AI Guides › Playbooks
By Nigel Guy · 7 min read
Most people studying solo AI businesses read the revenue screenshot and skip the operating model underneath it. They then copy the surface (an AI wrapper, a launch tweet, a dashboard of numbers) and bolt it onto a working pattern that doesn't match their time, temperament or audience. It feels like research because the examples are real; it fails because you have copied the outcome, not the machine that produced it.
The rule: pick one operating model on purpose, write down what it demands of you, and judge yourself against that model, not against someone else's headline number.
This playbook looks at three named, verifiable founders who each run a different model. It then gives you the Model Fit Card, a one-page check that tells you which model you are actually equipped to run.
Everything below comes from the founders' own sites as they read at time of writing. Revenue figures are self-reported unless stated otherwise, and they change monthly. Treat them as illustrations, not benchmarks.
| Model | Example | How it makes money | What it demands | The catch |
|---|---|---|---|---|
| Portfolio | Pieter Levels | Several independent products, a couple carrying most of the revenue | Constant shipping, a large existing audience, tolerance for most bets failing | You only see the winners; the dead projects are the price of entry |
| Single focus | Danny Postma (HeadshotPro) | One AI product, one clear job, paid per package | Depth: quality, support, conversion and ad spend on one thing | One market shift or one platform change hits everything |
| Ship fast | Marc Lou | Many small launches, then double down on whatever earns | Speed, public building, accepting that most launches earn very little | Revenue drifts to whatever is working this month, which may not be what you planned |
Pieter Levels is the clearest example. His site lists a set of active products, including Photo AI (AI-generated photos and video from a model trained on your images), Interior AI (room redesign), Remote OK (a remote job board) and Hoodmaps (neighbourhood maps). Nomads.com, his longest-running project, has been made free.
The model started deliberately. In March 2014 he set himself the challenge of launching twelve startups in twelve months, mainly to get past the habit of never finishing anything. Nomad List and Remote OK both came out of that run; several of the others did not survive. That is the mechanism: many small, cheap bets, each kept alive only if it earns or grows.
How it actually works:
What it asks of you: a decade-scale following does a lot of the distribution work here. Without an audience, a portfolio is just several products nobody has heard of.
Danny Postma's site describes him as the founder of HeadshotPro, an AI photography company based in Singapore. HeadshotPro does one job: turn a handful of selfies into professional headshots. It sells fixed packages rather than a subscription, and the site says more than 250,000 orders have been placed since 2023 (a company claim, not independently verified).
How it works:
One honest note: HeadshotPro's own site refers to "Danny and his team", so this is a founder-led company rather than a strict one-person operation. That's normal once a single product works; the model tends to pull you towards hiring.
Marc Lou runs it in public. His site lists his products with a live monthly revenue figure next to each, plus two small acquisitions he has sold. The list is long, and the spread is the lesson: at time of writing, a few products carry almost all of the total while many earn close to nothing. His best-known product, ShipFast, is a code boilerplate for launching apps quickly, which is itself a product built around the model.
He also runs TrustMRR, a database of startup revenues that says it verifies figures through connected payment providers such as Stripe, rather than accepting screenshots. That makes his numbers more checkable than most, though the figures shift every month.
How it works:
Fill this in before you choose. Score each line honestly from 0 (not true of me) to 2 (clearly true).
| Question | Portfolio | Single focus | Ship fast |
|---|---|---|---|
| I already have an audience that will look at what I launch | Needs 2 | 0 is fine | Needs 1+ |
| I can live with most projects failing | Needs 2 | 0 is fine | Needs 2 |
| I enjoy improving one thing for years | 0 is fine | Needs 2 | 0 is fine |
| I can ship a usable version in under two weeks | Needs 1+ | 0 is fine | Needs 2 |
| I'm comfortable building and posting in public | Helps | Optional | Needs 2 |
| I have money for paid acquisition on one offer | Optional | Helps | Optional |
Read it as: any "Needs" line you score below its requirement rules that model out for now. If two models survive, choose the one with fewer "Needs" lines you only just scraped.
Prompt for working through the card with a model. Fill in the bracketed parts with your own situation.
You are a sceptical small-business adviser helping a solo founder in the UK choose an operating model for an AI-built product. Your job is to rule models out, not to encourage.
My situation:
- Skills: [WHAT_YOU_CAN_BUILD_OR_DO]
- Hours per week available: [HOURS]
- Existing audience (platform and rough size, or "none"): [AUDIENCE]
- Money I can put in over 6 months, in £: [BUDGET]
- Product idea(s), if any: [IDEAS_OR_NONE]
Steps:
1. Score me 0-2 on each line of this card: existing audience; tolerance for most projects failing; appetite for improving one product for years; ability to ship a usable version in under two weeks; comfort building in public; budget for paid acquisition.
2. Compare my scores with three models: portfolio, single focus, ship fast. State which models my scores rule out, and why, in one sentence each.
3. Recommend one model and the first 30 days of work for it, as a numbered list of no more than eight items.
4. Name one thing I should explicitly not do in those 30 days.
Rules: do not quote revenue figures for any real founder; do not promise income; if any input above is missing or vague, ask me for it before scoring rather than assuming.
Before answering, check: did every score come from something I actually told you? If not, ask instead.
Hypothetical: Priya, a freelance designer in Leeds, has around ten hours a week, roughly 600 followers on LinkedIn, £1,500 to spend, and an idea for an AI tool that turns brand guidelines into social templates.
Her card: audience 1, tolerance for failure 1, appetite for one product 2, ship in two weeks 1, building in public 1, paid budget 1. The portfolio model fails on audience and failure tolerance. Ship fast fails on speed and public building. Single focus survives cleanly. So she builds one tool for one buyer (small agencies), charges per use, and spends her first month on output quality and ten conversations with agency owners, not on a second product.
Survivorship. You are looking at three people whose models worked, chosen because they worked. The losers in each model don't publish dashboards. Two specific versions of the trap: