DATA WORLD read 326 Japanese FDE-style job ads: 70% want one engineer from discovery to adoption, only 3% require industry knowledge

Japanese data publication DATA WORLD read the full text of 326 job ads at 168 companies that were live on October 17, 2026, all describing the work of taking AI into a customer's or the company's own operations. Only 154 of them use the FDE title. Another 108 are IT firms advertising the same job as "AI engineer" or "AI transformation engineer", and 64 come from non-tech companies hiring for themselves.

The headline finding is scope. 228 ads (70%) give a single engineer requirements gathering, PoC-to-production build, and operations and adoption support, and only 5% stop at a proof of concept. Yet just 10 ads (3%) make industry knowledge mandatory, and only 37% require LLM or generative AI experience. The most common hard requirements are hands-on development, such as Python (70%), and customer-facing requirements work (46%). About 60% of the roles build AI agents or workflow automation for paper- and Excel-heavy work such as drawings and quotes, dispatch, and approvals.

At least 41 manufacturers, banks and utilities posted 64 such roles, led by Aisin (6), Honda and Mizuho Financial Group (4 each), with Toyota, Denso, SMBC, Resona, Tokyo Gas, JR West and Kirin also named. Those in-house ads put much more weight on keeping systems in use (94% mention operations and adoption support). Among the 168 ads that list pay, 80% have a ceiling of ¥10M or more, 16 reach ¥20M, and the top is ¥30M. Only 23% ask the engineer to measure ROI or KPIs.

Why it matters for FDEs: in Japan, many FDE jobs don't carry the FDE title, so searching by title misses a big share of the market. The ads also suggest domain knowledge is something employers expect you to learn on the job, while owning a deployment end to end and being able to prove its impact in numbers are what set candidates apart.