Ask anyone in the games industry today whether the adoption of artificial intelligence (AI) is coming, and you’ll probably get a wide range of opinions. But according to Richard Tsao, Director of Innovation at DigiPen Institute of Technology Singapore, that question itself is already outdated. “It’s not a question of if,” he says. “It’s a question of order, and the order is predictable: pipeline first, playing-facing content second, AI-native game mechanics last.”
Having been in the industry for close to two decades, Richard says that this same question now surfaces in nearly every conversation he has with studios, faculty, and students alike. Here is how he sees it unfolding.
The economics make adoption inevitable
AAA game budgets have roughly doubled across every console generation for the last twenty years, while live-service titles demand a never-ending stream of new content. As such, any technology that cuts cost, raises content output, and eventually raises quality will get adopted. Richard explains that we’ve seen this story unfold before across different technologies like middleware, motion capture, and procedural generation. Nobody today asks whether using tools like SpeedTree instead of hand-modelling every single asset is “cheating.”
To studios managing a budget, AI is in the same category of decision-making, but on a bigger scale. Furthermore, the technology does not need to be perfect in order to be adopted — it just needs to be better than the cost of doing it by hand, one task as a time.
Stage one: the invisible pipeline
“The first wave of adoption is already happening in the game development pipeline,” Richard says. Industry surveys back this up: GDC’s 2026 State of the Game Industry report found that half of surveyed respondents (52%) reported generative AI use in their companies, with AI use concentrated in research and brainstorming (81%), code assistance (47%), and prototyping (35%). Meanwhile, a Google Cloud survey from August 2025 found that roughly 90% of developers used AI somewhere in their workflow.
Richard explains that teams mainly use generative AI tools for rapid prototyping where they create mockups and test ideas. “Throwaway work is where AI is strongest, because nothing it produces ships,” he says. Concept art was the first discipline to be hit, with Chinese studios in particular cutting illustration outsourcing sharply back in 2023 when image generation became usable. An example of this is Tencent’s Hunyuan Game Visual Generation Platform, which was launched in May 2025. It is a production engine for concept art and character iteration. During GDC 2023, Ubisoft also introduced their Ghostwriter tool, which generates first drafts of non-playable character (NPC) barks, which are short lines such as, “Hey, watch it!” Games need these by the thousands, and writers pick and edit what Ghostwriter generates.
Still, Richard notes that these tools work in the development pipeline because “failure is cheap and nobody needs convincing.” This does not mean that AI can take over the entire development process. He points to a useful cautionary tale, where Keyword Studios tried to build a game using only generative AI in 2023. The studio tested over 400 tools before concluding that the tech was “unable to replace talent.” The experiment needed people from seven studios to rescue it. “As a replacement for developers, AI fails. As an assistive layer inside the pipeline, it already pays for itself, and that is exactly how studios are using it,” Richard says.
Stage two: player-facing AI
The next stage of AI adoption in game development is in content that players actually touch, such as dialogue, NPCs, and in-game creation tools. Richard explains that adoption here is gated by player acceptance — not AI capability — and this acceptance moves at very different speeds around the world. Chinese studios are already shipping LLM-driven NPCs at scale, from NetEase’s Justice Mobile to Tencent’s Game for Peace.
The West, meanwhile, is still negotiating. The Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) struck against game companies for nearly a year, primarily over AI voice replicas. Steam has also required AI disclosure since early 2024. There has been ongoing public backlash amidst all these, and the GDC survey captures this mood precisely: the share of developers who view generative AI negatively climbed from 18% to 30% to 52% across three annual surveys, over the same stretch in which Steam’s AI disclosures grew eightfold. This signals that adoption and resentment are rising together.
Richard sees a familiar pattern here. “Free-to-play was born in Korea and China and dismissed in the West for a decade before Fortnite made it the industry’s biggest revenue engine,” he says. Mobile game adoption had a similar trajectory, where they were initially not considered “real games” in the early 2010s, yet make up more than half of global game revenue today. “The West does adopt what works — it just runs one market-proof cycle behind,” he explains. Still, he’s careful to note that Chinese adoption isn’t blind either. Chinese players have boycotted AI art they see as cost-cutting, even as AI companions are welcomed as headline features. “The objection is to lazy AI, not to AI itself,” he adds.
Stage three: AI-native mechanics
The most interesting — but likely slowest — stage to arrive will be games where AI is not a production tool but the mechanic itself, and generation is the gameplay. Early experiments like GoodAI’s AI People and Anuttacon’s Whispers from the Star showed novelty appeal but unproven retention. Still, this doesn’t mean that the category is dead; it means it is early. Wider adoption can take time and Richard points out that free-to-play systems took around fifteen years to translate from MapleStory’s cash shop to Fortnite’s peak. “Somebody will crack the design loop where AI generation is the fun rather than the gimmick, and in hindsight it will look obvious. My honest bet is that the game which defines this category has not shipped yet, and its designer may still be a student,” Richard quips.
What this means for developers, and why indies should pay attention
For game developers, Richard’s advice is direct: reposition now. “The scarce skills are shifting from production technique toward design judgment, taste, and the ability to direct AI tools across disciplines you did not train in,” he says. The rules are being reset, and reset rules favor people with nothing to unlearn. This is a shift that DigiPen (Singapore) is already building into its curriculum, where students work in small, interdisciplinary teams to ship ambitious work with AI in the pipeline from day one.