[GamesIndustry.biz] 87% of game developers are using AI agents in their workflows, says new survey

SomeNorseGuy

Gold Member

87% of game developers are using AI agents in their workflows, says new survey

While 97% of game developers believe AI is "reshaping" the games industry

A new survey by Google Cloud and The Harris Poll has found that 87% of developers are using AI agents in their workflows.

The
research survey, published on August 18, 2025, was conducted in late June and early July 2025 and asked 615 game developers in the U.S., South Korea, Norway, Finland, and Sweden about the current state of AI in the industry – and where it could be heading next.
Google Cloud found that respondents "largely" agree AI is "having a positive influence across a wide range of creative efforts, business settings, and internal workflows," with "more than 90%" of respondents saying it is " helping with an array of challenges, including driving innovation, and enhancing the player experience."
"97% of those surveyed said generative AI is "reshaping" the gaming industry, while 95% said it's "reducing repetitive tasks in workflows," and 94% said it's "driving innovation."

47% of respondents reported that AI is speeding up playtesting and balancing of mechanics, 45% said it assists in localization and translation of game content, and 44% said it improves code generation and scripting support."

In addition, 89% of developers believe AI integration is "changing player expectations," with 37% reporting that they've found gamers are looking for "more lifelike experiences."

According to the survey's findings, 40% of respondents see AI-driven game engines and AI for balancing gameplay as the most promising trends, followed by AI-powered testing and QA at 36%.
However, the report noted that "game developers face some hesitancy around the adoption of gen AI, particularly due to data and ownership rights."

63% of developers expressed concerns about data ownership, with 35% worried about player data privacy, 32% concerned about unclear licensing, and 32% concerned about ownership of AI-generated content.
"Google Cloud found that AI has changed team compositions, too. 62% of respondents say new AI-focused roles have emerged, while 56% of respondents said existing roles have "evolved" to include AI-related tasks.
The biggest barriers to AI adoption, according to the survey, are "difficulty measuring the success of AI implementations" (25%) and "cost of AI integration" (24%).

However, 94% of respondents expect AI to reduce overall development costs in the long term (in three or more years), and 40% said it is "creating new business models or strategies."

"Overall, the research found widespread adoption of gen AI in the games industry – and a surprising level of optimism for it," the report noted.

"AI is already making a big difference in developer workflows, including productivity and creative tasks. Developers also see promising possibilities with AI agents and other emerging AI tools to accelerate game development and enhance player experiences.

"And while developers raise important concerns about IP issues and the ownership of AI-generated content, the overall feedback on how AI can impact the holistic games industry is trending positive – with some even expressing that a more inclusive and democratic future lies ahead."

------------------------------------

udlXju1vaS4qzbDE.jpg


kTL0GLr3SToGFmdQ.jpg
 
Sounds good. I'm sure they're much better than their human counter-parts because over the past few years it's like human devs have forgotten how to optimize games for their respective platforms, put out unfinished games, whine and complain about how they are treated, complain about pay, take too long to put out games etc.

I, for one, welcome our new AI overlords.
 
And it's growing at an inside rate. The is a vocal minority that's opposed to AI in gaming, and I think the origins of that fall under seeing actual assets in gaming (visually, audio, etc), however, I think AI is another engineering facet that will overtake the gaming community in all ways sooner or later.
 
I dont know anything about AI, but supposedly AI tools have been used for years commercially. It's just that it was a hush hush kind of thing and AI only became a hot topic when the masses could use the million AI websites that popped up the past year to goof around and make funny images.
 
why wouldn't you?, that doesn't mean the code creation is not strictly supervised, it's easier and faster.

I can copy and paste my code into copilot and it can find out what I'm doing wrong in a second vs a couple of hours of googling. It's pretty bad ass.

Imagine game devs doing the same to find inefficient code, memory leaks, etc. This is good.
 
I can copy and paste my code into copilot and it can find out what I'm doing wrong in a second vs a couple of hours of googling. It's pretty bad ass.

Imagine game devs doing the same to find inefficient code, memory leaks, etc. This is good.
It's going to be a tit for tat kind of thing. No different than when an office implements a new financial ERP program.

Some people love it, makes their job easier, and it helps with data and dashboards.

