Many large Japanese companies with translation departments, including mine, have been evaluating machine translation and AI assistants to improve quality, increase speed, and reduce labor costsโparticularly those associated with employing people like me. So, with my head potentially on the chopping block, you might expect me to explain why inโhouse translators, especially native English-speaking translators, are essential to a companyโs success.
But I canโt. My AI assistant won’t let me. Instead, I am here to write my own obituary on behalf of myself and other native English-speaking translators.
Since maritime traders began using English as their common language, international trade has endured and even prospered under business correspondence that was as cryptic as a crafty pirateโs treasure map. It will certainly survive a few AI indiscretions, with or without a native English translator there to save the day.
What English translators have done by translating understated Japanese into clearly stated English and spending hours tightening prose, untangling modifiers, and hunting down wandering antecedents is to add another layer of trust for a colleague, product, service, or company. Gone are the days when consumers in English-speaking countries threw their operation manuals down in disgust.
Nevertheless, the exodus of teary-eyed translators walking out the companyโs door with a cardboard box, flowers, and a farewell card signed by their colleagues seems inevitable.
โSayลnara, Mr. Native English Translator!โ the in-house Japanese translators will say while secretly gleeful that their nemesis with the red pen is leaving, unaware that they too may just be whistling past the graveyard.
That said, before you begin stacking logs onto my funeral pyre and publishing my obituary, I have some caveats.
The Caveats
Itโs certainly true that machine translation and AI assistants provide access to an extraordinary amount of knowledge about English composition and grammar through a browser window, allowing anyone with a keyboard and an internet connection to write like a budding Ernest Hemingway.
The process is lightning fast, and when applied appropriately, these translation tools help Japanese translators instantly correct awkward expressions, confusing prepositions, and pesky articles. Likewise, English translators, who had once spent years wrestling with the complexities of kanji and the nuances of Japanese, now have the ability to paste a document into an AI assistant and receive a translation that is often remarkably good.
But using digital tools to improve translation is nothing new. Translators have been quietly using their predecessors, a.k.a. software translation applications and online translation sites, for years while feigning ignorance of their existence.
Having begun translating professionally in an era when English translators had to prove their fitness with a translation test, a couple of sheets of paper, a Japanese-English dictionary, two sharpened HB pencils, and an eraser, I am of two minds. One, itโs not fair, damn it!! Two, these tools give all translators the opportunity to better serve their customers.
The Risks of AI-Assisted Translation
When we consider entrusting our lives and livelihoods to AI, we have to consider its risks and reliability. Ninety-five percent reliability might seem acceptable until you are talking about driverless cars and fatalities. Then you might start thinking that the crazy driver in the next lane is actually Mario Andretti.
Working with AI-generated prose requires translators to learn a different set of pitfalls from those associated with traditional machine translation. While machine translation often risks saying too little, AI assistants increasingly risk saying too much. What makes AI-assisted errors particularly troublesome is that they are often veiled in fluent, compelling English:
- Presenting inferences as established facts
- Prioritizing fluency over fidelity
- Overlooking or misinterpreting specialized terminology
- Adding specificity not present in the source text
- Normalizing writing to a common style
- Softening assertions it perceives as controversial
- Projecting the image of the ultimate authority
Often readers will forgive the occasional grammatical error or misspelling, but when the subject becomes confused with the object and facts are replaced with suppositions, they start to distrust what they are reading. Consequently, that layer of trust surrounding a colleague, product, service, or company and earned over decades can instantly erode.
Like autonomous driving, AI-assisted translation works remarkably well right up until it makes a wrong turn into a lake, turning autonomous into amphibious. Thatโs why a competent driver needs to be at the wheel, ready to take over when necessary.
A New Skill Set for Translators
One lesson from the rapid adoption of AI is that translators now need training not only in language but also in how to use these tools effectively. Knowing when to trust a translation, when to challenge it, and when to verify terminology has become as important as producing the translation itself.
For that reason, I believe both Japanese and English translators should continue translating documents themselves, even if only as a first draft. AI can then serve as a valuable second opinion, offering alternative interpretations, suggesting improvements, and identifying possible errors. This means keeping the translator in the driverโs seat, responsible for the final decisions, and the AI assistant as his trusted navigator.
