Rob Alvarez Bucholska
0

AI + Game Thinking = Retention Breakthrough | Episode 397

 

Click the link below to book a quick 10–15 minute chat with me.

No sales, no pressure—just ideas and clarity on how to engage your community and grow revenue.
🗓 professorgame.com/discoverycall

Strategy Meets AI
Game design + AI = game changer. In this episode, Amy Jo Kim returns to share how she’s combining decades of design wisdom with cutting-edge AI to build smarter, faster onboarding and retention systems. If you’re struggling with churn, this conversation is your blueprint to better systems, better questions, and better engagement.

Previous episodes with Amy Jo Kim:

 

Rob is a host and consultant at Professor Game as well as an expert, international speaker and advocate for the use of gamification and games-based solutions, especially in education and learning. He’s also a professor and workshop facilitator for the topics of the podcast and LEGO SERIOUS PLAY (LSP) for top higher education institutions that include EFMD, IE Business School and EBS among others in Europe, America and Asia.

 

Links to episode mentions:

 

Lets’s do stuff together!

Looking forward to reading or hearing from you,

Rob

 

Full episode transcription (AI Generated)

Rob (00:00)
What do you mean you’re interviewing people who are not people and getting results from that? Does that work? So yes, in this episode, we’re gonna be talking about exciting things and exciting ways in which AI is revolutionizing the way that we think about how games can improve the way we do other things. Among them, of course, retention, loyalty, engagement, and so on and so forth. But before we get started with the interview with Amy, make sure that if you’re interested in getting more insights into the kind of things that we discussed here, be it AI, onboarding, retention,

engagement and all these exciting topics that we get into on the podcast. Make sure you go on the link in the description so you can find all those free resources that we can get for you. So, Engagers, welcome back to another episode of the Professor Game Podcast. We have an exciting interview today. We have Amy Jo Kim back on the Professor Game Podcast. Amy, are you prepared to engage once more? Let’s do this. And if you want to know, you know,

Rob (00:55)
the origin story, the fails, everything you know we talk about on the podcast. You can definitely go to the show notes. You’ll find it there or you go to ProfessorGame.com. Just type in Amy Jo Kim. You’ll definitely find all of that, all the past episodes also. However, today we’re going to be speaking about a trending topic, which is AI, but especially in particular how Amy Jo Kim and her team have actually created a specific tool for the type of work that she does.

that uses AI to enhance all sorts of fantastic, phenomenal things. So, Amy, without giving away anything on my side, how is it that AI and game thinking are coming together? Like, what’s that look like? Very briefly, and we can dive in all kinds of directions.

Amy Jo Kim (01:37)
Absolutely.

Well, let me tell you how it came about. So about a year and a half ago, I was talking with one of my former clients who is a full stack engineer in India. We were talking about what was going on in AI, comparing notes, seeing all of the developments. And he said to me, you know, I really love your material. I love your framework. We used it on our startup. It helped us so much. It helped us do a major pivot, go in the right direction.

find our audience. It helped us with all of that, but it was really hard. He said, you know, I’m just as a, he said, as a technical founder, it was just really hard to learn how to do the research and then the storyboarding and what story beats are. And my brain loved it, but implementing it was so much work. Well, perhaps you can relate, right?

Rob (02:29)
There’s a lot of things to do and I always like to say when people are talking about you know these these hacks and these things to do things quick I always sound like they’re all fine, but nothing beats actual strategy in place and strategy takes work Don’t get me wrong. You can be more productive and I’m sure you know you can make the best of of tools like what Amy Jo Kim is using but don’t come in thinking that the fact that we’re using AI now to do this makes it such that there’s no hard work, especially hard thinking

and a lot of stuff you have to do. But that’s my message. Maybe that has changed massively with what you’re working on.