Some other people hate it as they cant figure out how to use it to add value to their work. So they are shitty bricks they'll be out of a job when bosses realize they are falling behind.
 
And the remaining 13% will start doing it soon or loose their job. As a dev myself (business applications though) AI heavily improved my productivity and completely replaced google search for me. AI won't be something you can ignore or actively fight. It's just too damn useful and will be part of most professions soon.
 
It's going to be a tit for tat kind of thing. No different than when an office implements a new financial ERP program.

Some people love it, makes their job easier, and it helps with data and dashboards.

Some other people hate it as they cant figure out how to use it to add value to their work. So they are shitty bricks they'll be out of a job when bosses realize they are falling behind.

The pros far outweigh any cons when it comes to software development. Those who don't utilize AI are going to look incredibly slow compared to those who do.
 
A lot of tools already had constant "automation" or workflow improvements with every update; even super old packages like 3DSMAX 5 felt way better when updated to 3DSMAX 7, same things with photoshop, illustrator etc.
There's really no reason to not keep using tools/plugins/updates that increase productivity just because it might have an LLM backend now; it's really business as usual for developers, it's just the use of full generative LLM for dodgy assets or "vibe coding" that's seen as a problem
 
The pros far outweigh any cons when it comes to software development. Those who don't utilize AI are going to look incredibly slow compared to those who do.
I have a couple of devs who absolutely refuse to use it and they're adamant they're better coders than copilot. I told them that their reality is becoming just like the one blacksmiths faced with the Industrial Revolution. It doesn't matter how high the quality of your horseshoes is, a horseshoe factory can produce good enough horseshoes much faster and cheaper than you can produce really great ones. So invest in a factory before someone else does and you're out of a job. Coding has become a commodity that I can get anywhere. What I increasingly need are people who can build systems from commodity parts.
 
I have always thought that Math, Logic, and Philosophy should be taught together under Programming in schools, starting from elementary school.

Understanding the concepts is what's important. AI agents can help you get around the syntax quirks and features of any programming language
 
I have a couple of devs who absolutely refuse to use it and they're adamant they're better coders than copilot. I told them that their reality is becoming just like the one blacksmiths faced with the Industrial Revolution. It doesn't matter how high the quality of your horseshoes is, a horseshoe factory can produce good enough horseshoes much faster and cheaper than you can produce really great ones. So invest in a factory before someone else does and you're out of a job. Coding has become a commodity that I can get anywhere. What I increasingly need are people who can build systems from commodity parts.

Yeah, exactly. AI is a tool. It isn't about being a better coder than the AI. The AI isn't writing code for an entire application. It is solving specific problems quickly and efficiently so the developer can continue making progress. It's just a better google.

Recent example is an issue I was having pushing my app into Azure. Docker spit out a whole bunch of issues. I couldn't make heads or tails of that so I copied and paste the entire docker log into copilot and it spit out the answer. I would have spent hours trying to figure it out. Copilot told me to add one line of code and bam.....done.
 
Last edited:
I dont know anything about AI, but supposedly AI tools have been used for years commercially. It's just that it was a hush hush kind of thing and AI only became a hot topic when the masses could use the million AI websites that popped up the past year to goof around and make funny images.
I've been using AI for professional/commercial development since mid 2023. It continues to get better at a rapid rate. Now I am a full blown AI junky, I no longer write code.

System design and planning is more important than knowing how to write code.
 
Last edited:
I mean, yeah, for asking quick questions to Copilot or stuff like that, not to make our games code themselves lol. I asked Copilot for examples of code for my Unity project once and I won't ask the same thing next, the thing takes more time thinking and still being out of context that if I had to code using it I wouldn't finish stuff in a timely manner.

Same for my real job, I spent a whole afternoon in prompts, review, etc and didn't even reach bugs fixing (leaving it for later this week) in a ticket I would have complete by myself already, just because management wants us using AI with no way around it.
 
Last edited:
why wouldn't you?, that doesn't mean the code creation is not strictly supervised, it's easier and faster.
Yep, pretty much any dev out there is going to use modern LLMs for code assistance. Folks shouldn't confuse this with pure "vibe coding" though.

Devs use this for auto complete, help with particular functions, etc…. But "most" don't let AI code and submit without oversight.

Edit: The biggest issue are going to be newer devs who have no idea of what is actually good/decent vs crap code as they have never solved issues by themselves.

