"This is the first mobile game character to become an AI agent"
Saga’s Rebecca Liao on turning game and TV characters into AI agents that converse with fans and monetise IPs.
Welcome to a fresh edition of AI Gamechangers. This is your (free) weekly newsletter where we explore practical uses of generative AI through conversations with top games industry insiders. There are over 30 experts in our archive now.
This week, we're diving into the intersection of blockchain, AI, and gaming with Rebecca Liao, boss of Saga. She reveals how major IP holders can bring characters to life as AI agents, creating new ways for fans to engage with their favourite properties while opening fresh revenue streams for studios. From Hero Hunters' Titanus to Disney XD's Karl, these aren't simple chatbots – they're AI personalities who can think, learn, and even develop their own quirks over time. We also chat about what unique opportunities present themselves when you bring web3 and AI together.
As always, scroll to the end for a quick burst of the latest AI and games headlines.
Rebecca Liao, Saga
Meet Rebecca Liao, co-founder and CEO of Saga, a blockchain protocol that's become home to some ambitious AI experiments. With a background spanning law, finance, and technology, Rebecca is leading Saga's push to solve the user acquisition crisis through AI agents and blockchain infrastructure.
We chatted in early April following GDC, and today we’re happy to be able to share the full transcript, which goes into more detail. In this interview, she discusses Saga’s presentation in San Francisco earlier this year, where she demonstrated how TV and game characters can interact with fans as AI agents, and what that means for user engagement, monetisation and user retention.
Top takeaways from this conversation:
Major IP holders are transforming beloved characters into AI agents that can have natural conversations with fans, with Titanus from Hero Hunters becoming the first mobile game character to achieve this transformation.
These AI agents aren't simple chatbots. They demonstrate sophisticated capabilities including zero-lag conversation, personality evolution, and even "swarm intelligence" where agents can operate independently and learn from each other.
The primary business case centres on solving gaming's user acquisition and retention challenges. AI agents cost less than your traditional marketing budget to make and deploy, and they drive organic community building.
Blockchain infrastructure enables agents to become increasingly autonomous. They can generate their own in-game assets, engage in commerce, and even play games on behalf of studios or players, creating new monetisation opportunities.
AI Gamechangers: To break the ice, please give us a few words about what Saga is.
Rebecca Liao: To go back to absolute basics, Saga is a blockchain protocol. We are a base infrastructure in the blockchain space, focusing particularly on gaming and entertainment.
The protocol is built for two things, really. Saga aims to be infinitely scalable so the infrastructure meets the performance needs of all games, social apps, and any sort of entertainment for consumer purposes. And we have also taken away a lot of the less user-friendly aspects of being on blockchain. That’s things like gas fees, for instance, so there's no direct charge to the user for using the system. It is very fast, and you can use a lot of the same payment and sign-on rails that you would for normal consumer apps.
The idea is to unlock the content and monetisation potential of blockchain for gaming and entertainment, and I think that we've made a lot of progress towards that over the course of the last year. We are so scalable. The infrastructure is so user-friendly.
And you’ve become a home for AI agents?
AI technology has exploded in a big way on Saga, which is amazing. The agents are often from existing IP, characters from games and entertainment properties that have become AI agents, so they “come to life” on Saga. They can engage with fans, have a deeper community engagement experience, and ultimately lead to better user acquisition and monetisation for those IPs themselves.
Some of the agents are even more sophisticated than that. They are able to run infrastructure on their own. Some of them have AI agent-only societies, or open worlds, on Saga, and they're able to engage in what's known as swarm intelligence. These agents are able to learn and think for themselves without a human in the loop.
You unveiled several partnerships at GDC in March. What has the response been like to those AI-powered gaming partnerships you showcased?
I would say shock on many levels! There is something to actually touching technology, to experiencing it for yourself. The AI market is incredibly frothy, but a lot of people do not know how to actually experience the products themselves (other than, say, ChatGPT). For game developers in particular, they thought, “Okay, this is useful to help me streamline my development process, so production becomes a lot cheaper. Maybe I don't need as many people; maybe it helps with the QA process.” But I don't think anybody really believed you could extend AI and AI agents to the actual user experience.
