← Back to the book

Free preview

Look inside AI-Powered Online Course Creation

This is the introduction and the opening chapter in full — the same text you get in the bundle, not a rewritten sample. The complete book runs to 45 sections.

See what's included

Introduction — Two Hundred and Fourteen Apologies

Two hundred and fourteen seats, three refunds, and nineteen people who finished.

Wren built a course in eleven weeks. It sold two hundred and fourteen seats over two years, and three people asked for their money back.

Nineteen people finished it.


The version that looked like a success

The build took eleven weeks of evenings and weekends — Wren logged it afterwards at about a hundred and ninety hours. Twenty-four lessons, four modules, a workbook, a slide deck, a set of downloadable templates.

It went on sale at $149.

Over the next two years it sold 214 seats. About $31,900 before fees, spread across two years, most of it in three launch windows.

Three refunds. A refund rate of 1.4%.

The reviews were good. Not effusive, but good — people said it was clear, that it was well organised, that they had been meaning to learn this for a while.

Wren spent two years believing this was a successful course, and there was nothing available on any dashboard to suggest otherwise.


The afternoon that changed it

In month twenty-six, Wren was preparing a second course and wanted to know which of the first course's lessons had worked best. The platform had progression data. Wren had never opened it.

Here is what it said.

Of 214 buyers, 19 had reached the final lesson.

41 had never opened lesson two.

The median learner had stopped at lesson three of twenty-four — and had stopped within the first forty-eight hours after purchase.

Eleven people had never logged in at all.

Wren checked the arithmetic four times. It was a 9% completion rate, which Wren later found out is unremarkable — not good, not catastrophically bad, roughly what a self-paced course sold to a general audience tends to do.

That was somehow the worst part. It was normal.


The apologies

Wren emailed sixty non-finishers. One question: what got in the way?

Thirty-one replied, which is a high response rate and tells you something in itself.

And the answers were almost identical:

"I got busy, that's completely on me."

"I keep meaning to get back to it — it's a great course, I'm just terrible at finishing things."

"Sorry, that's my fault, not yours."

Twenty-six of the thirty-one apologised. To Wren. For not finishing a course they had paid for.

Not one of them asked for a refund. Not one of them said the course was bad. Several of them said they intended to buy the next one.

Which is why the refund rate was 1.4% — and why a refund rate of 1.4% had told Wren precisely nothing.


What was actually wrong

Wren went through the drop-off data lesson by lesson, and the diagnosis took about an hour.

Lesson one was forty-one minutes long. It was an introduction. It contained a welcome, an overview of the modules, a section about Wren's background, and an explanation of how to use the workbook.

The learner did nothing in it. There was no exercise until lesson six.

The sales page did not state how many hours the course required. Wren had never worked it out. When Wren finally did — adding the video runtime, the exercises and the workbook — it came to about twenty-two hours.

Buyers had, reasonably, assumed it was about four.

So they bought twenty-two hours of work believing they were buying four, opened a forty-one-minute lecture that asked nothing of them, and quietly went back to their lives — and then apologised for it.


What Wren changed

The rebuild took nine days, not eleven weeks, because the material already existed.

Lesson one became six minutes and ended with a two-question assessment the learner could complete in four.

The first real exercise moved from lesson six to lesson two.

The sales page gained a line, in the same size type as the price: "This course takes about 22 hours. Roughly 6 of those are video; the other 16 are you doing the work."

And Wren started measuring one number: the proportion of buyers who completed the first assessment within seven days.

Over the following eighteen months, across 176 new enrolments:

First-assessment completion went from an unmeasured baseline of roughly 25% to 61%.

The completion rate went from 9% to 34%.

Refunds went UP — from 1.4% to 3.1% — because the hours line caused some people to realise, correctly, that they did not want to do this.

And revenue went up, because 34% of learners finishing produced testimonials, referrals and a second-course conversion rate that 9% never had.


The line this book is built on

Wren had not sold 214 courses. Wren had sold 214 intentions, delivered 19 outcomes, and been thanked by almost everybody — because a person who does not finish a course does not ask for their money back. They apologise.

And every commercial signal available — sales, reviews, refund rate, repeat purchases — was consistent with a course that worked.


What this book is

Thirty-seven chapters on building and selling online courses as a side hustle: choosing an outcome narrow enough to be testable, designing a curriculum backwards from it, writing lessons and assessments, video and accessibility, launching to a founding cohort, support, pricing, updates, and the arithmetic of a business whose entire cost is paid before the first sale.

It is built on four rules, and every chapter refers back to them.