I firmly believe that newer generations of young workers will be less productive with less critical thinking and problem resolution capabilities because they let AI do the "thinking" for them.

It's one thing for an experienced dev who spent 10+ years in the industry and years of dev work in school before who now uses AI for assistance.

It's another issue for a new dev who got through College with ChatGPT and is now using AI for work. I can just imagine the slop being submitted.
 
Last edited:
Yeah, exactly. AI is a tool. It isn't about being a better coder than the AI. The AI isn't writing code for an entire application. It is solving specific problems quickly and efficiently so the developer can continue making progress. It's just a better google.

Recent example is an issue I was having pushing my app into Azure. Docker spit out a whole bunch of issues. I couldn't make heads or tails of that so I copied and paste the entire docker log into copilot and it spit out the answer. I would have spent hours trying to figure it out. Copilot told me to add one line of code and bam.....done.
Exactly. I use it for triage use cases all the time. Over-engineered code crapping the bed? Let my buddy Claude analyze it and tell me exactly what's wrong in under 30 seconds versus spending hours researching and debugging. Then it does me the solid of documenting the problem and the solution as well as creating test cases and unit tests. Devs need to know how to put the robots to work.
 
As they should, AI will be the only way to keep AAA development sustainable in the future, to reduce unacceptably long (5-7 years) development cycles and also to introduce a new level of realism in the way digital worlds feel alive and NPCs credible.
I don't get the ideological crowd opposing this because they think everyone will be fired, it's anachronistic, things have become progressively more automated since the first industrial revolution, the reality is that some jobs will become obsolete and will be replaced by other new figures.
 
What I increasingly need are people who can build systems from commodity parts.
Let's not pretend this is a recent development - tech has been doing it for literal decades. One of the reasons many engineering jobs 'are' threatened is that they haven't been doing work that is particularly challenging (for a human) for a long time.
Granted - the best way to protect yourself at those jobs it so simply scale-up the throughput, and AI is certainly helpful there - but let's be real - the general trends aren't positive here. Tech (at least mainstream) keeps getting worse and that trend was established a long time before AI became viable to help it along.

but supposedly AI tools have been used for years commercially.
I mean about half of the quoted use-cases in the article predate gen-AI by at least a decade, so yes.
But there's significantly more noise around it now so even if you do nothing new, just calling it agentic AI is automatically groundbreaking.

however, I think AI is another engineering facet that will overtake the gaming community in all ways sooner or later.
Ok I've read this like 5 times and I still have no idea what you're trying to say. Facet overtake the what in the where now?
 
Yep, pretty much any dev out there is going to use modern LLMs for code assistance. Folks shouldn't confuse this with pure "vibe coding" though.

Devs use this for auto complete, help with particular functions, etc…. But "most" don't let AI code and submit without oversight.

Edit: The biggest issue are going to be newer devs who have no idea of what is actually good/decent vs crap code as they have never solved issues by themselves.

I firmly believe that newer generations of young workers will be less productive with less critical thinking and problem resolution capabilities because they let AI do the "thinking" for them.

It's one thing for an experienced dev who spent 10+ years in the industry and years of dev work in school before who now uses AI for assistance.

It's another issue for a new dev who got through College with ChatGPT and is now using AI for work. I can just imagine the slop being submitted.
The pure vibe coding approach is what I've been asked to do in my job and it's not only boring, it can lead to more delays and really bad code that I can't author myself since it's the AI generating it thus don't feel good if it breaks I get to be blamed for it.

For autocomplete, documentation, help with bugs detection, etc. It's amazing tho, that the way it should be used and I've used it that way for around a year now, it helped me skip all the SEO bs in blogs and get right to the solution.
 
The pure vibe coding approach is what I've been asked to do in my job and it's not only boring, it can lead to more delays and really bad code that I can't author myself since it's the AI generating it thus don't feel good if it breaks I get to be blamed for it.

For autocomplete, documentation, help with bugs detection, etc. It's amazing tho, that the way it should be used and I've used it that way for around a year now, it helped me skip all the SEO bs in blogs and get right to the solution.
Rightfully so, you should be the one to blame. I don't understand your excuse "it's not my fault, AI wrote the code". You're stamping your name on the PR when you push the code, its your fault.

