"We still have game artists but they’re using our AI generation funnel"

Welcome to another edition of AI Gamechangers, your weekly newsletter where we sit down with games industry leaders who are putting AI to practical use.

This week, we’re in conversation with Erkay Uzun, co-founder and CEO of Unico Studio. With more than 50 games and 1.6 billion downloads under its belt, Unico is one of the biggest names in casual and hybrid-casual mobile gaming. Now, AI has been deeply woven into the company’s operations, from automated review analysis and AI-generated store assets to agents that play-test levels thousands of times before a player ever sees them.

Remember, our archive of conversations with industry leaders is free to browse, so dive in.

Erkay Uzun, Unico

Meet Erkay Uzun, founder of Unico Studio and a computer science-trained entrepreneur. Unico Studio is a mobile games company headquartered in Palo Alto with more than 50 successful titles. He recently hired a dedicated AI team and is also co-founding a new stealth AI startup alongside a former Google AI engineer.

In this interview, he shares how Unico has embedded AI across its entire pipeline. We discuss how Unico uses AI agents to handle review analysis, generate assets, support code and test game balancing, and why the bigger prize is adaptive, personalised gameplay that changes in real time.

Top takeaways from this conversation:

1. Unico’s AI journey began with a ChatGPT-powered system that analyses and answers thousands of player reviews, replacing a labour-intensive workflow and delivering immediate value.

2. The company now applies AI across multiple asset pipelines, including game art, marketing materials and app store creatives for A/B testing, with the aim of automating even more.

3. Erkay sees the next frontier not as pre-launch tuning alone, but real-time personalisation inside games: moving from fixed level design and predefined difficulty curves towards adaptive systems.

4. His advice is to stop thinking of AI as one monolithic initiative and instead break the business into tasks, then evaluate where AI agents can improve efficiency and unlock new capabilities.

AI Gamechangers: Unico Studio recently surpassed 1.6 billion downloads worldwide. Now the whole company is focusing heavily on AI. Please tell us about the moment you realised AI was going to become part of your business.

Erkay Uzun: As you know, AI is not one thing – it has a lot of different aspects. From a gaming perspective, AI can be applied in the production of the games, in user acquisition, or in customer interaction. It’s a whole journey.

When we first saw ChatGPT, it came into our mind: how are we going to utilise this technology in any part of our gaming journey? That’s an LLM, so our first intention was to create an agent to analyse the reviews we get, and answer those reviews – we have millions of daily active users leaving thousands of reviews each day. It requires a lot of time just to analyse and answer them. The first project that we focused on was to develop an agent to analyse and answer reviews. That was two and a half years ago. Before, we might have had two or three people just working on these. But now we have a ChatGPT-powered agent just answering thousands of reviews, classifying them for us: it’s a system giving us feedback, and we saw immediate value here on the AI side.

Then, of course, image creation came to market. We thought, “This is great – how can we use this in our projects, in our production pipeline?” So we started using the APIs from OpenAI, Gemini, and other image-generation models.

And then a couple of tools came to market that added extra layers on top of these LLMs. We tried some of those, but we also developed our own internal tool to generate both game assets and marketing assets. And last month, we started generating store assets as well. So now there are really three different aspects of asset generation for us.

Marketing assets are a little bit easier. For games, it requires consistency (style consistency, character consistency), so early on, we couldn’t get as much benefit. But this year the models have become much better. Now we’re able to generate most of our game assets with AI. Of course, we still have game artists and designers, but they’re using this AI generation funnel. We’re using OpenAI and Gemini so they can generate assets through these models as part of their workflow.

And most recently, we started generating store assets for A/B testing. We have more than 50 games, so it requires a huge amount of store asset generation. We thought, “Why not use AI?!” Now we provide existing screenshots, icons and feature graphics to the model, and generate alternatives. Sometimes they’re iterations of the existing ones, sometimes completely different versions. Then we upload them and start A/B tests.

“We help developers optimise everything: gameplay balancing, live ops, and user acquisition. By combining those tools with AI, we can test, improve and scale games more effectively”
Erkay Uzun

Our goal is to automate this fully. The APIs don’t yet allow full automation, but I think, starting with Google, that will become possible. So we’re building an internal system to automate not just generation, but the whole A/B testing funnel. Right now, there’s still manual work (designers or store managers uploading assets and starting tests), but the hardest part is the generation, and that’s where AI already helps.

We’ve started doing this across three or four games. It requires a different perspective, both technically and artistically, because each game is different.

So we had to personalise these AI agents for each game. Every game has a different agent, because some are puzzle games, some are idle games, and some are more brain-training games. Each one needs its own logic for generating store assets. That was one of the first areas we focused on.

After review analysis and asset generation, we also moved into code generation. Claude is great at generating code, and since we have more than 30 developers, why not integrate it into our development stack? Now they’re using Claude in their daily workflow. But honestly, that’s the part almost every software company is doing these days.

