All 9 of our rounds, plus 6 new research tracks (300+ sources). Every number is tagged. Where a figure is our model, it says so.
Building got cheap. The money is in making AI work inside a business, paid monthly. Six businesses already paid you for it. Make it one offer, sell it to people who trust you, then own the product.
How this was built. Six research agents ran in parallel on 26 Oct 2026:
A seventh agent re-read all 9 of our past rounds, 35 report revisions and 58 of your messages.
Where data is thin, we say so. Odds and timelines for you are our estimates built on these base rates. They are not forecasts.
$10k a month = price × customers. Fewer customers means faster money. Each dot is one paying customer.
Media is even further right. A niche B2B newsletter earns about $0.10–0.33 per subscriber a month from sponsors, so $10k a month needs roughly 31–95k engaged subscribers. beehiiv's median time to the first dollar is 66 days for new launches, and 104 for business titles (beehiiv State of Newsletters 2026). Even GameDiscoverCo, the best-known games newsletter, lists $2,500 per issue (rate card).
Why few customers wins on speed. Each customer is a separate act of trust. Four businesses you can reach this month beat a thousand strangers you have to attract first.
Each dot is a way to make money. Up = more likely to reach $10k a month. Left = faster when it works. Green corner = fast and likely.
| Model | Reach $10k/mo | To first $ | To $10k/mo | Fast | Digital | You own it | Fits you |
|---|---|---|---|---|---|---|---|
| AI build-and-run studio (paid builds → monthly retainers)Build an AI workflow on a client's own data, then run and improve it monthly. $1.5–3k pilot + $1–3k/month. | ~30% | 1–4 wk | 3–9 mo | yes | yes | yes | yes |
| Fixed-scope AI subscription serviceOne menu, 3 monthly tiers, AI delivers and a person checks. White Glove Content, AEO Engine. | ~15% | 1–2 mo | 6–12 mo | yes | yes | fully | yes |
| Senior AI contracting / freelancingRent your hours. Fastest, but not a business you own. Upwork volume −4%, AI work +22%. | ~40% | 2–6 wk | 1–4 mo | yes | yes | no | maybe |
| Own media + lead desk (our last plan)Newsletter, tools and a lead desk for one buyer type. Sell sponsor slots and leads. | ~8% | 2–4 mo | 12–24 mo | no | yes | fully | no |
| B2B micro-SaaS from scratchSelf-serve software. Verified $5–50k MRR products have a median age of 34 months. | ~2% | 1–4 mo | 12–30 mo | so-so | yes | fully | maybe |
| App inside a platform store (Apify, Chrome, Shopify)A narrow tool listed where buyers already shop. Fast when it works, but Apify shows only 0.45% of developers at $10k. | ~2% | 2–6 wk | 4–17 mo | so-so | yes | yes | maybe |
| Consumer subscription app4.6% of new apps reach $10k MRR within 2 years. AI apps churn 30% faster. | ~4.6% | 1–3 mo | 4–24 mo | so-so | yes | fully | no |
| Courses / paid communitiesSkool groups over $10k are run by named coaches. The best-documented earners teach 'how to make money'. | ~4.8% | 1–3 mo | 12–24 mo | so-so | yes | fully | no |
| Newsletter / mediaNeeds 31–95k engaged subscribers for $10k. beehiiv's ad network pays only via Stripe. | ~3.5% | 2–4 mo | 18–36 mo | no | yes | fully | no |
| Creator (YouTube, TikTok, LinkedIn)About 4% of creators earn $100k+. Platforms down-rank AI 'slop'. Needs a face. | ~4% | 6–18 mo | 18–36 mo | no | yes | yes | no |
| SEO / affiliate sitesThe #1 Google result gets 58% fewer clicks when an AI Overview shows. Google referrals −40% a year. | ~2.5% | 4–9 mo | 18–36 mo | no | yes | yes | no |
| Scrapers / data tools on ApifyPays to Ukraine, first dollar in days. But payout per tool fell from $70 to $17 a month as tools grew 16× in a year. | ~0.45% | 1–4 wk | 12–24 mo | so-so | yes | yes | maybe |
| E-commerce / Amazon FBA / print-on-demandPhysical goods, so it fails your digital gate. About 6% reach ~$10k profit. | ~6% | 2–6 mo | 12–24 mo | no | no | yes | no |
What the axes mean. Up means the chance that a serious attempt reaches $10k a month within about 2 years. Left means the typical months to $10k a month among the attempts that make it. Bigger dots mean you own more of the business: brand, clients, product.