Amy Jo Kim (03:03)
Absolutely. That’s a timeless message. Strategy, insight, talking to real customers, putting concrete things in front of them they can react to, whether it’s storyboards or a prototype or even a full blown app that you’ve vibe coded. Those are things that there’s no substitute for. Those are rubber hits the road moments. The punchline is that what I’ve learned about what AI can do, and I’ll take you through how we learned it, is…

It can help you prepare to do all of that much more quickly. It can help you frame smart questions to ask as you go do your research. It can help you not have to execute the really grungy hard work that requires storyboarding skill or UX design skill. So it can help you do those massively faster. AI like our tool and like lovable and

and all of it puts the power to turn your ideas into something concrete and visual into the hands of people that just have ideas and can prompt. Product managers, founders, technical founders. So back to my technical founder, his name is Piyush Mahajan. So Piyush was saying to me, I love your stuff, but it’s just too damn hard, takes too damn long. You know, no offense, but I’m thinking that AI could really…

tackle this, go, God, I’m thinking the same thing too, but there’s so many different things we could do. We could do a screener builder because building a really good screener to find exactly the right people that locate your ICP, your ideal customer profile. That’s actually quite hard. And it’s something I do, you know, dozens of times every year for clients and with clients. We could do that.

Or we could tackle the storyboarding piece or, we could tackle some of the research piece and there’s so many different ways AI could work. So Piyush says to me, well, I’ll tell you what my dream is. My dream is that we could just get you to answer a few questions about like your elevator pitch, who your customer is, what you’re building, you know, a few things. And then somehow we could just magically make great storyboard.

And they would just come out the other side. That’s what I would like, just automate the whole thing. And I said, you’re nuts. I’m so against that. And I don’t even know if we should work together, Piyush. And he’s like, nah, that’s just, I’m open. I’m open, but honestly, for when I look at Aya and I think about where we’re going, that’s kind of what I want to do. I said, great. Where do we start? What’s our MVP? know? So Piyush and I partnered along with Scott Kim, my partner, who was our

prompt master, I’ll tell you about that in a minute. And he’s an amazing full stack engineer who’s very up on AI and who’s got clients. He’s supporting himself by doing AI work for clients. So he’s a great person for me to partner with. I’m a UX designer, I’m a strategist, I know game thinking, and I work with, you know, dozens of clients every year, right? So that was the partnership is what could we do? So we set about

Building modules and models. One of the things we realized is that there’s a lot that you can just go and do with Claude or ChatGPT. Whatever we do has to be different and differentiated and better, which is a key thing about product strategy. And you can do quite a lot with Claude. mean, our team uses it every day at this point, but we say, okay, what can we do? Well, one of the things we have is data. We have.

decades of data from client projects on what a good screener looks like, what storyboards look like, how you write onboarding story beats, how you would create an effective habit loop that brings people back, how you extract job stories from customer research. We have decades of data. So the first thing we did was we went through all our data. We got permissions from everyone we had worked with to use it as training data.

key issue. And then we started building our own models that we could use based on, you know, open AI and trained it on our data to see what we could come up with. The first module we built is a screener builder, where you answer a few questions and it builds you a six question screener that will help you find people that are actively looking for a solution like yours, which is the most important thing. Then we built another module.

that would take the output of that, create job stories and habit stories, which are job stories written around habits, just from that data using an agentic workflow. Do you know what agents are? It’s really booming right now in AI.

Rob (08:04)
I think I have a clear idea, but I’m not sure how many of those listeners are all of the dates on these things. So give us give us a gist.

