Studios are drowning in tools, running endless trials, and still coming up short. The hard truth? Most of the industry is asking the wrong questions entirely.
Five Hundred Tools. A Handful of Winners
Jon Gibson, Head of Transformation at Keywords Studios, recently completed an internal review of approximately 500 AI tools. The number that proved genuinely useful in real production environments? A small fraction.
Read that again, slowly.
Not because it’s a damning verdict on AI itself — the technology is real, the potential is real — but because it exposes something the games industry has been quietly avoiding: we’ve mistaken activity for progress.
Evaluating hundreds of tools and walking away with a handful of keepers isn’t a research success. It’s a symptom. It tells you that studios are exploring without direction, spending time and resources on trials that were never anchored to actual problems in the first place.
The tools exist. The ambition is there. The thinking, in too many cases, isn’t.
The Demo Trap
Walk into any AI showcase, and things look impressive. Models generate assets at speed. Pipelines automate in ways that feel genuinely transformative. The demonstrations are polished, the possibilities feel vast, and it’s easy to leave convinced that adoption is simply a matter of choosing the right product.
Then you get back to production.
Real game development doesn’t look like a demo. It’s iterative, chaotic, full of edge cases, interdependencies, and team-specific workflows that no vendor has ever seen. The gap between what a tool can do in isolation and what a team can actually rely on day-to-day is where ambition quietly dies.
As Gibson told The Game Business: “Everyone’s focusing on building better AI, and no-one’s really focusing on how to use it in a live production environment.”
That’s the real frontier. Not whether the technology is impressive, but whether it’s integrable. Whether it’s consistent under pressure. Whether it can be governed, trusted, and handed off across a team without everything falling apart. Capability, in isolation, is the easiest part of this problem.
When ‘Cool’ Becomes a Strategy
There’s a pattern that plays out across technology cycles, and AI in games is following it faithfully. The tool arrives first. The use case is invented to justify it. The actual business problem, if anyone remembers to define it, gets retrofitted around the original excitement.
Gibson puts it plainly: “A lot of people focus on what’s cool. They focus on the tool itself or the model itself, rather than what they’re trying to do.”
“A company will use a tool or build a tool without a specific use case and try and cram it into their production pipelines, rather than flipping that problem around and saying: ‘What are our pain points? What are we trying to solve?’ And then building a tool against that.”
This isn’t a critique of curiosity. Curiosity is essential. The problem is when curiosity replaces strategy, when “we’re exploring AI” becomes a substitute for “we know what we’re trying to fix.” Studios that can’t articulate the problem they’re solving with any given AI investment aren’t really adopting AI. They’re collecting it.
The People Problem Nobody Wants to Address
Technical complexity is one challenge. The human dimension is another, and it may prove harder to solve.
Developer concern around AI has not softened as the technology has matured. It has intensified. Gibson’s observation here is striking: “That statistic of 52% of developers being concerned about the usage of AI, that’s gone up every year for the last three years. As AI tools and AI models and AI technology has become more prevalent, the lack of understanding and the concern has increased.”
This is counterintuitive only if you assume that familiarity breeds comfort. What it actually suggests is that as AI has moved from theoretical to practical, the stakes have felt more real. Concerns about job security, creative ownership, transparency, and who ultimately benefits from automation aren’t abstract positions to be managed. They’re legitimate questions that deserve genuine answers.
Studios that treat developer unease as friction to be overcome will not build the internal trust that sustainable AI adoption requires. Studios that treat it as a useful signal will.
The Harder, Quieter Work
The games industry doesn’t lack for AI enthusiasm, and it doesn’t need more of it. What it needs is the less exciting stuff: governance frameworks, honest impact measurement, clearly defined problems, and leaders willing to say “we’re not ready for this yet” rather than reaching for a tool because the moment feels right.
Gibson’s “chaos phase” is real. And it has a cure, but it isn’t more experimentation.
It’s asking better questions before you open the demo.