Where do you see AI going next in your workflows?

Where we really want to push further is personalisation inside the games themselves. Most of our games are level-based. We already do a lot of segmentation, but it’s more structural: pre-defined segments, pre-established difficulty curves. What we want is something more dynamic and more personalised, happening on the fly.

Right now, we’re using AI before publishing a game. We’ve built agents that play existing levels thousands of times. The designers use these agents to analyse fail rates, level duration, move counts – all that data. Based on that, they adjust the difficulty curve across levels. We started this about two months ago, and we’ve already seen significant benefit.

“Start by listing all the tasks that need to happen to create and run your games. You can start asking how AI could help with each of these tasks. Could it make them more efficient? It’s much better to break the work down into smaller pieces and evaluate them one by one”
Erkay Uzun

We’re also analysing successful games in the market. We feed those games to our agents and see how they perform: where they fail, how long they take, what patterns emerge. Our designers then use those performance metrics to make adjustments. But again, this is still pre-release optimisation.

The future, for us, is real-time personalisation. At the moment, even though we’re using AI to define difficulty, every player is essentially seeing the same levels. Yes, we make structural adjustments – if someone is struggling, we might show an easier level; if they’re progressing quickly, we increase difficulty. But it’s still predefined logic.

We want true on-the-fly personalisation: not just tuning levels before launch, but dynamically adapting them during gameplay. That’s what we’re working on now. It’s challenging. We’ve seen some Asian companies attempt this with some success, so we know it’s possible. Our next goal is to move from designing fixed levels to designing adaptive systems.

It sounds like you’ve got a plan for AI agents to work at every stage of the process. So what role do your existing human designers play?

They’re still working together. AI isn’t replacing anyone yet, but it’s helping creative minds do better work. From the game design side, AI agents are especially helpful for level designers. They can test things very quickly and generate different variations, which speeds up the process a lot.

On the development side, it’s true that we may need fewer developers than before, because AI is accelerating the workflow. In particular, we’re hiring fewer junior developers. I think that’s something we’re already seeing across the industry.

The same applies on the graphics side. We’re probably hiring fewer junior graphic designers, but we still absolutely need experienced designers. They need to see the whole picture and give direction to the AI (defining the concepts, guiding the visual style, deciding what the final output should be). AI helps with execution, but the creative direction still comes from people.

“For a small studio, resources are always scarce. One way to think about AI agents is that they’re like employees. You can treat them as workers who can help with specific jobs”
Erkay Uzun

Even before we start developing a game, we now do a lot more concept testing on the marketing side: generating different creative ideas and visual directions. AI is very helpful there as well.

Where AI is still less successful is QA. We’ve experimented with some internal tools, and we’ve also tried a couple of AI startups working in this space. They work at some level, but detailed testing still requires a lot of human effort.

So overall, I wouldn’t say it’s replacing people. It’s increasing productivity. We still need brilliant minds. Gaming is a very creative field.

More broadly, do you think the games industry has fully understood how important AI is. Do you see it being something that everyone will embrace in time?

Not just gaming; I see a lot of software companies still struggling to understand the importance of AI. Gaming is an industry more open to innovation than others, because it changes frequently, from many perspectives – user acquisition, creation, everything. Gaming companies are actually adapting better. But the problem is people often already have a system that’s working, and it requires some effort to understand what’s new. You need resources for that. And everyone understands gaming is super competitive. So you don’t have much time to learn new things.

However, I do see a lot of gaming companies trying something, at least. Things like image generation come to mind. It’s a little bit slower than I expected, actually, because AI might bring high value to gaming companies from every perspective. For instance, we are trying to generate playable ads internally with AI.

“Before we start developing a game, we do a lot more concept testing on the marketing side: generating different creative ideas and visual directions. AI is very helpful there”
Erkay Uzun

Video creation and other creative work still take a lot of time. Most teams are already busy with their weekly tasks, so they don’t always have the time to investigate new tools. People are thinking: “I already have things to finish this week – when am I supposed to explore something new?!”

That’s why a lot of AI startups are now going directly to companies and showing them how these tools work, how they can optimise production, speed up creative processes, and automate parts of the workflow. But it still requires focus. You have to deliberately make time to explore these technologies.

One reason we’ve been able to move faster with AI is that we stay very connected to the wider ecosystem. Conferences like Pocket Gamer Connects and other industry events help a lot. When you see what other people are building, it motivates you. You realise what’s possible.

We just hired three people solely to work on AI. We are researching it further; we have sessions internally in the company open to everybody. Anybody from the company (more than 100 people) can join these sessions – our AI people deliver presentations about what’s new this week, so everybody sees how we are using AI.

As a CEO, I try to connect with other founders (especially in the US, and also developers originally from Turkey) and ask what they’re doing, what tools they’re using, what experiments they’re running. That kind of exchange helps you understand the trends.

But I also see many game companies that stay very focused on their internal work and don’t look outside very much. And when that happens, it’s much harder to keep up with how fast things are changing.

