Written 19 August 2026. Every number here links to the source it came from. Where no credible public figure exists, I've said so rather than inventing one. Full source list at the bottom.
Most "make money with AI" lists are ranked by how good they sound. This one is ranked by how easy each one is to start.
Number 1 needs one person and one hour. Number 13 needs a year. Everything in between is ordered by what you have to have in place before anybody can pay you, so you can start at the top and stop reading when you hit something you can't do yet. Picking a number 12 idea when you need rent money in three weeks is the most common way people quit.
Before any of it, one number worth sitting with. McKinsey's 2025 State of AI survey found 88% of organisations now use AI in at least one function, up from 78% the year before, while only 21% of the ones using generative AI had fundamentally redesigned any workflow around it. Nearly everyone has bought in. Almost nobody has changed how the work gets done, and most can't point to a profit line that moved. Every service on this list sells into that gap.
Read the order as a sequence, not a menu. The pattern that keeps working: earn with services first, build an audience out of that client work, test the bigger bets once something is paying, then kill everything except the one that worked.
The whole list at a glance
| # | What it is | What starting takes | How fast you get paid |
|---|---|---|---|
| 1 | 1:1 coaching | One hour with one person | Days |
| 2 | Corporate training | One warm intro inside a company | Weeks |
| 3 | Paid courses | Teaching it live, twice | Weeks |
| 4 | AI strategy consulting | One session watching them work | Weeks |
| 5 | Brand sponsorships | Eight weeks of daily posting | Month 2+ |
| 6 | Paid community | 20 people who already ask you things | Month 2+ |
| 7 | Paid speaking | One talk, given free and recorded | Month 2+ |
| 8 | Social media agency | 30 days running your own account | Month 1-2 |
| 9 | AEO agency | An afternoon of screenshots | Month 1-2 |
| 10 | Vibe coding agency | One free build you can show | Month 1-3 |
| 11 | AI automation agency | One automation of your own week | Month 1-3 |
| 12 | Your own product | Months of building before revenue | Month 3+ |
| 13 | Faceless AI channel | A year of posting | Month 3+ |
1. 1:1 coaching
What you sell. An hour of your time, sitting with one person while they set up and personalise their AI tools. Their prompts, their tone, their actual job.
Your first move. Message five people you already know who complained about AI in the last month. Not "do you want coaching." Ask what they tried and where it went wrong. Two of them will describe the same problem, and that problem is your offer.
Where the unserved buyers are. The US Census Bureau's Business Trends and Outlook Survey asks firms whether they used AI in any business function in the past two weeks. By the collection period ending 3 May 2026, 37% of firms with 250-plus employees and 32% of firms with 100 to 249 said yes. Among firms with four or fewer people it was under 20%. Big companies are pulling away. The smallest ones are behind, and they have nobody internal to ask.
Why it's first. No audience, no funnel, no landing page. You need one person and one hour. If you're worried you're not technical enough, the framing that helps is that you're the translator, not the technician.
The honest catch. It doesn't scale and it never will. Treat it as tuition that pays you. Every session teaches you what real people get stuck on, which is the raw material for everything below.
2. Corporate training
What you sell. A live session teaching a team to use the AI tools their company already bought.
Your first move. Find a company that has paid for something nobody uses. In its FY26 Q4 results on 29 July 2026, Microsoft said Microsoft 365 Copilot "reached over 30 million paid seats." That's seats sold, which is the only number a vendor has any reason to report. Whether the people holding them opened Copilot this week is a different question, and the person who signed off on the spend is the one who has to answer it. That's your buyer.
What it pays. One working practitioner publishes $5,000 to $15,000 for corporate AI workshops (sources below). I looked for an independent survey of AI-specific workshop rates and found only content marketing pages quoting each other, so treat that as one person's pricing rather than a market rate.
The honest catch. Procurement is slow and one no-show contact can stall a deal for a quarter. Start with a company where you know a human, not with a form on a website.
3. Paid courses
What you sell. Recorded lessons that solve one problem, sold over and over.
Your first move. Don't record anything yet. Teach the thing live, twice, to people who paid. Their questions become your module list, and their money proves the topic before you spend two weeks in a recording setup.