Hard data. - About 2% of 11,866 verified startups make $10k+ a month (TrustMRR stats). - 4.6% of new apps reach $10k MRR within 2 years (RevenueCat 2026). - 0.45% of Apify developers reach $10k a month (Apify). - 7.7% of 74.9M US independents earn over $100k a year (MBO 2026). - Skool: at most 14.5% of paid groups reach $10k+ (profitable.app).
Our estimates. The studio, subscription and lead-desk dots have no published base rate. We estimated them from verified services (AEO Engine: $170.5k MRR from 56 clients; Windmill: $3,999 top tier) and from the m4 comparison: productized retainers 25–40% within 12 months, contracting 35–50%.
Which channel the winners used matters as much as the model. On TrustMRR, 12–14% of startups that list referral, partnership or affiliate as a channel are at $10k+. For cold email or DM it is 5–6%. For Product Hunt, Hacker News, Reddit and X it is about 1%.
Building got cheap. The money moved to what AI can't fake.
Monthly app launches on RevenueCat went from ~2,000 to 14,700+. 110k MCP servers are listed. Apify tools grew 16× in a year, so payout per tool fell from $70 to $17 a month. Building is no longer an edge.
The #1 Google result gets 58% fewer clicks when an AI Overview shows. Google referrals to publishers fell 40% in a year. LinkedIn flags AI-slop posts, which get 40% fewer views.
24.6% of US firms use AI, but 64% of those changed nothing about how they work. 70.3% of EU firms that considered AI say they lack the expertise. Projects built with a partner reach production about twice as often.
9 rounds in 7 days. The research was good. The loop was the problem.
Apps, channels, marketplaces, 15 scored
Apify + Prozorro 4-week tests proposed, never run. You proposed the AI forward-deployed engineer instead.
open report ↗iGaming PromoDesk: ~7 clients × ~$1.4k via warm intros
The step was to send 10–20 iGaming contacts. Never done. You asked for a new iteration.
open report ↗AngleLab: video ads sold as winning angles
The 4-week test and $300–600/month outreach kit were never approved.
open report ↗9 services for $500–3k, incl. Dubai
You rightly said the AI gave you instructions instead of doing the work. The session ended.
Trader tool vs betting tool
Platform = shelf; you still own distribution. A 45-day test was planned and never started.
open report ↗Tier II → monthly desks → 7 offers, 138 named leads
Once a year, government buyers, then data-trust fear. Cold-email maths gave $0.3–3k a month by month 6.
open report ↗5 audiences to pick from
21 minutes later you asked for your own distribution instead.
open report ↗Trade media + lead desk
The step 'pitch 10 buyers per category before building' was never done.
Games, UK homes, Bali/Phuket, importers
First money in weeks 8–10. Year-1 net +$2.5k (Base) to +$9.6k (Push). Bali/Phuket commissions fail 'digital'.
open report ↗Ideas were never the problem. Nobody was ever asked to pay. The only real proof we have is that six businesses already paid you.
At $2.5k per client you need only 4 yeses. That means about 2 hours of calls a week. AI writes every message, proposal and invoice.
Start with doors you already own: past clients and your contacts. A public voice can come later, if you want one.
Charge a paid pilot now (one-off), and turn it into a monthly retainer at the demo.
The only proof is a paid invoice. Day-14 gate.
Sell first to those who already trust you with their data, and run it on the client's own accounts and keys.
Sell digital AI systems to Bali/Phuket businesses for a monthly fee, not commission on deals.
The full graveyard is in research_notes/Money path/m7_history_synthesis.md: 72 ideas, each with its round, price, modelled money, status, the reason it died, and the assumption nobody tested.
One business, three steps. Digital, fast and yours end to end.
Your business: an AI build-and-run studio under your own brand. Build one AI workflow on a client's data, then run it monthly. Sell it first to people who already paid you.
"We put one AI workflow to work on your own data in 2–3 weeks, judged on your past cases. Then we run it and improve it every month."
Why not a generic 'AI automation agency'. Chatbots, voice receptionists and AI SDRs for tiny businesses are the most crowded corner of the market. - Liam Ottley's own numbers work out to ~0.14 clients per trained agency over 4 years. - 40,000+ firms applied to Anthropic's partner programme. - A tool for chatbot agencies lost 64% of its revenue in a year. - 5 verified cases of cheap AI sold cold to tiny firms stayed under $5k a month.
Your edge is the opposite: bigger custom work for firms that already trust you.
Price and retention. AI products under $50 a month keep 23% of their revenue after a year. Above $250 a month they keep 70% (ChartMogul). Price high.
Which lane? We don't pick it for you on paper. We pick it in week 7, from what two real clients bought. The three lane cards are where your past builds point.