Rule One — You are selling an outcome you cannot deliver alone.The learner supplies the other half in hours, and most of them will not. Chapter 5.

Rule Two — Nobody who fails your course tells you.They apologise, they do not refund, and every metric you have looks fine. Chapter 6.

Rule Three — The build is paid for before you know anything.A hundred and ninety hours are sunk before the first buyer exists, and the update cost is never budgeted. Chapter 7.

Rule Four — Every claim about what a learner will achieve is one you must defend.Outcome language, credentials, testimonials and accuracy. Chapter 8.


What this book will not do

It will not tell you what the law is where you are, what you may use in a lesson, what an accessibility standard requires, or what you may claim on a sales page. It tells you what to establish and who to ask. Chapter 15.

It will not promise you enrolments, completion, learner results or income. That is Rule Four, and the book applies it to itself.

And it will not tell you that this is easy. The build is long, it is paid for in advance, and roughly nine out of ten of the people who buy it will not see the end of it unless you design deliberately against that — which is what most of the following thirty-seven chapters are about.


Where to start

Read Chapters 5 to 8 first, in order. They are the four rules.

Then read Chapter 20 — exercises and assessments — before you record a single lesson. It is where the first assessment goes, and it is the difference between the two versions of Wren's course.

Then Chapter 26, which is the number nobody publishes about their own course, including almost everybody selling you a course about courses.

Every figure in this book is in U.S. dollars and every figure is blank. Fill them in from your own records.


©2026 James Henderson / https://localhandyman.work

Chapter 1 — Why Online Course Creation Works as a Side Hustle

The marginal cost of the next sale is near zero. The cost of the first one is all of it.

The proposition is that you build it once and sell it many times. That is true. What it conceals is that you pay for all of it before you know whether anybody wants it, and that most of the people who buy it will never see the end.

1.1 The proposition: build once, sell many, and what that conceals

What is genuinely good about it:

The marginal cost of the next sale is close to zero.Which is rare, and is the real prize.

It can be built in evenings.In blocks, around something else.

It does not require you to be available.⚠ Unlike coaching, and unlike a community. Section 3.7.

And it compounds, ⚠ because the second course is built on templates from the first. Section 23.7.

What the proposition conceals:

Every hour is spent before the first buyer exists.Rule Three.

The subject decays, and updating it is never budgeted.Chapter 31.

And the outcome requires labour from the buyer that you cannot supply.Rule One.

Which is why the completion rate is what it is, and why almost nobody publishes theirs. Chapter 26.

1.2 What a learner is actually paying for

Not the videos, which is the thing most creators build first and longest.

A change in themselves.⚠ They want to be somebody who can do the thing. Section 2.2.

A shortcut through the sequence.The order to learn it in, which they cannot work out alone.

Permission to stop looking.⚠ A decision that this is the resource, so they can stop searching.

Structure.⚠ A finite thing with an end, which free material does not have. Section 5.5.

And, last, the information itself — ⚠ ⚠ which is almost always available free somewhere, and which is not what they are buying. Section 2.5.

1.3 Why it is easy to build and hard to finish

You know something. You record it. There is a platform. It can genuinely be on sale in a month.

The build is longer than anybody estimates.⚠ Section 7.2 — Wren's estimate was three weeks and it took eleven.

The scope grows while you build.Section 18.4.

You finish, launch, and sell some.Which feels like the end.

And then the completion data, which you may never look at. Chapter 6.

And month fourteen, when something in it stops being true and there is no budget for the fix. Section 31.9.

1.4 The four rules, and where they come from

All four came out of one progression table opened twenty-six months too late.Resource 3.

Rule One — you are selling an outcome you cannot deliver alone.Chapter 5.

Rule Two — nobody who fails your course tells you.⚠ Twenty-six of thirty-one apologised. Chapter 6.

Rule Three — the build is paid for before you know anything.Chapter 7.

Rule Four — every claim about what a learner will achieve is one you must defend.Chapter 8.

1.5 What this work rewards that other content work does not

Sequencing.⚠ Knowing what must come before what is the actual skill, and it is rarer than subject knowledge. Section 18.3.

Narrowness.Section 10.2 — a testable outcome beats a broad curriculum.

Ruthlessness about scope.⚠ The best courses are the ones with the most cut out of them. Section 18.4.

Finishing.⚠ A course that exists at 70% quality outsells a perfect one that does not. Section 23.9.

And patience with a slow feedback loop, ⚠ ⚠ because completion data takes months to arrive.

1.6 What it punishes that others forgive

Vagueness.⚠ Section 11.1 — an untestable outcome is unrefundable and unsellable.

