The People Who Need It Most Have the Least Time to Learn It

Imagine a contractor running a small crew. Four or five people, a couple of jobs live at any time, quotes going out in the evenings because the day belongs to sites and suppliers. He is losing hours every week to the same handful of tasks: writing up what was agreed on a call, chasing a client who went quiet, pulling together the numbers for a quote he has effectively written twenty times before. He is exactly the person an AI system would help most, and he is also the person least able to sit down and learn to build one.

That gap is the thing I keep running into, and I don’t think it gets said plainly enough. The people with the most to gain from automating a business process are usually the people with the least available attention to spend acquiring the skill. Not because they’re incurious — most of the owners I talk to are extremely curious about this — but because the hours genuinely are not there. A business that needs this badly is, almost by definition, a business where the owner is the bottleneck in three places at once.

Why the course format is the wrong shape for this buyer

The standard answer to this is a course. Several weeks, a set of sessions, exercises between them, and at the end you can build the thing yourself.

I want to be careful here, because I’m not making the argument that the material in those programmes is bad. Often it’s good. The objection is to the format, applied to this specific buyer. A multi-week course assumes the student has the one resource this buyer doesn’t have, and then assumes it twice: once for the sessions themselves, and again for the work between them. The sessions are the small part. The real cost is the homework, and the homework can’t be delegated, because the thing being built is specific to your own business — your quotes, your clients, your follow-up habits, your record-keeping. Nobody can do that exercise on your behalf. Which means a seven-week programme isn’t asking for seven evenings. It’s asking for seven evenings plus everything in between, from a person who told you at the start that he has zero.

Zero time doesn’t mean tight. In practice it means less than zero — the weeks where the course would sit are already oversubscribed, and something else will lose. Usually the course does, somewhere around session four, and then the money is spent and the problem is still there.

Buying the programme is a bit like buying the gym membership. The purchase is easy and it feels like progress. The workload stays exactly where it was.

What I do instead

The alternative I run is not clever, and I’d rather describe it plainly than dress it up.

First, document. Every word the client says — calls, correspondence, the actual back-and-forth of how the business really operates, not a tidied-up description of how it’s supposed to. This is the slow part and it’s also the part that makes everything after it possible, because a system built on a summary of your business will be about as good as the summary.

Then build it once. Not a template, not a starting point for you to finish — the actual working thing, running against your actual material.

Then remove the ongoing workload, which is the point of the exercise and the only outcome worth paying for.

And then a retainer, which amounts to babysitting: tuning, adjusting when the business changes, catching the places where it has started to drift. I say that last part out loud because the alternative is pretending these systems are set-and-forget, and they aren’t. Something that touches your live business needs someone watching it. A course answers that question badly too — after the final session you are alone with a half-built system and no one to ask.

Why the documentation step is the one that matters

There’s a reason the first step is the boring one.

The AI tooling itself is a commodity. Anyone can buy access to the same models I use; a competitor can have the identical stack by this afternoon. What can’t be bought is the accumulated record of how a particular business talks to its customers — and most businesses are sitting on far more of that than they realise, unread.

I found this out on my own material before I ever framed it as an offer. I had roughly two years of recorded calls — client conversations, internal meetings, several languages — somewhere near 500 hours of audio, plus all the written correspondence over the same period. None of it had been reviewed. It existed as an archive in the most literal sense: stored, and never opened. Pulling that into one readable body of text is what made the system I run on my own pipeline possible, and it’s the asset I’d point to first, ahead of anything about the tooling.

That’s what a course can’t hand you and what an implementation can. Your archive is already there. The question is whether anything ever reads it.

Where this argument stops

I’ll name the limit, because the case above isn’t universal and I’d rather say so than let it sound like one.

If you want to learn this craft — if the interest is genuine and the skill itself is the thing you’re after — then a course is a perfectly sensible purchase, and my whole argument evaporates. It’s built entirely on the premise of no time and no desire to acquire a new discipline. Take that premise away and it doesn’t hold. Some owners will genuinely enjoy building it themselves, and they should.

Equally, if the process you want automated is small and well-defined enough that an afternoon with a chat window would solve it, you don’t need an implementation and you don’t need me. Reaching for a built system when a good prompt would do is its own kind of waste.

The case I’m making is narrower than “don’t take courses.” It’s this: when the buyer’s defining constraint is time, a format that spends time is the wrong shape of answer, no matter how good the content inside it is. Match the format to the constraint first, then argue about the material.

If you want to see what the built version looks like when it’s running against real material, the system I use on my own pipeline is written up at /systems/ and in more detail in The AI That Runs My Pipeline. The part worth stealing, whichever route you take, is the first step — start documenting what your business actually says, now, before you’ve decided what to build on top of it.