The Demand Engine

What you could build

Three worked examples, in enough detail that you could start one on Monday. They are here because the honest version of "start your own thing" is not a course promising a crore a year. It is picking which relationship you want with money first, and then building something that fits it.

These are not three businesses so much as three different reasons to do the work. One pays you. One pays someone else back. One you would do anyway. Read all three even if you only want one, because the second and third only work when you are clear about the first.

01 Local businesses 02 Your community 03 The thing you love Build your own

Path 01 · This one pays

An AI business for the businesses near you

You are creative and you can build things. There are a few hundred businesses within ten miles of you whose website was last touched in 2019, who are paying someone monthly for something they cannot explain, and who have never once been asked what they actually want. Proximity is the advantage no remote agency can copy.

Designed to make money

Who you are actually selling to

One person, usually the owner, who decides in a conversation rather than a committee. They are not evaluating five vendors. They are deciding whether they trust you.

That means the whole machine collapses into something small: you need ten people who already know you, and one genuinely useful thing to do for them.

A ladder that works at this size

Something useful, free, once$0
A paid look at one thing$500–1,500
The fix, scoped$2,500–8,000
Keeping it working$500–2,000/mo

Four to six retainer clients is a living. That is the whole target, and it is closer than it sounds.

The room

A conversation, usually in their shop, usually while they are doing something else. There is no landing page in this business for a long time.

What converts is showing them something specific about their own business that they did not know. Not a template audit. One real finding.

What the tools actually do here

Genuinely well: research before a meeting, first drafts of everything, building pages fast, producing the audit, writing the follow-up you would otherwise skip.

Not at all: the trust, the walking in, the noticing that their busiest day is Thursday. That part is the business.

Your first month

Write down ten businesses you already have a reason to talk to. Not cold. The dentist you go to, your friend's restaurant, the shop you buy coffee from. Cold outreach is a different and much harder business.
Pick one and do something genuinely useful for free. Not a free audit, which is a pitch wearing a costume. Fix the thing that is actually broken. It will take you an afternoon.
Charge for the second thing. The moment they ask for something else is the moment you have a business. Name a price out loud. It will feel worse than it is.
Write down what you did, with the number. That is your first case, and you will use it for two years.

What actually goes wrong You build yourself a job rather than a business. Six clients who all need you personally, every week, is employment with worse benefits. The fix is boring and it is real: write down how you do the thing, so somebody else can do it.

The second risk is that a machine eats your product. One firm already sells a productized SEO audit at $750, delivered in twenty-four hours, by stripping out the call and the presentation. Anything you sell that a machine can produce alone will meet that price. What survives is the part that needs a person in the room: the interviews, the local knowledge, the noticing.

Masterclass track: C, selling to a business, though your committee is one person. Also read Track C and the Rung Two catalogue for what to call things.

Path 02 · This one pays someone else back

An AI business that feeds people

You care about people who have less, and you like putting food in their hands. There is a version of that which is a weekend, and a version which is an organisation, and the difference is almost entirely about whether you can fund it without burning yourself down.

Nonprofit does not mean unpaid

The one structural fact that changes everything

The person you serve and the person who pays are two different people. Every other path on this page has one customer. This one has two audiences with different needs, and almost every failure here comes from serving one and forgetting the other.

Practically: the work is with the people who need food. The communication is with the people who fund it. Those are different jobs, they take roughly equal time, and nobody warns you about the second one.

The ladder is funding, not pricing

Your own money and timeMonth 1
People who know youMonth 2–6
Small recurring donorsThe real base
Local business sponsorshipIn kind, usually
GrantsSlow, large

Recurring small donors are the rung that makes it survivable. Grants are slower and larger and will not save you in year one.

The room

The story, told with specifics. Funders decide on narrative, and vague narrative reads as no narrative.

The specifics that work are small and countable: this many meals, this street, this Tuesday, this is what it cost. Not "fighting hunger in our community."

What the tools actually do here

Genuinely well, and it is a lot: grant applications, which eat an enormous share of a small organisation's time. Donor updates you would otherwise not send. Volunteer scheduling. Route planning. Translating materials into the languages people on your street actually speak.

Must not: write in the voice of someone you serve, or produce a story about a person that did not happen. That is the same violation as an invented testimonial, and here it is also a betrayal of the person.

Your first month

Do one route, one week. Whatever you can carry, wherever you already go. Do not incorporate anything yet.
Count what happened. How many people, what it cost you, how long it took, what you ran out of. This is your entire case for funding later.
Tell fifteen people you know, with those numbers. Not an appeal. Just what you did. Some of them will ask how to help, and that is your first funding conversation.
Do it again the next week. The second week is the one that tells you whether this is a passion or a Saturday.

