After AI Work After AI Work

Trust and quality

Fair to the person paying. Fair to the person working.

A two-sided platform only works if both sides can rely on it. Here is exactly what protects a requester, what protects an expert, and what happens when they disagree. We describe only what is in place today.

Expert vetting

Reviewed by a peer. Proven on a paid trial.

Nobody sees a requester's job because they filled in a form. Every expert passed a portfolio review by an existing expert in the same craft, completed a paid trial piece of real after-AI work against a real definition of done, and verified their identity.

  • Craft-specific. Vetted for logo work, or for backend, or for B2B editing. Not "designer" or "developer" in general.
  • Ongoing. Approval rate, revision rate and dispute outcomes are tracked. Experts who fall below the bar are paused and reviewed.
  • Public record. Jobs finished, first-round approval rate and permitted before-and-afters are visible to requesters before they choose.
A small, skilled pool beats a large, cheap one for this work. "Almost right" output needs someone who can see what is wrong, and that is rare enough to be worth protecting.

Payment protection

Funded before work. Released on approval.

For requesters

Your money is held by the platform's payment provider, not sent to the expert. It releases only when you approve against the definition of done, or when a dispute reviewer decides.

For experts

You only see jobs that are already funded. You never start on a promise. When the requester approves, the funds are already there.

Cancellation

Before an expert is matched: full refund. After match but before work starts: full refund. Once work has started: handled as a dispute, on the record.

Milestones

Large jobs are split into funded milestones. Each is approved and released separately, so neither side carries the whole risk at once.

Scope

The definition of done is the contract.

Every job has a short list of observable criteria agreed before work starts. It protects the requester from an expert doing less, and the expert from a requester asking for more.

  • One revision round included. Changes within the definition, once, at no extra cost.
  • Scope changes are new work. Anything outside the definition is proposed, priced and funded as an addition. Both agree before it starts.
  • Either side can propose a tighter definition at Match. Better to fix vague criteria before money and time are committed.
Vague: "Make the logo look professional and finished."
Observable: Direction B as clean vector. Legible at 16 px. AA contrast on white and dark. Mono and reversed. 4-page rules PDF. SVG, PNG x3, favicon set.

Disputes

Decided by someone who knows the craft, using the record.

Most jobs never get here. When one does, the process is short and factual.

Raise

Either side opens a dispute on the job. Funds stay held. Both write a short statement.

Assign

A reviewer is assigned: a vetted expert in the same craft with no link to either party.

Read

The reviewer reads the definition of done, the AI source, the changes, the comments and the submissions. Not memories.

Decide

Release in full, release in part, or refund. With a written reason both sides can see.

Record

The outcome joins both records. Patterns of bad-faith disputes, on either side, lead to removal.

Reviewers are paid by the platform, not by either party, and are not told who raised the dispute until after they decide.

Conduct

Rules that apply to both sides.

  • Honesty about AI. Requesters say what was generated. Experts say what they changed. The record makes both easy.
  • No off-platform payment. It removes every protection above. Both accounts are closed if it happens.
  • Rights and licences. Requesters must have the right to the AI output they post; experts must have the right to what they add. Neither side uploads material they cannot license.
  • Confidentiality. Job material is private to the job. Experts may show a before-and-after publicly only with the requester's permission.
  • Respect. Feedback is on the work, against the definition. Abuse on either side ends the relationship with the platform.
  • Security. How we handle data is on the security page.

Questions about how you are protected?

The FAQ covers the common ones. For anything else, write to us.