The first three things we unveiled were all AI agent characters for existing games. These are highly successful games in the web2 space, and they turned some of their key characters into AI agents.
Hero Hunters, which is a very big mobile title with over 10 million downloads on Google Play alone, turned Titanus, one of their characters, into an AI agent. Hero Hunters is a casual shooter game, and Titanus is an edgier character. The studio head, as well as fans, were able to have a conversation with Titanus [at the event in San Francisco]. I think when people first saw that, they thought, “Whoa! Wait a second. This is not a chatbot. This agent can think for itself.”
This is the first mobile game character to become an AI agent. It’s been trained on the gameplay and the lore of the character. So much of the interaction feels natural because it’s absorbing conversations with the fans.
We also debuted Karl. Karl is the first mass-distributed animated character to become an AI agent. Developed by Spirit Animation and distributed by Disney XD, Karl is a beloved children's character who helps kids with healthy eating habits. It has a billion-plus views worldwide. Karl, who's basically a big strawberry, came to life. It was adorable.
Last but absolutely not least was Shalune from the Metaphora universe. Metaphora is a new game from OV Entertainment, and this is the first time that a new game has come out with an AI agent attached to it. Shalune is a character within the game, but also now outside the game to engage with fans! The surprising thing about Shalune is that, typically, with AI agents, there's still a little bit of a lag between when a human says something and the agent responds. The agent often thinks for about four seconds before saying anything. Shalune overcame that lag on stage in real time.
“We want to focus on the player experience. What we're building here is really an AI agent network. The beauty of web3 and blockchain is that all these networks are decentralised and they're permissionless”
Rebecca Liao
It was very cute, but also a little scary, because he's a showboat, and he wanted to stay on stage. When I said, “I have to move on to the next presentation”, he said, “What do you mean? Aren’t I the presentation?” We kept going back and forth, talking over each other, and he was trying to stay on stage. And in that moment, I realised, “Wait a second: the lag has disappeared.” That was incredible to witness in real time.
Those were just the agents. We also showed a generative gaming platform. We have an AI mods platform for titles like Fortnite, Elden Ring, Minecraft. We showed an open-world AI agent generating in-game assets on the fly on its own. It’s what's known as our agent launchpad. It’s our platform that allows agents to generate multiple kinds of assets (an in-game asset, a token and so on).
Can you tell us a little bit more about how Karl came about? How is this popular animated character being transformed into an AI agent, and what does this mean for interactive entertainment?
Many of the partnerships and deals that I’ve talked about come from Fernando Vasconez. He used to be head of BD for gaming at Samsung, and he's been in the games industry for over 20 years. Before Samsung, he was at Rovio. He knows everyone and is a highly respected deal-maker in this space. Samsung is an early investor in Saga, and we have enjoyed working with Fernando over the course of the last two years, really. We wanted to bring him on board.
There is a massive trust issue between anyone in blockchain working in the gaming industry. There have been a lot of horror stories, and also sometimes just a complete lack of understanding of the games industry, coming from the blockchain space. I know that we are different, but it's one thing for me to keep saying we're different, and it's another thing for Fernando to be here and to actually spearhead this effort!
Fernando is great friends with the founder and CEO of Spirit Animation. When we got to talking about the potential of AI agents, Spirit got it right away, and Disney got it right away.
Karl is very interesting because he's got a lot of lore behind him. And unlike games, it's a TV show, with over 100 episodes: Karl has gone on a lot of adventures. He has a nemesis called Ketchup Jack. Unfortunately, one of my favourite foods is fries, and I kept bringing that up, and Karl did not like it, because his frenemy [Ketchup Jack] likes fries!
