Once you can make something with AI, the next question is always the same: who buys it?
The usual answer is a marketplace, because that is where people who already decided to spend go to search. You do not have to create the market. That is the whole appeal of this route.
What follows is built on three listings we opened and read on 2026-08-15, on Kmong — Korea's largest freelance marketplace. The platform is Korean; the structure it enforces is not.
Demand grew. So did supply
Kmong created a separate AI services category in June 2025 and says cumulative orders passed 12,000 within nine months, up from 2,200 at the two-month mark, with average satisfaction of 4.9 out of 5.
The same announcement contains the other half. The number of AI sellers more than doubled and listed AI services passed 3,800. The fastest-growing area was "AI model" — generating model imagery without a photo shoot.
Step 1. Demand
12,000cumulative orders in 9 months (company figure)
Money already moves here. You do not have to build the market.
Step 2. Supply
3,800+listed AI services (company figure)
Sellers grew over the same period. The later you arrive, the more you are compared against.
Step 3. Biggest pull
AI imagerymost-searched keyword (company figure)
Replacing photo shoots. One of the listings below sits exactly there.
⚠️ These are company figures relayed by the press, not independently audited. Use them for direction, not as grounds for "so it will sell for me too." Source: Platum, 2026-04-14.
The three listings we checked
Below is what the platform displays, copied as-is. These are numbers the platform computes, not marketing copy the seller wrote.
| Chatbot builds | AI product imagery | Model development | |
|---|---|---|---|
| Transactions | 18 | 23 | 81 |
| Satisfaction | 96% | 99% | 100% |
| Entry price | ₩150,000 | ₩150,000 | ₩800,000 |
| Top price | ₩700,000 | ₩1,350,000 | — (on request) |
| Turnaround | 3–7 days | 3–10 days | 14 days |
| Account type | Individual | Corporate | Corporate |
| People | 1 | not shown | 2 |
The full analysis of each is in our case library, with a verification grade, the date we checked, and a link to the original listing.
What matters here is not "how much they earned." No revenue figure is published anywhere, and we did not manufacture one by multiplying. What matters is the price band. Inside one category, ₩150,000 and ₩800,000 coexist, and the gap comes from the scope sold and the track record required — not from being better at using AI.
What all three do the same way: they sell scope, not outcomes
Every one of them splits the price list like this.
Step 1. Scope
What is included, written so it can be counted — “up to three documents connected.”
Step 2. Turnaround
Three days, five days, seven days. The duration becomes the justification for the price.
Step 3. Revisions
Two, three, five. Leave this out and revisions never end.
Sell an outcome and the deal becomes “until you are happy.” Sell scope, duration and a revision count, and what you sold is a bounded amount of labour — which is the first thing you can actually compute an hourly rate against.
✓ Write it like this
- Up to 3 internal documents · 5 working days · 3 revisions
- 2 AI scene shots + 1 thumbnail
- Anything beyond this is quoted separately
✕ Not like this
- Unlimited revisions until you are satisfied
- We build you the perfect chatbot
- Timeline agreed after consultation
Four things to settle before you list
- 1
Narrow to one category
“We do all things AI” does not surface in search. Each of the three listings we checked does exactly one thing.
- 2
Build only the smallest package
Cut until you could deliver inside three days. Better than building something large you never ship.
- 3
Make three portfolio pieces with no client
A new listing has no track record; three self-made pieces stand in for it. One listing we checked had twelve.
- 4
Decide your response window
Profiles display things like “replies within 1 hour.” Put down a window you can actually hold.
What the platform puts on screen
Profiles show total transactions, satisfaction, average response time and account type. That is the platform telling you which signals it believes buyers choose on.
⚠️ How ranking is decided is not published. All we know is that these values are displayed. We have not verified any claim that faster replies move a listing up.
Check these before you sell (Korea)
| Item | As an individual | As a registered business |
|---|---|---|
| Signing up | Registers under your own name. One of the three listings is an individual account | Business registration required |
| Payment | Runs through the platform's escrow and settlement | Same |
| Tax invoice | May show as unavailable — one listing we checked does | Available. Effectively a requirement for corporate buyers |
| Income reporting | Yours | Yours |
| Minimum | Identity verification | Business registration certificate |
This table is not tax or legal advice and does not replace it. It is a list of questions to take to an accountant, not answers. Where filing depends on income level, we deliberately left the answer out — the cost of being wrong lands on you.
⚠️ We did not print a seller commission rate. We verified only what the listing pages display, and how much actually settles to the seller is not something we confirmed. Read the platform's own settlement terms before listing.
Readers outside Korea: the platform names differ, but the four decisions above and the scope-turnaround-revisions structure carry over. What does not carry over is the tax and invoicing column — check your own jurisdiction.
Do this today
Step 1. 1. Copy ten price lists
From the category you would sell in. Entry price, top price, turnaround, revision count, in one table.
Step 2. 2. Find the common scope
Write what all ten include, and separately what none of them offer.
Step 3. 3. Pick one three-day scope
From that table, the smallest thing you could deliver within three days.
Step 4. 4. Build one portfolio piece today
With no client. Repeat until you have three, then list.
Before setting a price, subtract tools, fees and tax — the break-even calculator and tool cost comparator do that arithmetic for you.
What we do not know
Stated plainly.
- How long it takes to get a first order. One listing we checked shows 18 transactions, but that listing itself is new with zero reviews — the 18 belong to the seller's other listings
- What any of them earn. The platform does not display revenue. Multiplying transactions by price is tempting and would be our estimate, not a fact
- How ranking works. It is not published
- How fast competition is growing. The 3,800 figure is from the company's April 2026 announcement and will be different now
You can act without knowing any of those four. Copying ten price lists takes thirty minutes today, and that table decides most of the rest for you.