Nobody Talks About This One Simple Trick for AI in Game Development
I just got out of a pitch meeting. The founder, a kid barely out of college, was vibrating with excitement. He showed me a deck full of buzzwords—'procedural narrative,' 'emergent AI,' 'fully autonomous NPCs.' He was promising a game that would change everything. I’ve seen this movie before. I passed.
This might be controversial, but someone has to say it. The entire AI in game development industry, or at least the public perception of it, is built on a house of cards. Everyone is chasing a ghost. And it’s leading to a lot of wasted time, money, and talent.
The Great AI Lie
Everyone is chasing the same thing: more. More realistic graphics, more complex NPCs, more sprawling open worlds. And they’re all convinced that generative AI is the magic wand that will get them there. They’re wrong. Dead wrong.
I’ve seen it dozens of times. A promising young studio, burning through my cash, trying to build the next big thing. They get obsessed with using AI to create the most “intelligent” enemies the world has ever seen. They spend months, even years, developing complex behavioral trees, reinforcement learning models, and neural networks. And what do they end up with? A game that’s buggy, unbalanced, and worst of all, not fun.
I remember one team in particular. Let's call them 'Project Chimera.' A group of brilliant engineers, Ph.D.s from Stanford and MIT, some of the smartest people I’ve ever met. They had a vision for a revolutionary new RPG where every single NPC was a unique, learning AI. They could remember your actions, form relationships, and have their own goals. It sounded incredible on paper. They raised $10 million in their seed round—I was one of the first checks in. They hired a big team and got to work.
Two years later, that $10 million was gone. They had a tech demo that was, I admit, impressive. You could talk to an NPC baker who would remember you bought a croissant yesterday and ask if you wanted another. But that was it. There was no game. The 'learning' NPCs were so unpredictable that they broke every quest they were supposed to give. The system was so computationally expensive it could barely run on a high-end gaming PC, let alone a console. They had fallen into the trap of chasing the AI dragon.
They were so focused on the “AI” part of the equation that they forgot about the “game” part. They were building a technology, not an experience. The project was eventually shuttered, a painful lesson for everyone involved.
The Simple Trick Nobody Talks About
Here’s the secret that nobody in those flashy pitch decks wants to admit: AI in game development isn’t about making games more complex. It’s about making them easier to create.
That’s it. That’s the whole trick. The real power of AI isn’t in creating hyper-realistic worlds or super-intelligent NPCs that can pass the Turing test. It’s in automating the tedious, time-consuming, and expensive tasks that bog down the development process. It’s about empowering small teams to create amazing things without needing a nine-figure budget.
Think about it. What are the biggest bottlenecks in game development? It's not a lack of grand ideas. It's the grunt work. Asset creation, level design, and testing. These are all areas where AI can have a massive, immediate, and practical impact.
1. AI-Powered Asset Creation: Your Art Team on Steroids
Instead of spending months modeling and texturing every single rock, tree, and sword in a game, developers can use AI to generate high-quality assets in a fraction of the time. I have an investment in a small indie studio that was building a fantasy RPG. Their art team was just two people. They were spending 80% of their time just creating environmental assets. We introduced them to a suite of AI tools for texture generation and 3D model variation.
Suddenly, they could create a hundred different-looking trees in the time it used to take to make one. They used AI to generate base textures for their armor sets, which the artists would then refine and customize. The result? They cut their asset creation time in half. That’s not a small improvement. That's the difference between shipping a game and running out of money.
And no, the AI didn't replace the artists. It made them art directors. It freed them from the monotonous work and allowed them to focus on the creative vision and unique hero assets that made the game special.
2. Procedural Content Generation (PCG): Building Worlds, Not Just Levels
This is another area where the hype often misses the point. People hear 'procedural generation' and think of infinite, bland, cookie-cutter worlds. That's the old way. The new way is AI-assisted PCG.
AI can be used to create vast, varied game worlds that still have a human touch. A level designer can set the high-level rules: 'I want a forest here, a mountain range there, and a river running through it. The forest should feel dense and mysterious, the mountains treacherous.' The AI can then generate the terrain, place the foliage, and even create points of interest based on those rules. The designer then comes in and adds the handcrafted touches, the unique locations, and the narrative elements.
It’s a partnership. One of my portfolio companies, building a space exploration game, is using this to great effect. Their universe is technically infinite, but the AI ensures that every solar system has a unique feel and a plausible layout based on astrophysical rules. It’s not just random noise; it’s structured, intelligent creation at a scale that would be impossible for a human team.
3. Automated Testing: The Bug Squasher
Testing is the unglamorous, unsung hero of game development. It's also a massive resource sink. You need to pay an army of QA testers to play your game for thousands of hours, trying to break it in every way imaginable.
Or, you can teach an AI to do it. I’ve seen this firsthand at Scale AI, one of my most successful investments. They are masters of data labeling and model training. The same principles can be applied to game testing. You can train an AI to play your game, rewarding it for finding bugs, exploits, and balance issues. It can run 24/7, testing every new build, and providing a detailed report of every issue it finds, complete with video recordings.
This frees up your human testers to focus on the things an AI can't judge: Is the game fun? Is this mechanic intuitive? Does the story make sense? You're using the AI for what it's good at—brute-force repetition—and your team for what they're good at—creative and subjective analysis.
The Unseen Artist: Where AI Fails
For all its power, there’s a hard limit to what AI can do. It’s a phenomenal tool for optimization and automation, but it is not, and will not be, the creative heart of a game. A game isn’t just a collection of assets and systems; it’s a curated experience. It’s the deliberate choice of what to show the player and what to hide. It’s the careful pacing of a story, the precise tuning of a difficulty curve, the emotional arc of a character.
This is the work of an artist, a designer, a storyteller. AI can generate a thousand swords, but it can’t design a single sword that tells a story, that feels perfectly weighted in the player’s hand, that becomes an iconic part of the game’s identity. AI can generate a sprawling, endless world, but it can’t create a single, memorable level like the Spencer Mansion in Resident Evil or the clock tower in Majora’s Mask. Those are products of intent, of a human creator making specific choices to evoke a specific feeling.
I worry that an over-reliance on AI will lead to a generation of games that are vast but shallow, technically impressive but soulless. We’ll have endless worlds to explore, but no reason to explore them. We’ll have NPCs that can talk forever, but have nothing to say. The magic of games comes from the connection between the creator and the player. It’s a conversation. AI can’t be a part of that conversation. It can only set the stage.
The Real Revolution
This is where the real revolution is happening. It’s not in the flashy, headline-grabbing AI features. It’s in the trenches, where AI is being used to make the process of game development faster, cheaper, and more efficient.
I’ve invested in over 200 companies, including some of the biggest names in AI like Anthropic, OpenAI, and Hugging Face. I’ve seen firsthand how this technology is transforming industries. The common thread among the winners is that they don't use AI as a gimmick. They use it as a tool to solve a real, painful problem.
In game development, the most painful problem is the sheer cost and complexity of production. AI, when used correctly, is the solution.
So, if you’re a game developer, I urge you to stop chasing the AI dragon. Stop trying to build the most complex, feature-rich game imaginable. Instead, ask yourself: What is the most tedious, time-consuming part of my workflow? And how can I use AI to automate it?
That’s the simple trick that nobody talks about. Forget about sentient NPCs. Focus on building a better, faster, smarter development pipeline. That’s how you’ll win. And that's the kind of company I'd invest in.
Frequently Asked Questions
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.