Once the translator is confident that the meaning has been accurately captured, AI can perform one of the tasks at which it excels: proofreading. At that stage, suggestions regarding style, clarity, grammar, punctuation, and readability can be evaluated without risking the loss of meaning during the original translation process.
In other words, the translator determined the destination while the digital tools help determine the best route to get there.
A Case Study: The Anniversary Book
Last year, I was asked to proofread the English translation of our parent company’s anniversary book which included interviews with 96 employees across the group. The interviews were conducted by three members of HR who relied heavily on machine translation and, at times, AI-assisted translation and editing.
At first glance, the translation appeared professionally done, and I expected the assignment to be straightforward. However, once I began reading for meaning rather than grammar and punctuation, I noticed a pattern of excessive intensifiers and unnecessarily flowery language that transformed genuine accounts of overcoming challenges and personal growth into something resembling a Marvel Avengers origin story. The over-the-top description became a distraction, causing me to lose trust in the content.
Needless to say, the interviews were painful to read. There was also a hotchpotch of rookie mistakes, including changing groupwide to groupwise in a title. More troubling than the occasional awkward sentence or sudden shifts in point of view was the paraphrasing and omission of key corporate terms and phrases used consistently throughout the company. The result was English that often sounded more glamorous and grandiose than the original Japanese while bearing only a passing resemblance to the values and lessons the employees were trying to convey.
In the past, I would have been involved in either the translation or editing process. This time, buoyed by the hype surrounding AI and their own experience using it to produce polished emails and reports, the employees never questioned what was being presented to them. What was needed was the judgment of experienced translators accustomed to questioning both themselves and their colleagues.
Thatโs not vanity. Thatโs experience.
Authority vs. Correctness
Machine translation users must ensure that proper terminology is used and that key corporate terms, phrases, and concepts are not omitted. AI assistants require a different kind of vigilance. Someone must still be willing to challenge the machine when it confidently asserts, โThis is the more natural expression,โ โNative speakers would normally say โฆ,โ or โThis wording is correct because โฆโ
What I have noticed is that the translator may not always possess the linguistic confidence to challenge those assertions. In this case, the AIโs conjecture often becomes fact; its preference becomes a rule, and its style becomes the standard.
As a native English-speaking translator, I can evaluate an AIโs suggestions and push back when I disagree. Many Japanese translators do not enjoy that luxury. Faced with a confident explanation of why one expression is preferable to another, they may understandably treat the AI as an authority. And if they donโt, they risk that their clients might.
The problem is that authority and correctness are not always the same thing, and when repeated often enough, an AIโs conjectures can harden into accepted facts.
What AI Still Canโt Do
What the anniversary book taught me was that the problem was not the English. Much of it was grammatically correct and stylistically polished. The problem was that the translation lacked the institutional knowledge possessed by people working within the company. AI could generate attractive prose, but it did not know which terms carried significance, which expressions reflected corporate values, or which wording had been intentionally standardized over time. Those are things learned through experience, conversation, and physical presence within an organization.
A translator sitting inside a company does more than translate documents. He attends meetings, asks questions, observes how people communicate, learns the history behind recurring terminology, and develops an understanding of the companyโs culture and priorities. Those insights rarely appear in source documents, yet they often determine whether a translation is authentic or merely sounds authentic.
The technology available today is phenomenal, but it is most valuable when used as a point of comparison against a translatorโs own work and as a resource for checking facts, terminology, and expressions. There is a reason virtually every AI provider includes a disclaimer reminding users that mistakes can occur.
Reports of Death Greatly Exaggerated
In retrospect, it seems that the timing of my obituary may have been greatly exaggerated. In-house translators, including English translators, bring something that machines cannot: physical presence within the company, an understanding of its history and culture, professional judgment, and intuition developed through experience. Those qualities ultimately reduce both translation costs and legal risk.
AI may not be helping translators write their obituaries after all, but simply revealing that a translator’s greatest responsibility is, and has always been, not to translate words, but to preserve meaning.





































































