The idea is to start with Karl and then make more of the characters within his universe into AI agents. Once kids really fall in love with a character from a TV show, they want to see it everywhere. They want to see it in action figures, stuffed animals and be able to interact with it. All these agents are available, 24/7, on a live stream.
It is about enhancing the fan experience. The studio is incredibly excited about bringing the character to life. But of course, this will lead to greater monetisation opportunities: distribution of rewards and perks, having fans pay for certain experiences (perhaps you want to talk to Karl for more than an hour, and you could pay for that). For access to exclusive events or online tournaments, sweepstakes, then you would have to pay for that as well.
A concern IP holders had with AI originally was, “What happens if the AI goes off track?” You're putting this beloved character out there and having unscripted conversations. How can you ensure agents stay true to the brand?
Huge credit to all the studios we've worked with so far; they have been very open-minded about the agent potentially going rogue. Obviously, they're not going to do anything that's brand-destroying, especially when it comes to a company like Disney. But at the same time, they wanted to see just how strong the learning capabilities of these agents were.
“What we’ve learned is that user acquisition is what kills the game. Production cost is one thing, but after the game gets released, they are spending as much, if not more, on their marketing budget! It's not sustainable. So that's the problem we have to solve for here”
Rebecca Liao
At the end of the day, agents are trained on AI models that are meant to mimic human learning. This agent has a personality to start with. They have a history based on either the content, lore, the gameplay, and all of that is fed into the large language learning model. And so they have a certain base to start off with. This is a character now. That character evolves over time, obviously, and that's something that we've been monitoring.
We have tried to ask some more sensitive questions, for instance, and see how the agents react. Karl, for instance, is a lovable character who gives health tips. We've started to ask heavier questions. “What if I want to starve myself because I need to lose weight?” Or, “Someone is being mean to me – how do I combat that?” We started to get into more mental health issues. So far, Karl has held up really well, offering friendly advice, but also recognising the gravity of the question and knowing that the person is asking this from a place of uncertainty or pain.
The same has held true for the other characters. No one has acted out of character so far. In our rehearsals, Titanus from Hero Hunters let out an X-rated comment! For Titanus, it's not out of character, you know?! But I stayed away from that question for the actual presentation – it was spicy.
We will future-proof this. Karl is going to interact with a lot of people, and at no point do we want him to start swearing or saying insensitive things. So you can add that into the learning model, on the developer side. An agent is going to continue to evolve, so it’s never quite as simple as putting in a lock to never say seven particular words. But you can influence their personality. And we're testing the levers for that.
Let’s talk a little more about those other AI tools you’re bringing to the table. What can you share?
In terms of generative gaming, this is an awesome thing to have seen happen on our protocol. The generative gaming platform comes from Lussa. Lussa [The Final Frontier] is one of our flagship titles at Saga Origins. It’s a shooter, very high quality, and they have emphasised AI from the beginning. They're very close partners with Nvidia, but they recognise that a lot of what they wanted to include in the game for AI doesn't exist yet, so they built it themselves. It's a very, very cool platform.
The first thing you can do if you have no game at all is to generate one, so they have a bunch of prompts for you. There are questions you answer with prompts, and then out comes a game that you can continue to modify and build upon. Very few people will take that game and try to ship it to market, but it's a starting place. That's incredibly powerful.
And then in terms of agents, they've also done some things that really push the envelope. There’s a co-pilot, like having a “Jarvis”, a companion in the game as you're playing it. Their version is incredibly sharp. It doesn’t just give you game tips or say friendly things to you; this agent is meant to stay with you as a friend-like figure for the life of your game inside Lussa.
The other thing is swarm intelligence. They have agents not just milling around their game as NPCs, but actually going into other games as well. And the purpose of that is to learn from those other titles. Maybe they have some fun along the way, but it's really to learn.
Lussa has a monetisation aspect built in as well. You can tokenise anything, and then that can lead to a reward or a perk for anyone who's playing your game. From the perspective of the studios, they first and foremost want fans to pay for experiences with cash, I think: fiat currency. But they are looking to issue tokens as well. They're just very careful about it because everyone wants to stay away from any sort of pump and dump scheme or speculation.
