After AI Work After AI Work

How it works

From AI output to finished work, with a person in between.

Five stages. The requester owns the outcome and the money. The expert owns the craft and the changes. The platform holds the payment, the record and the rules that keep both sides safe.

The flow

Who does what, in order.

Requester

1. Post

Upload or link what AI produced: files, code, text, audio, video, or a description. Write what finished means in a few observable lines. Set a budget, or ask for quotes. Fund the job; the money is held, not paid.

Example: "4 generated logo directions. Done = one direction redrawn as vector, works at 16 px, passes contrast, dark version, brand rules PDF." Budget: fixed.
Both

2. Match

Vetted experts in that craft see the job. Each can accept the budget or send a short quote with a plan. You see their record: jobs finished, approval rate, before-and-after examples. Pick one, or accept our suggested match.

Most jobs are matched within a day. You can ask questions before choosing. Nothing starts until you choose.
Expert

3. Work

The expert refines, develops, edits, tests, and posts progress on the job. Every change sits next to the AI source it acts on, so you can see exactly what a person did to what the model produced.

You can check in, answer questions and see interim work. You cannot change the definition of done mid-job without both agreeing; that becomes new scope.
Requester

4. Approve

The expert submits. You review against the definition of done, line by line. Request changes (one round included) or approve. Approval releases payment to the expert.

If you and the expert disagree about whether the definition is met, either side can ask for a review by a human in that craft. The reviewer decides on the record.
Both

5. Ship

Final files are delivered. The job closes with its full record: the AI source, the prompt if you shared it, every change, every comment and decision, the approval, and a ready-made disclosure of how AI and a person produced the result.

The record is yours to keep, export and show a client. The expert can add the job to their public record with your permission.

The definition of done

One short paragraph does most of the work.

Every job starts with a definition of done: five to eight observable lines describing finished. It is what the expert quotes against, works toward and submits against, and what you approve against. It is the contract and the quality gate in one.

  • Observable: "passes WCAG AA" not "looks accessible"; "all nine statistics sourced" not "facts right".
  • Contextual: names the real platform, brand, codebase or audience.
  • Short: if it needs more than eight lines, it is two jobs.

Not sure how to write one? Each kind of work has a starter you can edit, and the expert can propose one at Match.

Done means: direction B redrawn as a clean vector. Legible at 16 px as a favicon. Passes AA contrast on white and on #0C1410. Mono and reversed versions. Type and spacing rules in a 4-page PDF. Files: SVG, PNG at 3 sizes, favicon set.
This is a real-shaped definition for a logo job. Note that every line can be checked by someone who is not a designer.

Money

Paid on approval. Funded before work starts.

The requester funds the job at Post. The expert sees a funded job before accepting. Payment releases at Approve. Nobody works unpaid, nobody pays for work they have not approved.

Fixed price or quote

Set a budget when you post, or ask experts to quote. Most after-AI jobs are well-scoped enough for a fixed price.

Milestones for bigger work

Larger jobs (an app to production, a full brand system) can be split into funded milestones, each approved and paid separately.

One service fee, on completion

Free to post, free to join. A service fee is charged only on completed, paid work. See pricing.

The record

Not just the file. The path to it.

Every job keeps what AI produced, what the expert changed, who said what and when, and the approval. That record is the thing most other ways of getting AI work finished throw away, and it is the thing you will need later.

  • For clients and platforms: an honest, one-click answer to "how was AI used in this?"
  • For disputes: a reviewer reads the record and the definition of done, not two memories.
  • For experts: proof of judgement. The before-and-after is the portfolio.
Job record: landing page copy
Shipped, 24 Sep
Origin: 2 AI drafts (requester supplied), 1 promptSource
Anita N.: 5 changes, 9 statistics sourced, intro rewritten, 2 questions answeredExpert work
Requester: 1 revision request ("shorter intro"), resolvedRevision
Approved by requester. Payment released. Disclosure generated.Approved

AI creates possibilities. A person turns them into finished outcomes.

Post what AI made and get it finished by someone who knows the craft, or join as that someone.