Being wrong.⚠ A wrong lesson is taught to everybody who ever buys it, forever. Section 8.6.

Silence about the hours.Section 5.3.

Building before testing.Section 7.9.

And assuming the learner is motivated.Section 5.7.

1.7 The central inversion of this book

You are not selling information. You are selling a change in somebody, and you can only supply half of it.

The other half is their hours — and most of them will not spend those hours, will not tell you, and will apologise to you for it.

Not "how many did I sell" but ⚠ "how many finished".

⚠ **Not "what's my refund rate" but ⚠ **"what does the progression table say". Chapter 6.

⚠ **Not "how long is my course" but ⚠ **"how many hours am I asking for, and have I said so". Section 5.3.

1.8 Who does well at this, and who struggles

Somebody who can cut. The instinct to include everything is the single most common failure of design here.

Somebody who has taught this in person, badly, at least once.⚠ You learn the sequence from watching people get stuck.

Somebody who can finish an unglamorous thing.Section 23.9.Resource 1.

And somebody who will look at the progression data.Rule Two.Resource 12.

Somebody who needs the money this quarter.⚠ The build is unpaid and long. Section 7.1.

Somebody who wants to be seen as an expert more than they want somebody to learn something.

And anybody who cannot stop adding lessons.Section 18.4.

1.9 What the first year actually looks like

Month What happens
1 Outcome, twenty conversations, first module
2–3 The pilot cohort, paid, small
4 The first drop-point data. Uncomfortable
5–6 Rebuild the first thirty minutes
7–9 Second launch. Better completion, similar sales
10–12 The first update, and the first refund you decline

The build is the first two months and roughly forty per cent of the year's hours. Almost none of them are paid at the time. Section 7.1.

1.10 Chapter summary

The marginal cost of the next sale is near zero. The cost of the first one is all of it.

Learners buy a change, a sequence and an end — not information, which is free.

The build takes longer than anybody estimates, and the update cost is never budgeted.

You supply half the outcome. The learner supplies hours, and most will not.

Sales, reviews and refund rate can all look healthy while nine in ten buyers see nothing.


A realistic scenario

The three-week course that took eleven.

Wren's estimate, written down in week one: "three weeks, maybe four."

The actual log, kept afterwards:

⚠ Task Estimated Actual
Outline and curriculum 4 hrs 11 hrs
Slides 10 hrs 34 hrs
Recording 12 hrs 41 hrs
Editing 8 hrs 52 hrs
Workbook and templates 6 hrs 23 hrs
Platform setup and sales page 4 hrs 29 hrs
TOTAL 44 hrs 190 hrs

Where the overrun came from, in Wren's own analysis:

Editing, which was estimated at two-thirds of recording time and turned out to be more than recording time.

Re-recording six lessons because the slides changed.⚠ Section 21.8 — the sequencing error that costs the most.

Scope growth: the outline grew from 16 lessons to 24 during the build, all of it justified at the time.

And platform setup, which Wren had budgeted four hours for and which involved learning a system, testing a purchase, fixing access, and writing a sales page from nothing.

What Wren's second course took:

61 hours. Fourteen lessons.

Because the templates existed, the slide format was fixed, the recording setup was already configured, and — most of all — the outline was locked before recording began and nothing was added to it. Section 18.9.

"The first course cost 190 hours and taught me the sequence. The second cost 61 and sold about the same. Nobody warns you that the first one is mostly tuition, and that you are the student."

Wren's note: "I estimated forty-four hours. If somebody had told me one hundred and ninety, I would have built a smaller course — which is exactly what I should have built."


AI prompts for this chapter

Prompt 1 — Is This Business a Fit?

"Ask me ten questions to work out whether online course creation suits me. ⚠ Cover: whether I can cut material rather than add it; whether I have taught this in person and watched people get stuck; whether I can finish an unglamorous thing; whether I can spend roughly a hundred to two hundred unpaid hours before the first sale; whether I need income this quarter; whether I will actually open the progression data; whether my subject will still be true in three years; whether I can state honestly how many hours my course will take the learner; and whether I want somebody to learn something more than I want to be seen as an expert.Then tell me plainly which answers are warnings. Do not reassure me."

Prompt 2 — What My Learners Are Actually Buying

"Help me separate what I think I am selling from what a learner is actually buying. ⚠ I believe they buy: information, videos, my expertise.Argue the case that they are buying a change in themselves, a sequence they cannot work out alone, permission to stop searching, and a finite thing with an end — and that the information itself is almost always free somewhere.Then ask me what change my course produces, in one sentence, and push back if my answer is a description of the content rather than a description of the person afterwards."