Currently AI is only as good as the user prompting it. It will spit out messy, unmanageable code if not steered towards logical and maintainable.
 
Sounds good. I'm sure they're much better than their human counter-parts because over the past few years it's like human devs have forgotten how to optimize games for their respective platforms, put out unfinished games, whine and complain about how they are treated, complain about pay, take too long to put out games etc.

I, for one, welcome our new AI overlords.

It's because tools like this is making everyone retarded.
 
Rightfully so, you should be the one to blame. I don't understand your excuse "it's not my fault, AI wrote the code". You're stamping your name on the PR when you push the code, its your fault.

Currently AI is only as good as the user prompting it. It will spit out messy, unmanageable code if not steered towards logical and maintainable.
But why? I'd be literally put the AI to do something and directly make into the code base, unless you're supposed to check line for line, in which case you better off do it yourself and use AI as a support tool as many of us already do
 
I can copy and paste my code into copilot and it can find out what I'm doing wrong in a second vs a couple of hours of googling. It's pretty bad ass.

Imagine game devs doing the same to find inefficient code, memory leaks, etc. This is good.
Exactly. This is how I imagine AI is implemented at first.
 
Let's not pretend this is a recent development - tech has been doing it for literal decades. One of the reasons many engineering jobs 'are' threatened is that they haven't been doing work that is particularly challenging (for a human) for a long time.
Granted - the best way to protect yourself at those jobs it so simply scale-up the throughput, and AI is certainly helpful there - but let's be real - the general trends aren't positive here. Tech (at least mainstream) keeps getting worse and that trend was established a long time before AI became viable to help it along.
I never said it was a recent development. Automation has been bootstrapping code development for more than a decade. But this is the first time you've been able to literally tell your computer what kind of problem you're trying to solve and have it built a complete custom solution tailored to your needs. Then when you want a change two sentences to an LLM can make the change and create a pull request. It has never been like it is right now.
 
The progress in agentic coding in the last ~1 year has been amazing ever since Sonnet 3.5. Copy and pasting between a chat app and an IDE feels so slow now when you have the Agents going all over your codebase actually producing valuable work editing multiple files and 100s of lines of code per prompt, and it will only keep getting better. For example, the current internal research OpenAI model gets gold in Programming Olympiad, got 2nd place at the AtCoder 2025 competition (you need to work for 10hs in this one) and also gets gold in Math Olympiad. Supposedly it should be integrated into main GPT models by end of year.
 
I can copy and paste my code into copilot and it can find out what I'm doing wrong in a second vs a couple of hours of googling. It's pretty bad ass.

Imagine game devs doing the same to find inefficient code, memory leaks, etc. This is good.

That's not how you use agents, though.

We are adopting it (healthcare industry), at a slower pace though because base code needs to fulfill certain guidelines set by ONC which the AI doesn't really know about until you have told them. AI agents have been a game changer (no pun intended), though, speeding up development. As I read somewhere, those who are not using look to be static rocks compared to the speed the other developers are working.

Kind of ironic that some gamers don't have problem with developers using AI but cannot stand them using it for assets.

For example, the current internal research OpenAI model gets gold in Programming Olympiad, got 2nd place at the AtCoder 2025 competition (you need to work for 10hs in this one) and also gets gold in Math Olympiad. Supposedly it should be integrated into main GPT models by end of year.

I don't really trust certain competitions tailored for OpenAI. Everything should be taken with a pinch of salt.
 
Last edited:
That's not how you use agents, though.

We are adopting it (healthcare industry), at a slower pace though because base code needs to fulfill certain guidelines set by ONC which the AI doesn't really know about until you have told them. AI agents have been a game changer (no pun intended), though, speeding up development. As I read somewhere, those who are not using look to be static rocks compared to the speed the other developers are working.

Kind of ironic that some gamers don't have problem with developers using AI but cannot stand them using it for assets.

Sure, but it is the same concept. I have Copilot access embedded in Visual Studio. You can also use editors like Cursor. You are still telling AI to do some work. I'm just providing examples of where I've found AI to be most helpful.

But I do see how it would be incredibly valuable when you have guidelines that must be followed.
 
Last edited:
At this point if your work allows you to use AI productivity tools and agents and you're not getting up to speed, you're choosing to be inefficient.
 
Top Bottom