How to build a one-person business with AI: a 90-day plan

Most advice about running a one-person business with AI is a list of tools. Tools are the easy part. The hard part is the order you do things in, and knowing which jobs AI can take off your plate and which ones it can't.

This guide is the order. It's the plan we teach in our community: one person, a day job, and a first paying customer within 90 days. It's written for software, and most of it works for any small product people pay for monthly.

What a one-person business looks like now

A few years ago, a small software product needed at least a developer, a designer and someone to sell it. Teams of three or four built products that still run and pay their founders well, but the build took most of a year.

AI coding tools collapse that build from months to weeks. One person can now do what that team did, if they work in the right order. Two things haven't changed:

• People still only pay for things they already want. AI doesn't help you guess better. It helps you check faster.
• Quality still sells. Products built with AI that skip design all look the same, and buyers notice. A small product has to look trustworthy to get paid.

So the job is: pick something people already pay for, build a smaller and nicer version of it, and charge from day one.

The 90-day plan at a glance

1. Weeks 1–2Pick an idea that already makes moneyDone when you have one idea, a competitor list and a written angle.
2. Week 3Cut the scope, name it, open a waitlistDone when a one-page site and waitlist are live.
3. Weeks 4–9Build with AI, post weekly updatesDone when 100+ people are waiting.
4. Weeks 9–10Set your launch priceDone when prices are live in Stripe and on the site.
5. Weeks 10–12Charge your first usersDone when someone has paid.

The email to your waitlist in week 12 has a fixed date. Whatever state the product is in, it goes out. That deadline is what turns a side project into a business.

Weeks 1–2: pick an idea that already makes money

Don't look for a new idea. New ideas need you to teach a market that it has a problem. Look for a category where people already pay, and an angle the existing products ignore.

Every tool you or your last workplace paid for

Categories with 3–5 established products

…where reviews keep repeating the same complaints

…that one person can run

One idea

1. List every piece of software you or your last workplace paid for. Each one is a category with paying customers.
2. For each category, find the three to five main products. Note when each was founded and roughly how many people work there. Products that have survived for years with a real team are proof of a real market.
3. Screenshot their pricing pages. Anything charging $100+ a month for something cheap to run is a gap.
4. Read their 1-, 2- and 3-star reviews on G2, Capterra, the App Store or Trustpilot. "Hard to use", "ugly", "expensive" and "support is slow" are the openings you're looking for.
5. Search Reddit and Google for "[product] alternative". If people are asking, those are your first customers and your first blog posts.
6. Cross out anything risky for one person: ideas that depend on someone else's API, have costs that rise with every user, or are genuinely new.
7. Pick one and write a one-page "why we win" note: who it's for, which product you're taking customers from, and your angle (simpler, cheaper, or for a niche the big product ignores; usually two of those three).

Give yourself 14 days. If you haven't started building by then, the idea isn't pulling you. Go back to the list.

Week 3: decide what "good enough" looks like

Open each competitor's help centre and changelog and list every feature. Then mark the ones that show up on their homepage, in their onboarding and in their most-read help articles. Those are the features customers actually use.

Cut your list to three to five of those, done properly. Write a one-page spec: what's in, what's out, and what waits until after your first customer.

Then, in the same week:

• Pick a short name and a domain. Don't spend more than an afternoon on it.
• Sketch the main screens before you write code. If you pay for one thing in this whole plan, pay a designer a fixed fee for the five main screens. It's the cheapest way to look like a real company.
• Put up a one-page site with a headline, three benefits, a few screenshots and a waitlist form.
• Ask for early users. Post in the subreddits and Facebook groups where your future customers spend time. Describe what you're building and ask who wants early access.

Weeks 4–9: build it, in public

Build feature by feature from the spec with an AI coding tool, working from your designs so the result looks like your product rather than the tool's default. Use a mainstream stack the tool knows well, hosted login and database, Stripe for payments, and analytics with session recordings from day one.

Two habits matter more than the code:

• Post a short update to your waitlist every week. It keeps people interested, and it's your build-in-public content.
• Ship to a test address every few days and use it yourself.

Aim for 100 people on the waitlist by week 9. Founders who've run this plan have reached 200 before writing any code; 100 is enough to find your first paying customers.

Weeks 9–12: charge from the first user

No free plans. A free trial is fine; a free-forever plan attracts people who will never pay and won't tell you the truth.

1. Set a launch price. Look at what similar products charge and sit in that range. An early "founding member" price rewards the people who trusted you first.
2. Write the one email that turns your waitlist into customers: what's ready, what isn't, the founding price, and the deadline. Send it in week 12, finished or not.
3. Turn on Stripe notifications on your phone. Every payment pings you.
4. Reply personally to every paying customer within a day. It's where your best feedback comes from, and it's the one thing a solo founder does better than a big company.

One reply that pays is the milestone. Everything after that is turning one customer into a business.

After day 90: from one customer to a business

The first customer proves people will pay. The next phase is about finding more of them without hiring anyone.

Growth at this stage is steady, not explosive. That's fine. A product that adds a few customers every week compounds.

What AI does well, and what it doesn't

The founders who do well with AI treat it like a capable assistant who needs clear jobs and has its work checked. The ones who struggle ask it to "run the business" and wonder why nothing happens.

Frequently asked questions

Can one person really run a software business with AI?

Yes, if it's a focused product in a market that already pays. AI covers most of the building and a lot of the admin. You still do the deciding, the designing and the selling.

Do I need to know how to code?

It helps, but it's no longer required for a first version. AI coding tools can build a working product from a clear spec. What you do need is the patience to test it yourself and fix what's wrong.

How much does it cost to start?

Mostly your time. A domain, hosting, an AI coding tool and a few subscriptions add up to tens of dollars a month. The one cost worth paying for early is design.

Should I quit my job?

No. Keep it until the business pays more than the job. Staying employed is what lets you outlast competitors who need to make money quickly.

How long until I make money?

The plan above aims for a first paying customer within 90 days. Replacing a salary takes longer, usually years of steady growth. Anyone promising faster is selling something.

Do it with other people

The plan is simple. Doing it alone, around a day job, is the hard part. Heyron members get their own AI agent, with AI usage included, and the Solo Software Studio community, where people work through this plan step by step and share what's working. If that sounds useful, here's what membership includes.