What actually goes wrong, and one thing to get right early The work is real and the funding never stops being work. People who burn out here are almost never tired of the service; they are tired of asking. Build the asking into the routine early, at a size you can sustain, rather than as an emergency every six months.

Charitable solicitation is regulated, state by state. If you start asking for money publicly, there are registration requirements, and they vary. Look them up before the first appeal rather than after. Claims about how donations are used are enforceable in the same way as any other claim, so if you say ninety percent goes to food, be able to show it.

And say this out loud to yourself early: being a nonprofit does not mean you work for free. Organisations that refuse to pay their founder tend to lose their founder.

Masterclass: the path modules 00, 0A and 0B, then Foundations and Guardrails. Module 04 matters more here than anywhere, because the people in your stories cannot consent after the fact.

Path 03 · This one you would do anyway

An AI business built around dancing

This might make a little. It might make a lot. The reason to be careful here is not that it will fail. It is that the failure mode of a passion business is that it stops being a passion, and that happens through a specific and avoidable mechanism.

Money optional, by design

Decide this before anything else

What does this have to earn for you to keep doing it? Write the number. If the honest answer is zero, because something else covers your rent, you have just made every subsequent decision easier and you can build the version that stays fun.

If the answer is a real number, you are building a business that happens to involve dance, and it needs the same discipline as path one. Both are fine. Pretending it is the first while needing it to be the second is what goes wrong.

The ladder

Free, in a room, regularlyThe base
Paid workshop$20–60
A series people commit to$120–300
Recorded, sold online$27–95
Teaching other teachersHighest, slowest

The room, literally

The class is the room. People buy the feeling of the room they were already in, which is why the free regular thing is the base of the ladder rather than a marketing tactic.

Be realistic about the online rung. Selling a physical practice as recorded video is genuinely hard, and the numbers are sobering: median free-to-paid conversion for online audiences runs around 0.62%, roughly ten times below what the platforms advertise.

What the tools actually do here

Genuinely well: everything around the thing. Scheduling, class descriptions, the newsletter you keep not sending, editing video, working out music licensing, finding spaces, handling the fifty messages.

Not at all: the dancing. Which is the point, and is also why this path is more resilient to automation than the other two.

Your first season

Run it free, on a schedule, for one season. Same day, same time. The schedule matters more than the promotion.
Notice whether you still want to do it in week seven. Week one is easy for everybody. Week seven is the honest test, and it is worth more than any market research.
Ask the people who keep coming what they would pay for. They will tell you something you did not expect, and it is usually not the thing you were planning to sell.
Charge for exactly that, and keep the free thing free. The free room is what feeds everything else. Closing it to monetise it is the most common way this dies.

The specific way this one goes wrong You monetise the part that was the joy, and then it is a job you are worse paid for. It usually happens by accident, in a specific order: the free thing starts feeling like unpaid work, so you close it, and closing it removes the room that made people want the paid thing.

The people who avoid this tend to do one structural thing: they keep the free thing genuinely free and unhurried, and let something else carry the income. Which is a real argument for running path one at the same time. Path one funds path three, and that is not a compromise, it is a design.

Masterclass track: A, an individual selling online. Module 07 in that track is the arithmetic one, and it is the one to do before you build anything for the online rung.

If none of these are yours

Good. They are examples, not a menu, and the whole point of the first three masterclass modules is that an idea generated from a list of trending businesses produces something nobody wants to run on a Tuesday in February.

The three path modules do this properly. Module 00 gets you honest about what this has to earn versus what you would do anyway, which decides more than any market research. Module 0A generates ten candidates from what you already have, meaning what you get asked for, what you find broken, and what access you have that most people do not. Module 0B kills nine of them cheaply, before you build anything, and makes you test your favourite first rather than protecting it.

The shortest version, if you do nothing else. Write two sentences: the number this has to earn for you to keep going, and what you would still do if it earned nothing. Where those two point at different work, you have found the thing worth knowing, and it is usually the most useful hour anyone spends on this page.

Then the same five stages apply to whatever you pick, the same as they do to a $200,000 agency engagement. Who exactly, and where do they already gather. What one thing will you be believed for, and what backs it. Where do they decide. What are the rungs. And how will you know it worked.

Start at module 00. Then read the three tracks to work out which one your idea sits on, because it changes what you build next.