OPINION – The video game industry is simultaneously struggling with runaway development costs, increasingly long production cycles, studio closures and a wave of layoffs that has lasted for years. Into that crisis comes generative artificial intelligence with the promise of faster testing, coding assistance, cheaper localization, smarter NPCs and, perhaps most importantly, a way to cut some of the increasingly absurd cost of AAA development. The problem is that the same sentence that means “efficiency gains” in an executive presentation can sound suspiciously like “the same work with fewer people” to a developer. AI really could help save the games industry, but first we may have to decide exactly who inside that industry is supposed to be saved.
Artificial intelligence, of course, is hardly a new guest in video games. Enemy pathfinding, procedural generation, animation systems, matchmaking, cheat detection and machine-learning-based testing existed long before “generative AI” began appearing on every other corporate slide. The current argument therefore is not about whether games should be allowed to use algorithms; that question was answered decades ago. The real change is that generative and increasingly agentic AI is moving into workflows previously handled by writers, artists, programmers, QA testers, localization specialists and producers.
And the timing could hardly be more sensitive. The GDC 2026 State of the Game Industry survey, based on responses from more than 2,300 industry professionals, found that 28 percent had personally experienced a layoff during the previous two years, rising to 33 percent among respondents in the United States. Seventeen percent had been laid off in the previous twelve months alone, and 48 percent of those who had lost their jobs had still not found another one when surveyed. This is the environment in which developers are now being told that a new technology has arrived to “free their creativity.” It is understandable if not everyone reaches for the champagne.
The issue is also too easily reduced to two convenient extremes. One says AI is the industry’s savior, finally capable of breaking the insanity of $200 million budgets and six-year production cycles. The other treats every AI tool as a digital scythe whose sole purpose is to remove as many names as possible from a payroll spreadsheet. Reality is more uncomfortable than either version: AI genuinely can automate enormous amounts of unnecessary work, and precisely because it can do that, it can also be used by companies to attempt the same output with fewer people. The technology makes neither decision by itself. Management does.
The Math Is So Tempting That Finance Departments Will Struggle to Resist
If you look only at the numbers, enthusiasm is difficult to avoid. An international Google Cloud and Harris Poll study involving 615 game developers found that 95 percent believed AI was reducing repetitive tasks in their workflows, 47 percent reported faster playtesting and mechanics balancing, 45 percent cited localization and translation, and 44 percent pointed to coding and scripting support. More important still, 94 percent expected AI to reduce overall development costs in the long term.
That is almost exactly the medicine the games industry would order for itself. Modern AAA development has become so expensive that a single failure can wipe out years of work and sometimes an entire studio. If AI can accelerate prototyping, automatically run thousands of tests, help identify bugs, prepare first-pass localization or remove monotonous work from programmers, its practical value is easy to see. In many of these jobs, the best AI may be the AI the player never even realizes was involved.
Electronic Arts CEO Andrew Wilson made exactly that case in April. According to his comments, some form of machine-learning or AI-driven algorithm is now involved in roughly 85 percent of EA’s quality-assurance work, while Wilson claims the company employs more QA staff than ever. Asked whether AI was replacing workers, he said: “So far, it’s been almost entirely augmentation.” It is worth noting that much of the QA automation Wilson described involves conventional machine learning rather than necessarily generative AI, but the broader point remains valid: there is little creative virtue in paying skilled humans to repeat mechanical tasks when a system can handle them faster.
For smaller studios, these tools could become genuinely democratizing. A twenty-person developer does not have a dedicated localization department, one hundred QA testers, a research team and fifty technical artists. If that same studio can use AI to prototype faster, create temporary assets, process documentation or prepare a game for multiple languages, projects that previously would have been financially impossible may suddenly become viable. Viewed from that angle, AI is not the creative worker’s enemy. It is additional labor an indie team could never have afforded to hire.
AI Is Most Useful in the Places Nobody Wants to Put in a Trailer
The most valuable applications may not involve a machine drawing an image or producing dialogue at all. Game development is full of time-consuming work that demands human oversight but offers little creative satisfaction: build verification, compatibility testing, log analysis, bug reproduction, animation-data cleanup, searching enormous internal documentation or running thousands of variations. If AI removes some of that burden, it is difficult to argue that the game has somehow become less artistic because a senior designer no longer spent three hours on Thursday afternoon reorganizing a spreadsheet.
There are more ambitious experiments as well. After Ubisoft’s NEO NPC project in 2024, the company revealed the playable Teammates prototype in 2025, where generative-AI-enhanced companions can understand spoken player commands in real time, react to their surroundings and operate beyond a simple library of prerecorded dialogue choices. This is not primarily about eliminating three concept artists. It is about attempting a form of interaction that would be practically impossible to script manually for every possible player action.
Virginie Mosser, narrative director on the earlier NEO NPC experiment, told Ubisoft: “For the first time in my life, I can have a conversation with a character I’ve created.” This is a far more interesting face of AI. Rather than replacing a writer with a machine that generates a hundred mediocre lines, it gives a personality created by a writer a degree of responsiveness that could never realistically be authored for every possible player behavior. Of course, that still requires human work, and probably a great deal of it, to stop the character from talking complete nonsense five minutes later.
That is exactly why the simple “AI versus humans” framing is misleading. Good implementation often means a human defines the objective, rules, personality, style and quality threshold while AI expands the space within which the system can respond. Bad implementation begins when somebody notices that if a machine can produce something ten times faster, perhaps nine people are no longer required. Same technology, completely different corporate philosophy.