What it pays. The figure that gets passed around is Write With AI at $400k ARR. Read that as a ceiling, not a median. There's no published benchmark for what a typical AI course earns, and the people best placed to publish one sell courses about selling courses.
The honest catch. Courses look passive from the outside. In practice you're now in the marketing business, and the course is inventory. The people making real money here sell every day.
4. AI strategy consulting
What you sell. Working out where a business is burning hours, then telling them exactly which parts AI should take over and in what order.
Your first move. Ask one business owner for 30 minutes to watch how a single task gets done. Not a discovery call, an actual shadow session. Write down every step, then mark which steps a person has to do and which ones only need a person to approve. That document is the deliverable, and it's usually the first time anyone has written their process down.
Why the work exists. McKinsey's finding is that the strongest link to actual profit is fundamentally redesigning workflows, and only 21% of organisations using generative AI have done it. Companies didn't buy the wrong tools. They bolted them onto processes designed for people, which is precisely the thing an outsider gets paid to notice, and it's why the deliverable above is a process map rather than a tool recommendation.
Where the first clients are. Warm network first: family, friends, ex-colleagues, LinkedIn. Then Upwork, where the rule is uncomfortable but correct: pick one niche, apply to 20-plus jobs in it, and apply within three hours of the job going live. Speed beats polish.
The honest catch. Consulting pays for opinions, and opinions need evidence. If you've never actually implemented anything, your recommendations will be generic and clients can smell it. Do a couple of coaching sessions first.
5. Brand sponsorships
What you sell. Access to your audience's attention.
Your first move. Post daily for eight weeks before you email anyone. Sponsors don't buy follower counts, they buy predictable views, so you need a screenshot showing what a normal post does for you. Then reach the brand manager directly rather than the support inbox.
What the money is doing. Influencer Marketing Hub's 2026 benchmark report, from a survey of 600-plus respondents, found 87.49% expect their influencer budgets to rise and 72.22% expect a rise of 50% or more. More useful for a small account: 51.43% plan to work with more nano creators and 52.83% with more micro creators, while interest in macro creators is flat. Budgets are moving toward smaller accounts, not away from them.
What it pays. One practitioner's published range is $1,200 to $15,000-plus per piece. Published CPM tables for creator tiers are everywhere and almost all of them trace back to nothing, so I'm not repeating one here.
The honest catch. This is the item most people put first and it belongs at five. It's downstream income. You get sponsorships because you were already making content about the client work from numbers 1 to 4.
6. Paid community
What you sell. A room where people get help and, more importantly, get asked whether they did the thing.
Your first move. Start it free with 20 people and run it for a month. If nobody posts without you prompting them, you don't have a community, you have a mailing list. Fix that before charging.
What it pays. The arithmetic is the appeal: 200 members at $50 a month is $10,000 a month recurring. Skool is the usual home for these. Nick Saraev's community gets cited at around $4 million a year, which is the top of the distribution rather than the middle of it. Skool doesn't publish revenue or churn data, and the sites that claim to have it are guessing.
The honest catch. Recurring revenue means recurring obligation. A quiet week in your community shows up as cancellations six weeks later, and you'll feel that pressure every single month.
7. Paid speaking
What you sell. A talk, at someone else's event, on their budget.
Your first move. Give the talk once for free at a local meetup and record it. That recording is the only thing an event organiser actually wants to see. One talk, refined, beats five you've each done once.
What it pays. National Speakers Bureau puts its entry-level tier, which it defines as up-and-coming experts, first-time authors and local business leaders, at $1,500 to $5,000, then mid-range at $5,000 to $15,000. A working practitioner's published workshop figure, $1,500 to $4,000, lands in the same place. Two unrelated sources agreeing on the entry number is about as much confirmation as this corner of the market offers. Bureaus also take a commission, so a booked fee is not a banked fee.
The honest catch. Bookings cluster around conference seasons, so income is lumpy. It also pairs almost too well with number 2, since the same material sells to both, which is why most people who do this end up drifting toward corporate training where the cheques are bigger and the travel is less.
8. Social media agency
What you sell. Done-for-you posting. You take a business's raw material and turn it into a consistent feed, using AI to do the volume that used to need three people.