Prompt 3 — The Honest Build Estimate

"I am about to estimate how long my course will take to build. ⚠ Ask me for my estimate per task: curriculum and outline, slides, recording, editing, workbook and templates, platform setup and sales page.Then tell me the failure patterns: editing routinely exceeds recording time rather than being a fraction of it; scope grows during the build and every addition feels justified; re-recording after slide changes is the most expensive sequencing error; and platform setup is systematically underestimated because it includes learning a system from nothing.Ask me what I would build if the true figure were four times my estimate — and treat that answer as the real plan."

Prompt 4 — The Central Inversion, Applied

"My course sells reasonably well and has a low refund rate. ⚠ Argue that neither of those numbers tells me whether the course works, because a learner who does not finish does not ask for a refund — they apologise and blame themselves.Then tell me exactly what to look at instead: the proportion of buyers who reached the final lesson, the median stopping point, how many never opened lesson two, and how many never logged in at all.Ask me for those four figures. If I do not have them, tell me where in my platform to find them and treat not having looked as the finding."


⚠ AI checkpoint for this chapter

One — did it treat sales as the measure of a course? Nine per cent completion sells identically to thirty-four.

Two — did it estimate the build from lesson count? Editing and platform setup are where the hours actually go.

Three — did it describe the product as information? The information is free. The sequence and the end are not.

Four — did it assume the learner will do the work? Rule One is that you supply half.


Do This Now

1. Open your platform's progression data. Today, before anything else.

2. Write down what change your course produces, in one sentence, about the person.

3. Estimate your build hours, then multiply by four and decide whether you would still build it.

4. Work out how many hours your course asks of the learner. Add it up honestly.

5. Write down what you would cut if you had to halve the course.


©2026 James Henderson / https://localhandyman.work

That's where the preview ends

The rest of the book — 43 further sections — comes with your purchase, along with the worksheets, the resource library and the full set of AI prompts.

Everything in the book

  1. 01 Introduction Two Hundred And Fourteen Apologies — included above
  2. 02 Why Online Course Creation Works As A Side Hustle — included above
  3. 03 Understanding The Industry And What Learners Actually Buy
  4. 04 The Business Models
  5. 05 What This Work Is Not
  6. 06 Rule 1 You Are Selling An Outcome You Cannot Deliver Alone
  7. 07 Rule 2 Nobody Who Fails Your Course Tells You
  8. 08 Rule 3 The Build Is Paid For Before You Know Anything
  9. 09 Rule 4 Every Claim About What A Learner Will Achieve Is One You Must Defend
  10. 10 Courses You Do Not Build
  11. 11 Choosing The Subject The Learner And The Outcome
  12. 12 Writing The Outcome Statement
  13. 13 Researching Demand And Competitors With Ai
  14. 14 Creating A One Page Business Plan
  15. 15 Startup Costs And A Realistic Budget
  16. 16 Legal Copyright Privacy Accessibility And Claims
  17. 17 Choosing The Platform And What It Decides
  18. 18 Pricing Tiers And Cohorts
  19. 19 Curriculum Design From Outcome Backwards
  20. 20 Lessons Scripts And The First Thirty Minutes
  21. 21 Exercises Assessments And The Evidence Of Learning
  22. 22 Video Audio And What Production Quality Actually Buys
  23. 23 Accessibility Captions Transcripts And Who Cannot Use It
  24. 24 Building The Course Without Building All Of It
  25. 25 The Launch And The Cohort That Goes First
  26. 26 Support Questions And The Learner Who Is Stuck
  27. 27 Completion The Number Nobody Publishes
  28. 28 Selling The Course Pages Proof And Honest Claims
  29. 29 Finding The First Hundred Learners
  30. 30 Records Money And Tax
  31. 31 Build Hour Arithmetic And What A Course Is Worth
  32. 32 Updates Decay And The Course That Stops Being True
  33. 33 The Shape Of A Working Month
  34. 34 Refunds Complaints And The Learner Who Did Not Finish
  35. 35 Standard Operating Procedures And Quality Control
  36. 36 Tracking Money And Learning
  37. 37 The Thirty Day Ninety Day And One Year Plans
  38. 38 Using Ai Responsibly In Your Business
  39. 39 Resources Part One Getting Started
  40. 40 Resources Part Two Setup Compliance And Pricing
  41. 41 Resources Part Three Curriculum And Delivery
  42. 42 Resources Part Four Money Pricing And Systems
  43. 43 Resources Part Five Learners Growth And Records
  44. 44 Resources Part Six Scale Review And The Prompt Library
  45. 45 Four Rules One Progression Table And The Learners Who Apologised