If you were advising a small startup studio today, what’s the first step they should take with AI?

For a small studio, resources are always scarce, so the first step is to look carefully at your development process. One way to think about AI (especially AI agents) is that they’re like employees. You can treat them as workers who can help with specific jobs inside the company.

“We analyse successful games in the market. We see how they perform: where they fail, how long they take, what patterns emerge. Our designers then use those performance metrics to make adjustments”
Erkay Uzun

So start by listing all the tasks that need to happen in order to create and run your games. Maybe that list ends up being 15 or 20 different tasks. It could include development, marketing, user acquisition, QA, anything that different teams are responsible for.

Once you have that list, you can start asking how AI could help with each of these tasks. Could it make them more efficient? Could it automate part of the process? In some cases, could it even replace certain tasks completely?

Instead of trying to integrate AI across the whole company at once, it’s much better to break the work down into smaller pieces and evaluate them one by one. Look at each task and ask how AI might support it or improve it. I think that’s the best starting point.

Please tell us about your roadmap. What is Unico working on next?

Right now, Unico is focusing heavily on hybrid-casual games. Last year, we launched both a hybrid-casual studio and a publishing arm because I believe this genre is going to grow significantly.

Hybrid casual combines the strengths of both sides of the market. You get cheaper user acquisition similar to hyper-casual games, but you also get stronger monetisation and longer engagement like traditional casual games. At the same time, these games are still faster and easier to produce because they don’t require the same level of high-end graphics as bigger casual titles. It sits somewhere in the middle, and I think that’s exactly what players and developers want.

So you get faster development, more efficient user acquisition, and stronger value per player.

Today, most of our team is working on hybrid-casual games and on the publishing side. We’re developing some titles internally, but we’re also working with external studios. At the moment, we’re collaborating with more than 20 studios across countries like Turkey, Vietnam, China, Pakistan and India.

Our role as a publisher is not just user acquisition. Competition is extremely high now, so success requires optimisation at every stage. Many developers can build great games, but struggle to optimise them for scaling.

“We started generating store assets for A/B testing. We have more than 50 games, so it requires a huge amount of store asset generation. Now we provide existing screenshots, icons and feature graphics to the AI model, and generate alternatives”
Erkay Uzun

That’s where we come in. We’ve built a backend platform where developers can integrate their games, and then we help optimise everything: gameplay balancing, live ops, and user acquisition. By combining those tools with AI, we can test, improve and scale these games much more effectively.

Our goal is to become a major publisher in the hybrid-casual space, especially working with developers across Asia.

Do you have a programming background? Are you naturally interested in the AI tech yourself, or do you leave the details to the team?

Yes, I’m from a technical background. I am a software engineer – my background is computer science. I’m really interested in AI.

I’m very involved in AI projects in the company, especially our plans to work on generating playable ads with AI.

Actually, this is a twist: we are now making another company on this, with a co-founder from Google AI!

Meanwhile, I’m reading a lot on AI, not just on the technical side, but also on the business side. I’m following all the new developers, spending at least one or two hours each day reading about AI.

Subscribe now! It’s free, and you’ll receive interviews with industry insiders and studio leaders in your inbox. Coming soon: CALM Consultancy, Meshy, DeemosTech and more.

Further down the rabbit hole

Some useful news, views and links to keep you going until next time:

• Research conducted by the University of Bristol and developer Meaning Machine shows that gamers enjoy interacting with AI-powered NPCs.

• The GDC Festival of Games took place last week, and it gave us plenty of opportunities to speak with people in the space. Google Cloud used the show to set out its vision for “living games” - autonomous AI agents embedded in both development pipelines and gameplay itself (calling it the industry’s biggest transformation since the move from 2D to 3D). AI Gamechangers spoke with Jack Buser last year in Cologne.

• Meshy announced its new AI incubator at GDC, alongside its first title from that platform, called Black Box: Infinite Arsenal. We played the 2D survivor-style shooter, in which the player can deploy weapons that are dynamically generated during gameplay, in the expo hall during the event.

• Aptoide’s Game Genie is now live. It’s an AI-powered discovery feature inside its store that lets users describe what they’re looking for in natural language rather than typing exact search terms. For select titles, it also works as an in-game companion, analysing screenshots in real time to offer tips and strategy advice (currently available in Portugal, France and Japan).

• CCP Games teamed up with blockchain platform Sui to launch the EVE Frontier × Sui Hackathon 2026, an online competition running until 31 March with an $80,000 prize pool. The challenge invites builders to create live mods for the space survival MMO. And with AI coding agents now in the mix, CEO Hilmar Veigar Pétursson reckons even a solo gamer with a big idea could take the top prize.

• AI performer Tilly Norwood has released a music video, singing about how “AI’s not the enemy, it’s the key.” Prominent US cultural historian and critic Ted Goia described it as “cringe on steroids”.