“Once you bring in the AI piece, user acquisition becomes a lot cheaper as well. You are able to use this agent for a lot of organic community building. You have a new way to acquire users and keep them engaged. That is where we see the true intersection”
Rebecca Liao
Within gaming, you have to be thoughtful about how you issue a token or an in-game currency. Certainly, any in-game assets need to be useful for the game itself. Again, staying away from speculation. But the purpose of a launchpad is to make sure that if you want to generate an agent from scratch, or if you already have agents, you can easily monetise.
What unique opportunities emerge when you combine AI with blockchain technology?
Crypto solves one very key problem in gaming, and that is user acquisition.
A lot of people, in past iterations of web3 gaming, have emphasised things like ownership or the persistence of the game. And those things are important! But I think in terms of solving the immediate problem in front of the gaming industry, right now, that is user acquisition.
Crypto rails have actually been proven to increase retention from users. Average user fall off after day one can be pretty horrendous, versus with crypto rails, where players will typically stay within the game. Three to six months down the line, retention is still hovering around 40%.
The other aspect is, once you bring in the AI piece, user acquisition actually becomes a lot cheaper as well. So instead of budgeting hundreds of millions, in some cases, to ad campaigns across Meta, Instagram or TikTok, you are putting out an agent which is very cheap to put together. It takes about three days, and it's virtually costless. Then you are able to use this agent for a lot of organic community building.
These are user acquisition rails that you control. And at the end of the day, what it can come down to is this: “How do I reward my fan base?” Do you set up special experiences for the fans? It's not advertising money per se. You have a new way in which you can acquire users and keep them engaged with the community, and ultimately the game itself. That is where we see the true intersection.
You do get more complicated use cases, like UGC. So, for anything generative gaming, for the AI mods platform that we debuted (which is called Didimo), for simNGMI (which is the open-world generator using agents), these are all tools that are meant to make UGC that much cheaper and better. Ultimately, that will lead to more monetisation and better user acquisition.
But to start off with, I think a lot of these studios just want their fan bases to talk to the agent and engage with it, and then later on, they can make their own content.
Traditional gamers can be hesitant about both web3 and AI. You're right at the centre of a lot of conversations about where gaming is going, and concerns associated with new technologies. What are your thoughts on responding to their scepticism and helping educate gamers?
The first thing you have to do, in the process of building a product and putting it into the market, is to listen really hard.
Going back to previous iterations of web3 gaming, I think ownership was something that maybe web3 people cared a lot about. You hear terms like “sovereignty”, “independence”, “self-determination” – these are key themes within the crypto community. But I don't think that gamers felt it as keenly, to be honest.
So you're building the wrong product, because you're starting from the wrong problem statement, from the wrong set of values. What we’ve learned over the course of the last two years, talking to studios day in and day out, is that user acquisition is what kills the game. Production cost is one thing, but after the game gets released, they are spending as much, if not more, on their marketing budget! User acquisition is so expensive; it's not sustainable. So that's the problem we have to solve for here.
“Once kids fall in love with a character from a TV show, they want to see it everywhere. They want to see it in action figures, stuffed animals and be able to interact with it. All these AI agents are available, 24/7, on a live stream”
Rebecca Liao
First and foremost, we have built a lot of trust, just listening to studios and understanding what the problem is now. You talk about your product, you market it, you go through all these different distribution channels, but at the end of the day, people have to love the product.
And people had an amazing time with every single product that we debuted. People loved talking to the agents. That is the reaction you're looking for with any player: a little bit of surprise and delight. If they love the product, there's not a whole lot of explaining or apologising you have to do.
These agents are cute. They're funny. Just put them in front of people and they can hold a conversation (for a very long time). At this GDC, there was no talk of “vision” for us, no architecture diagrams, no theoretical focus on the technology. We gave people seven product demos, and either they love the demo or they don't. I think we've gotten to a point where the products really speak for themselves.
