AI Guides › Playbooks
By Nigel Guy · 8 min read
The usual way into "making money with AI" is a list of fifteen ideas, a weekend of tool sign-ups and no first sale. Each idea sounds plausible, so you keep collecting more, and collecting feels like progress. What's missing is a first move small enough to do this week and a number, set in advance, that tells you whether to carry on.
The rule: pick one route, take its first move inside seven days, and decide in advance what result earns a second week. If you can't name the buyer and the first move, you don't have a route yet. You have an idea.
Below are five routes, a card for each and a filter for choosing. None of it promises income; anyone quoting a monthly figure for these routes is selling something.
| Route | What you actually sell | Who pays | First move (this week, no spend) | Platform rule to read first |
|---|---|---|---|---|
| 1. AI-assisted freelance service | A finished piece of work: edited copy, cleaned spreadsheets, summarised research, first-draft proposals | Small businesses and agencies, directly or through marketplaces such as Upwork | Write one fixed-scope offer and produce three sample pieces in that scope | The marketplace's fee page. Upwork, for example, states its freelancer fee ranges from 0% to 15% per contract |
| 2. Workflow set-up for small businesses | A working process: a set of prompts, templates and instructions that handle one recurring task | Owners who repeat the same admin every week | Map one recurring task for one business you already know, and offer to set it up | Data protection. If you handle clients' customer data, check the ICO's data protection fee and whether you are exempt |
| 3. Digital products | Templates, planners, prompt packs, printables | Buyers on marketplaces such as Etsy, or your own audience | Make one product that solves one narrow job, and list it | Etsy's Creativity Standards require you to say in the listing description if an item was created with AI |
| 4. Self-published books | A book you wrote, with AI used for editing or research support | Readers on Amazon KDP and similar | Outline one short, specific book you could write from your own knowledge | KDP asks you to disclose AI-generated text, images or translations. Content you wrote and only refined with AI counts as AI-assisted, and you don't have to disclose that |
| 5. Content channel with AI in the workflow | Attention, which turns into ad revenue, sponsorship or sales | Platforms, sponsors, your own product | Publish three pieces that carry your own view or expertise, using AI for research and editing | YouTube's monetisation policies exclude mass-produced, template-driven AI content that adds no original insight |
Read the rules column as closely as the money column: each platform can switch off your income, and has written down why it would.
Fill in one card for the route you choose before you open any AI tool. It has six lines:
Score each route 0–2 on three questions and add them up. Take the highest score. If two routes tie, take the one with the faster first move.
| Question | 0 | 1 | 2 |
|---|---|---|---|
| Do you already know the subject well enough to spot the AI's mistakes? | No | Partly | Yes, I'd be paid for this without AI |
| Can you reach a buyer this week without paying for ads? | No | Through a marketplace | I already know people who'd consider it |
| Can you tolerate the route's delay before any money arrives? | No | A few weeks | Months is fine |
Services (routes 1 and 2) tend to pay soonest because someone is buying your judgement directly. Products and channels (routes 3 to 5) usually take longer and need an audience you may not have yet.
Use this with any capable chat assistant. Fill in the bracketed parts. Leave a field as "unknown" rather than inventing an answer.
You are a sceptical small-business adviser helping me test one way of earning money with AI tools. Your job is to make my plan smaller and more testable, not more exciting.
My situation:
- Skills and experience I'd be paid for without AI: [YOUR_SKILLS]
- Hours per week I can give this: [HOURS]
- People or businesses I can contact this week: [YOUR_CONTACTS_OR_MARKETPLACE]
- Route I'm leaning towards: [ROUTE_NUMBER_AND_NAME]
- Country: United Kingdom
Steps:
1. If any of the inputs above are missing or vague, ask me up to three questions before doing anything else. Do not guess.
2. Score my route 0-2 on each of these: subject knowledge, reach to a buyer this week, tolerance for delay. Explain each score in one line.
3. Fill in a First-Move Card with exactly six lines: offer in one sentence, named buyer, first move achievable in seven days, proof threshold, the platform or legal rule I must read, and a kill date.
4. List the two most likely reasons this route fails for someone in my position.
Output format: the scores as a three-row table, then the card as a numbered list, then the two failure reasons as bullets. Keep it under 300 words.
Constraints:
- Do not give income estimates, earnings figures or success rates.
- Do not recommend paid tools, courses or ads for the first week.
- If you are unsure of a platform's current rule, say so and tell me to check the official page, rather than stating the rule.
Before you answer, check: is the first move something I can finish in seven days with no spending? Is the proof threshold a number I can count? If not, revise it.
Fill in: your skills, weekly hours, who you can reach, and the route you're leaning towards.
Priya is a hypothetical part-time bookkeeper with eight spare hours a week. She scores route 2 (workflow set-up) at 5: she knows bookkeeping well enough to catch errors, two café owners she already works for have complained about chasing late invoices, and she can wait a few weeks to get paid. Route 3 (digital products) scores 2, because she has no audience and no marketplace presence.
Her card reads: "I set up a weekly late-invoice reminder process for independent cafés." Her buyer is the two café owners. The first move is a 20-minute demonstration, using one café's anonymised invoice list, of an assistant drafting polite reminder emails that the owner approves before sending. The proof threshold is one owner agreeing to a paid month. The rule check is the ICO data protection fee page, because she will be handling the cafés' customer names. The kill date is three weeks out.
She skips the website, logo and course, because none of them would tell her whether anyone will pay.
The trap is building the machine before finding a buyer: brand kits, multi-tool automations, ten listings at once. It feels like work and postpones the one test that matters. The second trap is volume. Every platform in the table has a written rule against unoriginal, mass-produced or undisclosed AI output, and running into that rule can cost you the account, not just a sale.