A lot of studios have bought into the idea that generative AI makes game development easier and faster. Krafton, the Korean company behind PlayerUnknown’s Battlegrounds, has gone all-in on it. The reality, though, hasn’t been quite as smooth as the pitch. It reveals that the hidden human and financial costs that aren’t getting as much attention as the hype are quietly piling up.
During a recent panel at Nexon Developers Conference 2026, Krafton VP Lim Kyung-young reflected on the company’s ongoing ‘AI First’ initiative. “Last November, we announced that we would operate everything at Krafton under an ‘AI First’ philosophy,” said Lim via Korean media Gamemeca through Google Translate. “In a survey conducted this February, 97.6% of employees said they were using AI.”
That’s definitely an impressive adoption rate, something that’s probably getting closer to becoming the standard. Not just in the gaming industry, but in tech in general. However, he also admitted that its AI push hasn’t been without growing pains. According to the discussion, one of the biggest challenges has been employee burnout.
Lim explained that as AI tools and models evolve at a breakneck pace, developers are constantly expected to learn and adapt to new tech. The irony here, of course, is that keeping up with the latest AI developments can quickly become exhausting instead of saving time.
The PUBG company isn’t alone in this issue, however. Kang Deok-won, the AI lead at Nexon, added that his company encountered the same problem. While Krafton pushed AI from the ‘top down’ through executive mandates, The First Descendant and MapleStory publisher applied a ‘bottom-up’ process. It encourages development teams to experiment with AI tools on their own. Yet in the end, both companies arrived at the same conclusion: employee fatigue.
Then there’s also the matter of money. Aside from speeding up asset creation, generative AI is often marketed as a cost-saving solution. But Krafton and Nexon acknowledged that AI usage ‘token-based’ costs continue to climb as more employees rely on large language models in their daily work.
To mitigate these problems, the companies are working on cost prediction models to see if their AI spending is driving productivity. Krafton is also building an LLM price comparison tool and is developing its own AI model.
Ultimately, repeating a point previously made by Guild Wars publisher NCSoft, both companies argued that developers shouldn’t be competing with AI. They should be focusing on the things generative AI still can’t do particularly well instead. That is focusing on creativity and emotional core to answer the fundamental question of, “Why is this game fun?”
