There is a version of this story that reads as straightforward industry news: a mid-sized Czech studio, looking to cut costs between projects, replaces a contract role with automated tooling. It happens. The games industry has been shedding headcount since 2023, and AI integration is accelerating. Move on.
But the Warhorse Studios case is not straightforward, and the reason it isn't tells you something important about what localisation actually is, and what happens when the people running the money side of a studio don't understand the answer to that question.
"Yesterday, March 27th 2026, with no forewarning, I was invited to a meeting and promptly told that, in an effort to make the company more effective and save finances, as of next month, my position at the company would be made obsolete and that they are switching to using AI for all translations going forward."
Max Hejtmánek, LinkedIn, 28 March 2026
What Was Actually Cut
Kingdom Come: Deliverance 2 was one of the most acclaimed RPGs of 2024-2025. Its nominations for Game of the Year were driven substantially by writing quality, narrative authenticity, and the coherence of its dialogue. The English-language version of that experience was Max's work. Four years of it.
This is not a story about replacing a generic translation vendor with a cheaper one. This is a studio firing the person who built the linguistic identity of a GOTY-nominated title, between projects, on the basis that AI can now do the equivalent job. That assumption is worth examining carefully.
The Cost Calculation Studios Get Wrong
The framing of AI replacing a translator as a "cost saving" treats localisation as a production cost rather than a quality determinant. The calculation: translator costs X per year; AI costs Y per year; Y is less than X; save X minus Y. This arithmetic is accurate on a spreadsheet and wrong about what it measures.
What it doesn't measure: the cost of a GOTY-nominated title shipping with mechanical dialogue instead of the writing that earned those nominations. The cost of community comparison threads that will run the moment the next title ships, side-by-side screenshots of KCD2 prose versus AI-assisted output. Studios that built reputations on writing quality and then systematically removed the people who produced it are not saving money. They are spending reputation capital they don't know they have until they've spent it.
The AI Localisation Claim
AI tools perform well on structured, high-volume, low-register content: UI strings, menu text, patch notes, system messages. They perform poorly on content where register, cultural authenticity, and tonal consistency are the point. Czech-to-English literary translation of dialogue for a historically-grounded RPG sits entirely in the second category. The outputs are measurably different, and audiences who bought the first game because of its writing quality will notice.
The category error
Applying AI to all translations without distinguishing between content types is the localisation equivalent of deciding that because a calculator can do arithmetic, it can also write your annual report. The tool is not wrong. The application is.
What This Signals to the Industry
The studios making this calculation without understanding its product implications will pay for the error in the market. The ones that understand it will build the distinction into their production models: automated workflows for structured content, human expertise for content where quality is the product. Not because this is a more principled position, but because it is the correct cost-benefit analysis when you account for what the quality of your writing is actually worth to your commercial outcomes.
Related Insights