What the work actually looks like

You start from a defined user goal — chasing a late payment, drafting a 見積書, reconciling a supplier spreadsheet, planning a small local campaign — and turn it into a prompt a real owner would actually type. You then run that prompt through several AI chatbots, capped at five turns per conversation, and compare what comes back. The output is a structured evaluation: which model was most useful, where each one was factually wrong or culturally off, and why. Transcripts and scoring go back with the assessment, so your reasoning has to be legible to someone who wasn't in the conversation.

Many tasks are multimodal. You attach your own Japanese business documents — spreadsheets, PDFs, images of receipts or product photos — and judge how well the models read, summarise and act on them. This is why the document requirement is a hard gate rather than a nice-to-have: without real source material in Japanese, you can't run the task.

What the screen is looking for

Turing's screening is less interested in whether you like AI than in whether you can be specific about small business operations. Expect probing on the mechanics of your business: how you invoice, what your cash cycle looks like, what a bad supplier email costs you. Expect equal probing on evaluation discipline — whether you can separate a response that sounds polished from one that is actually correct, and whether you can apply a rubric consistently across five conversations rather than drifting toward whichever chatbot flatters you.

  • Native or business-level Japanese, including keigo conventions in written correspondence
  • Ownership or hands-on operational experience in a small business
  • Willingness to work from a written rubric, not personal taste

Logistics

Fully remote, asynchronous, project-based with a defined task count and a stated duration of roughly 16 weeks. Pay is undisclosed on this listing; Turing typically quotes a per-task or hourly rate at the offer stage, and rates observed on comparable Turing generalist evaluation work vary widely by language and project. Volume is not guaranteed week to week — treat it as a supplementary engagement alongside running your business, not a replacement for it.