Your first move. Use your own account as the case study. Run your system on yourself for 30 days, screenshot the before and after, then approach businesses whose last post was in March. They're the easiest sell on this entire list because the problem is visible from the outside.
What you're selling against. In that same Influencer Marketing Hub survey, 66.3% of brands run their creator programmes entirely in-house, against about 11% who work through agencies. Your competition usually isn't another agency. It's a marketing coordinator who has been told to handle it on top of their existing job, which changes the pitch from "we're better" to "you get your Thursdays back."
Tools worth knowing. Blotato for scheduling and repurposing, Canva for carousels. If you need the video and image side, my guides on free AI video tools and free AI image tools cover what's genuinely free and what's licensed for commercial work, because that licence question will come up the first time a client asks whether they own the output.
The honest catch. Content agencies live and die on approvals. Build the approval step into your pricing or you'll spend your margin chasing a client for a yes.
9. AEO agency
What you sell. Getting a business mentioned when someone asks ChatGPT, Perplexity or Google's AI overview for a recommendation in their category.
Your first move. Open ChatGPT and ask "what's the best [their category] in [their city]" ten different ways. Screenshot every answer. If your prospect isn't in any of them, you've got a sales conversation that needs no explaining. If they are, find the competitor who shows up more often and lead with that.
The number that sells this. Ahrefs published their own analytics in June 2025: AI search sent 0.5% of their traffic and produced 12.1% of their signups, roughly 23 times the conversion rate of regular search. Small channel, wildly disproportionate outcome. Semrush's clickstream study, built on over a billion lines of US panel data through February 2026, found ChatGPT's outbound referrals to the web grew 206% during 2025 and reached 170,000 unique domains. Both are first-party datasets from companies that had no reason to undersell traditional search.
What the work actually is. Structured data on their pages, clear answers written near the top of the page rather than buried under marketing copy, and making sure they haven't blocked AI crawlers in robots.txt. That last one catches people constantly. Each bot has its own user agent, and a blanket block quietly removes you from that engine's citations.
The honest catch. Nobody can promise placement here. There are no published ranking signals for AI answers, so sell the work and the measurement, never the position. Anyone guaranteeing a spot is guessing.
10. Vibe coding agency
What you sell. Custom software, built the traditional way for the client's purposes, except you're using AI tools to do it in a fraction of the time.
Your first move. Build one thing for free for someone whose business you understand. It can be small. A booking form that writes to a spreadsheet counts. You need one screen you can show, because "I use AI to build software" means nothing to a buyer and a working demo means everything.
The economics. The numbers put on this model are an MVP in one to three weeks instead of three to six months, at around $1,000 in credits against a traditional $5,000 to $60,000 build. That gap is the entire business model.
The result that should temper your quotes. METR ran a randomised controlled trial in early 2025 with 16 experienced open-source developers across 246 real issues in repositories they already knew well. With AI tools allowed, they took 19% longer. They had predicted a 24% speedup beforehand, and after finishing they still believed AI had made them 20% faster. Worth being precise about what that does and doesn't cover: METR tested mature repositories the developers already knew, which is closer to a client's existing system than to a blank project. My read is that the one-to-three-week figures hold up best on greenfield builds, which is where you should be aiming your quotes anyway. The transferable finding is the perception gap. Your gut will lie to you about this specific thing.
The honest catch, and I mean this one. You are now responsible for other people's data. AI-built apps fail in predictable places, most of them in access control rather than the login form. Read don't vibe code your login before you ship anything for a paying client, and if you're on Supabase, the anon key guide covers the specific way row level security fails silently. Shipping a breach for a client is worse than never starting.
11. AI automation agency
What you sell. Automations that remove manual work. Lead comes in, gets qualified, lands in the CRM, follow-up sends itself.
Your first move. Automate something in your own week first, on n8n or Make, and time how long it used to take. That number is your pitch. Then look for businesses running on copy and paste between two tools, which is most of them.
Why the market is real. Go back to the McKinsey split: 88% adoption against 21% who have redesigned a workflow. That gap is what an automation retainer is actually being paid to close. The tools are already bought and mostly idle.
