inquiry comes in, you read it, judge fit, write a reply, and file it somewhere.
- A discovery call ends, you write up notes, then draft a proposal from those notes.
- A client signs, and you send the welcome packet, schedule the kickoff, and set up the folders.
- Friday arrives, and you pull numbers and write the client update.
Here's how to find yours in ten minutes. Open your sent mail. Scroll back two weeks. Find the email you typed the most times with only minor variations. That's the one. Not the biggest problem in your business, the most repeated one. Repetition is what makes automation worth building.
The four-hour sprint, hour by hour
Block the four hours. Not four scattered hours across a month. One block, door closed, phone face down.
Hour one: write down what you actually do. Not the idealized version. The real one, with the sticky note and the workaround where you paste from one system into another because they don't talk. Every step, in order, with the name of the tool you use at each step. Most people discover their workflow has six to nine steps, and two of them are pure copy-paste. Circle those two. They're the ones that are about to disappear.
Hour two: do it once with ChatGPT or Claude sitting next to you. Go step by step. Feed it the real messy input, the forwarded email with the weird signature and the typo, not a clean sample you invented. Save the prompts that produced something usable into a plain Google Doc. If it took three tries before the draft reply sounded like you, that's not failure. That's calibration, and you now have a reusable prompt.
Hour three: wire the plumbing in Zapier or Make. Pick Zapier if you've never built an automation before, Make if you're comfortable and want more control. Trigger on the thing that actually starts your workflow: a form submission, a new email with a subject line pattern, or a new row in Google Sheets. Then the AI step, then a row written back to Sheets, then a draft in your inbox. Use Sheets before Airtable if this is your first build. It's one less new system to learn in the same afternoon. Keep a human approval step. Nothing sends without you reading it first.
Hour four: run it live, once, on a real job. Real inquiry, real call notes, real client. Watch it break on something small. It will break on something small, usually a signature block or a forwarded thread or a name it guesses wrong. Fix one thing. Then stop. Do not fix all of it today.
What "working" means at hour four
It does not mean polished. It means it saved you twenty minutes and it ran without you remembering to run it.
That's the bar. Twenty minutes a day is roughly seven hours a month, from one boring workflow you built on a Thursday. It will not be the thing you demo to anyone. You'd be embarrassed to show it. That's a good sign.
The failure mode is overbuilding. Three branches, a lead scoring rubric, custom code, a dashboard nobody asked for. That's the version people abandon in week three. Keep it boring, keep it small, keep yourself in the loop until you trust it, then take yourself out of one step at a time.
You also want to be honest about what a first AI workflow for a small agency will and won't do. It won't run your business. It will remove one repeated, annoying task from your week and give you back the attention you were spending on it.
The second sprint is where this compounds
Here's what nobody tells you. The first one takes four hours. The second takes two. The third takes ninety minutes, because by then you're not learning a tool, you're applying a pattern you already know: trigger, AI step, record, human approval.
That's the point where founder operations systems stop being an abstraction. When intake, follow-up, and delivery handoffs all have a path that doesn't route through your head, you stop being the bottleneck. Not because you hired, and not because you bought software. Because the work has somewhere to go that isn't you.
If you want to see what that looks like built live in a room with other founders, we run it in the workshop. And if you'd rather see how the longer version of this works across a whole business, here's the overview.
Pick the boring job this week. Block the four hours. You'll learn more about how to learn AI for your service business when you have no time in that one block than in every tutorial you've bookmarked since 2023.
If this resonates and you want to build this kind of infrastructure in your own business, the Mastermind is where we do the work live. You can learn more at mastermindshq.business.