Resources
Practical guides for the work after AI.
For people posting work and people doing it. Written to be useful whether or not you use the platform. Original checklists and workflows, not keyword pages. Each guide says how it was made.
How to brief a human expert on AI-generated work
What to include, what to leave out, how to write a definition of done an expert can quote against, and the five mistakes that make good experts skip your job.
How to review AI-generated output before you ship it
Seven checks that catch the "almost right" problems in code, copy, design and media. Experts use it to scope; requesters use it to approve.
AI output vs finished work: understanding the gap
Why the last 30% takes 70% of the time, what "finished" requires in each medium, and how to write a definition of done that an AI draft can be measured against.
How teams organise the work after AI
Ownership, definitions of done, provenance, templates. A lightweight operating model for teams where everyone can generate and nobody is sure who finishes.
How to disclose AI use to clients honestly
A five-part disclosure format, what to record from the start, what Google and the US Copyright Office actually require, and why the disclosure is the best argument for your rate.
By craft
What finishing looks like, medium by medium.
Each one ends with a definition of done you can post as-is.
What to test after AI writes code
AI code passes security tests only 56% of the time, unchanged in two years (Veracode, 2026). The testing order a developer follows before an AI prototype goes to production, with the OWASP checks AI most often misses.
Finishing an AI-generated logo
From a 1024 px PNG to a working identity: direction, redraw, optical weight, small sizes, contrast, variations, rules, files. And why the human work is what makes the logo yours to own.
How to humanize AI writing responsibly
Not detector tricks. Judgement: who is speaking, what only they know, which position they take, which model patterns to cut, and what Google actually rewards.
Finishing an AI-generated song
Selecting takes, structure, lyric, live parts, mix, master, metadata, rights and disclosure. Including what the Copyright Office says about owning AI music.
Making AI-generated UI usable
The states, devices and people the model forgot: real flows, empty and error states, ugly data, responsiveness, WCAG 2.2 AA, components, testing with users.
Coming next
Topics on the list.
Published when there is something genuinely useful to say. RSS feed.
For requesters
- What a good after-AI job costs, by kind of work
- How to choose between two experts' quotes
- Writing a definition of done: worked examples across media
For teams
- Building a preferred-expert bench
- Client review on the record
For experts
- How to quote after-AI work without underpricing the judgement
- Proposing a tighter definition of done at Match
- Turning your before-and-afters into a record clients trust
- Common failure modes of AI output, by medium
By craft
- Finishing AI-generated video: continuity, audio, captions, versions
- Fact-checking AI-generated content: a source-by-source method
Questions the guides do not answer?
Try the FAQ, or write to us with the piece of work you are trying to finish.