Getting paid. For services: FOP group 3 plus a Payoneer payment request or a direct USD wire. For a self-serve product: Paddle, which supports Ukrainian sellers but bans gambling. Apify also pays out to Ukraine. Shopify pays via Hyperwallet to a real USD bank account, not Payoneer. Stripe is effectively closed, even through an Estonian e-Residency company (Stripe, Paddle).
The 2027 tax cap. The FOP group-3 limit rises to UAH 11.14M in the draft budget, about $265k (DTKT). $10k a month is far below it.
Optional AI-only side bet from week 2: an Apify tool portfolio. 6–10 reliable scrapers in ad-library and resale-price niches where the current leaders are weak (for example, the #2 Google Ads Transparency tool is rated 1.93★). First paying users come in 1–3 weeks, and Apify pays out to Ukraine. There is roughly a 35% chance of $3k a month by month 9 (judgement, not data). Risks: site terms (Google, LinkedIn and eBay forbid scraping) and Apify releasing official tools. It is not the $10k engine. Full analysis: research_notes/Money path/m3_owned_products.md.
Where the claim that MCP and ChatGPT stores pay comes from. Verified MCP businesses such as CreativeClaw ($92.9k a month by month 5) charge on their own website. The directories themselves don't pay a Ukrainian seller: ChatGPT apps can't sell digital goods, the Claude directory has no revenue share, and MCP marketplaces run on Stripe.
Week by week. Purple = you (about 1–2 hours a day at the start). Black = AI. Yellow = money lands. Red = we decide.
Revenue per month in three cases. A model built from published prices and our estimated conversion rates. Not a forecast.
2–3 past clients take a retainer in month 1 and referrals convert. The average retainer is $2–2.5k.
One pilot a month from the warm list. The average retainer is $1.5–2k. We lose about 1 client every 6 months.
Past clients say no, so it runs on cold outbound to one lane. It still beats the 20-week media plan on speed.
Why warm first: past clients are 61% likely to refer you (Hinge). Cold email gets about 1 meeting per 6,000 emails (Belkins), so cold is a phase-3 channel, not the start.
The assumptions behind the model. - Prices: pilot $1.5–3k and retainer $1–3k, from the published price points in the menu. - Product: starts in month 4–5 and grows slowly. Verified products in the $5–50k MRR band are typically ~3 years old. - Churn: we lose about 1 client every 6 months in Base. - Not counted: taxes (5% + social contribution on FOP group 3), tools (~$100–300 a month), and the Claude Team seats for client work (~$50 a month for the 2-seat minimum).
Compare with our last plan: first money in weeks 8–10, ~$4.7k a month by month 12 (Base), and a year-1 net of +$2.5k on $26.6k of spending.
What would break this: past clients don't want a retainer, the warm list turns out much smaller than ~200, or you don't send the messages. That last one is the real risk; see 'Our 9 rounds'.
Hard checks on fixed dates. If a gate fails, we change the offer, not the whole plan.
Everything else we checked, and the line that rules it out for you right now.
Fails fast. Median 66–104 days to the first newsletter dollar, and $10k needs 31–95k engaged subscribers. No case without a founder audience got there in 12 months. Keep it as a lead channel for the studio later.
Chatbots, voice receptionists, AI SDRs. The most crowded corner: 5 verified cases stuck under $5k a month, Vapi has 1M developers, and 11x lost 70–80% of its customers.
4.6% reach $10k in 2 years. App launches are up 7×, so apps from 2025 on earn 3% of revenue. AI apps churn 30% faster.
0.45% of Apify developers reach $10k. ChatGPT bans selling digital goods in apps. The Claude directory has no revenue share. Use stores as a channel in step 3, not as the business.
Upwork volume −4% and Fiverr buyers −22%. Ukrainian AI developers on Upwork charge $35–50 an hour. Fine as a bridge; not yours.
Groups over $10k are run by named coaches. The FTC's Growth Cave case shows where selling the dream ends.
Fails digital. Commission is paid at the deed, 4–9 months later, and foreigners can't broker there. Sell those businesses digital AI systems instead.
Fails digital: physical goods, inventory, shipping.
Say "go" and AI starts today. Your first moves this week take about 90 minutes.
Research files: research_notes/Money path/m1–m7 (markdown + JSON). Tags inside: [Primary] = payment-API data, filings, platform data · [Survey] · [Press] · [Estimate].
Limits. This session's 200 web searches ran out mid-research. Later facts were checked by opening primary pages directly. Upwork's and Fiverr's Q3 2026 results come out on 29 Oct and 2 Nov.
Owner-specific figures (odds for you, time to the first invoice, conversion rates, the money ramp) are our estimates built on the published base rates. Nothing here has been checked with a buyer yet; the day-14 gate does that.