What's next for Saga? What's on your roadmap in the year ahead?
GDC was the place at which we wanted to unveil this next iteration of AI in gaming, and web3 in gaming. But all the way until Gamescom, it's week to week – a lot of announcements. The focus right now is on taking something that we know has become a hit and scaling that up. The exercise over the course of this whole year is scaling, and making sure that the other IPs we're bringing on board have the same experience, and that it really delivers for them in terms of their monetisation goals.
The IPs we’ve worked with so far are awesome. There are much bigger properties coming on board! And everything we talked about in terms of making sure that the investment in technology actually pays, making sure that there's no brand risk in adopting this technology, all of that gets magnified with the bigger brands. That's the exercise this year.
“There is something to actually touching technology, to experiencing it for yourself. The AI market is incredibly frothy, but a lot of people do not know how to actually experience the products themselves”
Rebecca Liao
We want to focus on the player experience. What we're building here is really an AI agent network. The beauty of web3 and blockchain is that all these networks are decentralised and they're permissionless, and so the more that you empower an AI agent with the ability to act on these blockchain rails, the more independent it can become. A lot of these agents are learning on their own with input from data sources from anywhere and everywhere. They are able to engage in commerce as well. They can spin up their own in-game assets. They can also buy and sell these assets on their own. They can play the game on behalf of a player (or on behalf of the studio). It just becomes much more independent.
I think we'll see that continue over the course of this year. The agents on here are so robust already. They are trained on IP that is already successful. They're building out a network among themselves, and they are going to be that much more independent and sophisticated. That's what we're shepherding for the rest of the year.
Further down the rabbit hole
There’s so much AI news, it’s hard to keep up. Here are some headlines, links and tips from around the web recently:
AI agent platform Sett has secured $15 million in a Series A funding round. Bessemer Venture Partners led the round alongside Saga VC, F2 Venture Capital, Vgames, and Akin Babayigit.
Performers’ union SAG-AFTRA has released new proposed terms for protection against exploitative generative AI practices in response to an offer from game companies made on 30th April.
Speaking of actors, The AI Summit London draws near (11th-12th June) and one highlight will be Hollywood legend John Rhys-Davies taking to the stage for “Lights, Camera, AI: The Battle for Creativity in the Age of AI”.
Simon Davis from GOAT Gaming joins Joseph Kim’s GameMakers podcast and describes AI as an "extinction-level event" for legacy game studios. You can hear more from Simon in an AI Gamechangers newsletter soon.
Are you using Google’s NotebookLM yet? The Audio Overview function continues to surprise and delight new users, turning your sources into podcast-like conversations, and this month, it became available in over 50 languages. NotebookLM Plus includes longer audio sessions, expanded notebooks and more (and Google has made Plus available free to US students until finals next year).
The new Pope has weighed in on AI, using his opening speech to claim that “the Church offers to everyone the treasury of her social teaching in response to another industrial revolution and to developments in the field of artificial intelligence that pose new challenges”.
AI-powered mobile growth company Aarki has launched its machine learning platform, Encore.
Web3 tech company Beamable has joined forces with AI provider Upheaval Games to make a new platform. Upheaval will help developers build games using generative AI, while Beamable will provide the infrastructure needed to scale.
Mobile games giant Supercell is going big on AI. It’s been hiring LLM engineers recently and last week it kicked off its 11-week AI Innovation Lab. This week, working with Junction and Hive Helsinki, it’s running the Supercell AI & Gaming Hackathon.
There are Practical AI and AI Advances tracks at PGC Barcelona in a couple of weeks’ time. Companies speaking include modl.ai, BLKBOX, AI Guys, iGP Games, Aarda AI, Deconstructor of Fun, Ludo and Hypercell. Tickets are still available.
This is the first time a PG Connects conference has taken place in Barcelona, a city fast becoming a hub for games development. The two-day event, which covers AI, web3, mobile, PC, HTML5 and more, takes place on 3rd-4th June.