Developers Aren’t Technophobes When Desks Are Actually Disappearing Around Them
Executive enthusiasm about AI is impossible to separate from what has been happening to employment. In GDC’s 2026 survey, 52 percent of industry professionals said generative AI was having a negative impact on the games industry, while just 7 percent regarded its impact as positive. Only two years earlier, the negative figure had been 18 percent. The groups most critical of the technology are also those closest to making the actual games: 64 percent of visual and technical artists, 63 percent of game design and narrative workers, and 59 percent of programmers viewed generative AI negatively.
Interestingly, Google’s research paints a substantially more optimistic picture. That does not necessarily mean one survey is “lying”: they use different samples, questions and methodologies. Instead, it illustrates how radically different the same technology can look to someone considering how it might accelerate a specific workflow and to someone wondering whether they will still have a job in two years. AI can simultaneously be a useful tool and an employment threat. Unfortunately, those ideas are not mutually exclusive.
The distinction is particularly visible at Xbox. In July, Asha Sharma announced the most significant restructuring in Xbox history, involving approximately 3,200 job reductions by the end of fiscal 2027, with 1,600 positions identified in the initial wave. Sharma explained that in a typical year the company had lost 64 cents for every dollar invested in certain parts of the business. She did not identify AI as the cause of those layoffs, so claiming that those 3,200 people were simply “replaced by AI” would be factually wrong.
Yet Microsoft gaming workers are worried about exactly that possibility. An August survey conducted by the CWA with researchers from Cornell University and Western University found 40 percent of respondents extremely concerned and another 20 percent moderately concerned that AI could replace some or all of their work. More than 54 percent of Microsoft-studio respondents believed automation- or outsourcing-related layoffs were somewhat or very likely within two years. Researcher Johanna Weststar summarized the findings bluntly: “Fears and frustrations about AI are real and warranted.”
More awkward still, 59 percent of respondents who actually use generative AI said it creates additional work because they have to correct its mistakes, 50 percent said it makes their work more stressful, and 56 percent disagreed that it improves game quality. None of that proves every AI tool is useless. It does, however, demonstrate the potential gulf between “30 percent productivity gains” on a presentation slide and the developer still sitting at a desk at 7 p.m. correcting something a machine confidently invented.
A Saved Labor Hour Can Become a Better Game, or a Deleted Job
This is the heart of the entire argument. Imagine a new tool genuinely makes a team’s work 30 percent faster. Several futures are possible. The studio can keep the same team and devote 30 percent more time to polish, experimentation and ideas it previously lacked capacity to pursue. It can make the same game more cheaply and reduce the chance that moderate sales result in a studio closure. Or it can remove part of the workforce and tell everyone who remains that the new tools mean the original deadline should still be met by a smaller team.
The technology has not changed in any of those scenarios. The only difference is who receives the productivity dividend. The developer in the form of more creative time? The player through a better and less buggy game? A small studio through the ability to attempt something more ambitious? Or exclusively the company’s operating margin? Until there is a convincing answer, the phrase “AI frees people from tedious work” will inevitably sound suspicious to workers who have already watched several previous “efficiency transformations” end in familiar ways.
Take-Two CEO Strauss Zelnick is interesting because despite being strongly pro-technology, he is considerably less convinced of AI’s omnipotence. In a March interview with The Game Business, he argued that better and faster creation tools are clearly beneficial, but creating a hit on the scale of the industry’s biggest games still requires “human engagement and creativity.” AI can help generate assets, storyboard ideas or explore alternatives, he said, but none of that automatically produces another Grand Theft Auto. He described the idea that somebody could press a button and generate a global hit as laughable.
That matters because the industry’s real problem is not that it cannot produce enough assets. More games already release than audiences have time to play. The deeper problems are rising costs, poor project management, constant scope expansion, trend chasing, excessive monetization, executive mistakes and a financial model in which an AAA game often is no longer permitted to be merely a respectable success. If AI is used only to pump even more content into the same oversized machine, it does not solve the problem. It simply manufactures the same problem more cheaply.
AI Could Save the Games Industry Without Necessarily Saving Its Developers
That is why I do not find categorical rejection of AI particularly useful. If a system identifies the cause of a rare crash in an afternoon instead of a week, automates mind-numbing compatibility tests, allows a fifty-person studio to release its game in ten languages, or finally gives an NPC more than four lines to repeat for fifteen years, it would be difficult to argue that throwing the technology away is some kind of moral obligation. The games industry has always built new possibilities out of technology.
It would be equally naive, however, to believe companies will automatically convert every saved labor hour into creativity. Developers have spent the past few years watching record revenues coexist with layoffs, successful releases followed by studio closures, and enormous acquisitions followed by cost-cutting programs. After that, it is hardly surprising if workers are interested not only in what AI can do, but in what the finance department intends to do with it.
The right question may therefore not be whether AI will take developers’ jobs. Some workflows will certainly disappear or be fundamentally transformed, just as earlier technologies changed other professions. The more important question is whether the resources released by those changes will build a healthier and more sustainable games industry, or whether the same system will continue operating with smaller teams, more output and tighter deadlines. In the first version, AI genuinely could be a lifeline. In the second, it is merely a very sophisticated way to ensure the next layoff email contains the word “transformation.”
The biggest danger is not that AI will write the next Grand Theft Auto tomorrow. Strauss Zelnick does not seem particularly worried about that, and neither am I. A far more realistic danger is an executive deciding that because parts of ten people’s jobs can be automated, five people should now be enough to do all of it. Two years later, everyone wonders why the game feels more generic, ships with more problems and was made by an exhausted team. Creativity is difficult to place on a KPI dashboard, but players notice surprisingly quickly when it disappears.
So yes: AI might actually save the games industry. It can make development cheaper, help control the cost spiral, give smaller teams tools they could never previously afford and free developers from enormous amounts of mechanical work. But if the final measure of those savings is simply how many chairs can be removed from the office, then it did not save the games industry. It saved the quarterly report.
-Gergely Herpai „BadSector”-
