AWS for Games Blog
Trust, not volume: The real AI advantage in game development
Key insights from the Perforce 2026 State of Real-Time Workflows Report: Game Technology & Beyond
For game developers, the past year of AI has shifted from a focus on speed to a harder reckoning: its real value and its real cost. The Perforce 2026 State of Real-Time Workflows Report, produced with Amazon Web Services (AWS), surveyed 608 practitioners across the games industry and other industries building on game technology. It puts real data behind the question of value and cost, with 69 percent of respondents now actively using generative AI. What follows is our summary of the report’s key findings, plus our own perspective on what those findings mean in practice, drawn from working with game studios and partners.
The AI paradox and governance decision
The benefits and challenges of using AI show up at nearly the same scale, which is why the report calls it an AI paradox. AI fixing a source code bug is broadly accepted, while AI-generated art draws far more scrutiny. Reasons cited in the report include compliance and intellectual property (IP) concerns. Those concerns are the right place to start because they frame the decision every company now faces. Which types of work do we want to use AI for, and which tools are the right fit for our requirements?
In games, we see that choice playing out across various categories: AI code assistance, reviews, and testing are largely embraced. Generative AI for content production spans voice, concept art, and 2D and 3D assets, and it poses the hardest questions about consent and creative ownership. Agentic live ops, an emerging category, has its own operational and trust considerations. No two studios will make the same choices, and they don’t need to. When AI is embedded into player-facing products, compliance workflows, and operational decision-making, teams need a framework for how AI is evaluated, monitored, and corrected. This should be a living set of principles that tells teams what they can deploy, how they measure its impact, and when they pull it back. Teams can use features like Amazon Bedrock Guardrails to define and monitor policy on AI outputs, while surrounding security and access controls keep sensitive IP and model data governed. Without a governance framework like this, the pace of adoption becomes a liability rather than an advantage.
Studios adopt a playbook of AI disclosure policies, human-in-the-loop for final assets, and strong data provenance. As AI and real-time tools push more content, code, and assets through every pipeline, one basic method has become non-negotiable: version control, especially as adoption grows. Adoption is currently at 94 percent, up from 86 percent in 2025. With AI agents now committing code and generating content alongside human team members, the differentiator is provenance. That means knowing where an asset came from, how it changed, and whether it can be trusted. Version control is the single source of truth for distributed teams, disciplines, and now AI agents, with a traceable history for compliance and a rollback safety net.
Hybrid infrastructure is the new normal
Games are built by teams spread across studios, cities, and time zones, increasingly working alongside AI agents that commit code and generate content. And version control rarely stands alone. It sits next to the build pipelines that turn commits into playable builds, the project management that keeps far-flung disciplines in sync, and the quality assurance that validates every change before it ships. When you see version control as the backbone of that wider workflow rather than basic file management, the case for running it in the cloud is clear. For years, teams treated cloud and on-premises as a binary choice. This year’s data shows that framing is gone. A mix of cloud and on-premises is now the single most common setup, nearly tripling to 27 percent from 10 percent in 2025. On-premises-only deployments dropped to 21 percent from 28 percent in the same time. Add in the quarter of teams running cloud only (25 percent), and cloud adoption is clearly rising, primarily in hybrid form rather than cloud-only. According to the report, when asked which cloud provider companies use, AWS led the field at 32 percent. The report also flags a real awareness gap: 20 percent of respondents weren’t sure how their team syncs files at all.
Hybrid doesn’t mean moving everything to the cloud at once. The teams who are most successful sequence the move, starting where the cloud adds the most value. Collaboration across sites is a natural driver, giving remote teams a shared, elastic place to work together. For those working through where their P4 depots and asset stores should live across a hybrid split, AWS has published a related post on choosing the right storage backend for Perforce P4.
Predictions and sentiment
Asked where AI is taking game development next, respondents and experts at Perforce described changes in how teams are built, how games are made, and how AI use is governed.
- Team sizes shrink. Studios report 20 people doing the work of 50, and one executive producer predicts teams could converge to “solopreneurs or 3-person lead-only teams (tech, art, business)” as AI collapses production pipelines (one studio cut a legacy pipeline from 2 months to 2 days).
- The bottleneck moves from production to design. As asset creation gets more affordable, a game development CTO argues, “the bottleneck shifts from asset creation to design vision,” and experts at Perforce agree that judgment, not speed, becomes the scarce resource (“it just raises the cost of losing sight of it”).
- AI moves inside the engine. Expect generative AI and procedural systems embedded directly in game engines, engines doubling as real-time simulation platforms (games, virtual production, training), and modular live content pipelines that keep updating after launch.
- The market stretches at both ends. Rather than converging, the industry splits: cheap, disposable, viral games as social media, and deeper, more immersive craft, precisely because craft is no longer the default.
- Governance tightens. Disclosure and restrictions are already arriving: Steam requires developers to disclose player-facing generative AI use, and the Godot Foundation has banned AI from generating substantial engine code.
Bottom line
AI can help teams produce more, faster, but that isn’t automatically better. The teams that succeed will be the ones with the judgment to know what’s worth building, the discipline to say “no,” and the version control hygiene to trust what they ship. Successful teams deliberately decide which work to apply AI to, govern how AI tools are chosen and used, and keep distributed collaboration secure and resilient.
The report closes on a human note worth carrying forward, from a respondent who said that if we want this industry to survive, we need to think generationally. The efficiencies available today are real. The work worth doing is to make sure they don’t come at the expense of tomorrow’s talent, expertise, and creativity. That’s a challenge game studios and the industries building on their technology now share.
Read the 2026 State of Real-Time Workflows Report, or explore AWS services and solutions on the AWS for Games website.