Turing is recruiting Korean-speaking small business owners to write realistic business prompts, run them against several AI chatbots, and judge which responses would actually hold up in day-to-day operations. The work is project-based over roughly 10 weeks and requires you to supply your own Korean-language business documents — invoices, spreadsheets, contracts, PDFs — as task inputs.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

Each task starts with a defined user goal — a shop owner reconciling a month of receipts, a café operator drafting a supplier complaint, a consultancy pricing a proposal. You write a realistic prompt in Korean that a real owner would actually type, often attaching a document: a spreadsheet, a scanned PDF invoice, a photo of a handwritten ledger. You then run that prompt against multiple chatbots, holding each conversation to a maximum of five turns, and evaluate the responses side by side for clarity, usefulness, and factual accuracy. The deliverable is the transcript plus a structured comparative assessment explaining which model helped and where the others failed.

The volume is real work, not opinion-gathering. Expect to spend as much time on document preparation and writing up the rationale as on the conversations themselves.

What the platform screens for

That you actually run or have run a small business. Screeners probe for operational specifics — how you handle 부가가치세 filing, what your supplier payment terms look like, how you price. Vague managerial language gets caught fast.
That you have Korean-language business documents you can use. This is a hard requirement. Tasks depend on real input files, and you are responsible for redacting anything sensitive before uploading.
That you can evaluate rather than react. The difference between "I liked this answer" and "this answer cited a filing deadline that is wrong for a 간이과세자, which would cost the user a penalty" is the whole job.
That you follow a rubric. Evaluation guidelines are fixed; personal preference about tone is not a scoring dimension unless the rubric says so.

Logistics

Fully remote and asynchronous, with a stated project duration of about 10 weeks and a defined number of evaluation tasks rather than fixed hours. Pay is undisclosed on this listing — expect rate and task-volume details to surface during the screening conversation, and ask directly. You will need a reliable machine, spreadsheet software, and the ability to handle PDFs and images as task attachments.

Small business owners (AI response evaluation) - Korean Business Document

What the work involves

What the platform screens for

Logistics

Interview prep — 10 questions