What sells fastest. Anything that touches money or admin. Lead capture and follow-up, invoice chasing, support triage. A good test for the offer is whether it saves time, increases revenue or removes admin. The ones that remove admin close quickest, because that pain is felt daily.
The honest catch. Automations break. APIs change, a client renames a field, and suddenly nothing has fired for nine days. Charge a maintenance retainer from day one or you've sold yourself an unpaid support job.
12. Vibe coding your own product
What you sell. Software you own, to many people, without a client relationship.
Your first move. Pick one workflow that annoys you personally every week. Not a market you've researched, one you've lived. Build the smallest version that removes the annoyance, then stop building and start telling people. The building is the fun part, which is exactly why it becomes the procrastination.
Where this comes from. It's the natural graduation from number 10. You've built five client tools, three of them were nearly the same thing, so you build that thing once and sell it repeatedly. My Claude Code commands guide covers the tooling side if that's where you're headed.
The honest catch. This is the highest ceiling on the list and by far the highest failure rate. It pays nothing for months. Do it while numbers 1 to 4 are covering your costs, never instead of them.
13. Faceless AI channel
What you sell. Nothing, at first. You build attention, then attach an offer to it.
Your first move. Pick one narrow topic and post daily. Sora 2 and Veo 3 handle the generation, so the constraint isn't production any more, it's consistency. If you want the pipeline for turning an existing site or product into video, website to video with Claude Code and Remotion walks through it.
Read the monetisation rules before you build the pipeline. On 15 July 2025 YouTube renamed its "repetitious content" policy to inauthentic content. To earn money, content must "not be mass-produced, generic, repetitive, or manipulative," and the examples name "AI-generated content made with generic or unoriginal templates giving the impression of mass production" along with videos carrying "minimal variation" across the channel. AI in the process is fine. A template running on repeat is the thing being described. Any faceless model whose only advantage is volume is building on ground YouTube has already announced it will pull.
The honest catch. The advice from people who've built these is brutal and worth repeating: post multiple times a day and commit for twelve months. Most faceless channels die at week six, having produced 40 videos and 300 views. The ones that work treat it as a distribution asset for something else they sell, not as an income source on its own.
Which one should you actually pick
If you need money this month, it's number 1 or number 4. Both are one conversation away and neither needs an audience.
If you already have an audience, jump to 5, 6 or 7, because you've done the hard part and are currently leaving the easy part on the table.
If you can build things, start at 10 and let 12 come out of it. Going straight to your own product with no client work behind you means guessing at what people will pay for, and guessing is expensive.
The one thing that shows up in every version of this list: pick one and stay on it for twelve months. Not because there's something magic about a year, but because the first six months of anything look identical to failure, and people who rotate every eight weeks never get past that stretch on anything.
A note on the numbers above
For coaching rates, course revenue, community revenue, social media retainers and automation retainers, no independent benchmark exists that I could stand behind. What's out there is pages inventing precise-sounding numbers and citing each other. The prices marked above as one practitioner's are exactly that: figures one working operator has published, useful as a reference point and nothing more. Everything else is survey or first-party data, linked in place.
Price your first offer at whatever makes you slightly uncomfortable to say out loud, then raise it after the third client says yes without pausing.
Sources
- Sabrina Ramonov, 12 Ways to Make Money in 2026 (with AI), the money hub and the AI Millionaire Blueprint. The thirteen categories and their order follow hers, as do the practitioner price figures above.
- McKinsey, The State of AI, 2025 global survey.
- US Census Bureau, Business Trends and Outlook Survey, AI use by firm size, collection periods to 3 May 2026.
- Microsoft, FY26 Q4 earnings release, 29 July 2026.
- Ahrefs, AI search traffic conversions, first-party analytics, June 2025.
- Semrush, ChatGPT traffic study, 17 months of US clickstream data to February 2026.
- METR, Measuring the impact of early-2025 AI on experienced open-source developer productivity, July 2025.
- Influencer Marketing Hub, 2026 Benchmark Report, 600-plus respondents.
- National Speakers Bureau, keynote speaker fee tiers, 2026.
- YouTube, channel monetization policies, inauthentic content.