Amy Jo Kim (08:12)
So

an oversimplified, but really useful way to think about agents is that they have a very particular purpose. They are in essence role playing. think a lot of our listeners probably know about role playing because we’re speaking to gamers and people that love games. So when you see prompts like you are an expert marketer with experience in large CBG companies, do this, give me feedback as that role. That’s role playing.

in LLMs. Agents are a form of this. Agents are a chunk of software that plays a particular role and can go execute activities. One of the big developments is that LLMs aren’t just, you know, back and forth with a human. You can have them go interact with other agents. They can make stuff happen on your screen. You can integrate them with your data, like in

Google or Facebook, you can give them access, et cetera. And they can then go execute on tasks in a vaguely similar way as a human would. And that’s essentially an agent. Where it gets interesting is when you have multiple agents working in synchrony, each with their specialization. And that’s what we did. So our first module was a trained model that could spit out

a pretty good 70, 75 % of the way there super fan screener or ICP screener that you can just take and use in your research. The second module built seven different agents, all working in synchrony under the hood, starting with an agent that would take everything that we had defined about our ICP and our market niche leading up to the screener, take that and create

synthetic personas out of that, multiple ones that fit within the area, and then use the screener and interviewed the synthetic personas and then looked for patterns across those interviews that were common in more than one, and then took those patterns and turned them into job stories.

Rob (10:26)
Okay, okay, okay. So pause there for a second because that super, I think I followed it. The only thing that sounds, you know, both interesting and maybe a little bit scary as well is what do mean you’re interviewing people who are not people and getting results from that? What is like, does that work? Like, do they give you something interesting? Is it just a robot answer? Like, what’s that?

Amy Jo Kim (10:51)
So, back when Piyush said to me, I want to automate everything, he also talked to me about synthetic personas because there’s a very controversial company called Synthetic Personas that made a big deal about it, you know, to get PR. And they now offer this as a service. So when Piyush first said that to me, I was like, absolutely not. I’m against it. That’s so bad. You’re going to just get crap research. You won’t get anything innovative. They can’t give you anything innovative. Guess what? All of that’s still true.

But I’ve learned since then that there are multiple good uses that help product designers and founders and marketers and people that want to build great experiences. There are multiple uses of synthetic personas that help you. So a synthetic persona, there’s different words for it, but let me just frame it and describe it. It is a software entity.

that you can interact with either through an API like we’re doing in our product or through an interface like you might with ChatGPT, a chat interface. And the interface really doesn’t matter. What matters is the synthetic persona. So a synthetic persona is essentially a bot that’s role playing. Just like I talked about, you could role play as an expert marketer and tell ChatGPT to help you with that. Or you could role play as something else. Well, this bot is role playing as

A particular person who fits right into who you’re targeting that has certain characteristics and the way you define those characteristics and the data that you train that synthetic persona on will be the determinants of the quality of that persona. But all that matters is what you want to use it for. Right? So let me give you some examples. Let’s say that you want to build a synthetic persona. Well, you’ll need some basic demographic information.

Because whether the person’s in Nigeria, Sweden, or Maine is going to make a big difference, or Barcelona. Like that’s the location’s going to matter. Their age is going to matter. Their gender is going to matter. What they do for a living might matter. What, you know, their family situation, their living situation, the stressors that are on them, where they are in their career. You can go deep with demographic information and also psychographic information.

For those who might not know, demographics are like verifiable facts about you, know, age, sex, marital status, work status, et cetera. Psychographics have to do with your tastes, your interests, your triggers, the groups you identify with, things like that. And in gaming, psychographics are heavily used because you about casual gaming, it’s eight to 80. So how do you deal with that demographically? You don’t.

You talk about a psychographic of the certain type of person who like loves to solve word puzzles or that sort of thing. That’s a good example of a psychographic. You know, they also might love other things in their life that are similar to maybe they also love reading. Someone who loves word puzzles, for instance. So those are, those are all things that really good marketers and certainly some founders will try and specify about their audience.

The problem with using personas like that without them being digital or synthetic is that companies get really locked into like Abby 28 works in finance. And then as the company evolves, they, those personas get locked in stone. So there’s a lot of problems that I have dealt with up close and personal and painfully with personas in rapidly growing product and game situations. So that said.

It’s not like personas just work all the time, but what personas are is they’re very useful and they’re used really widely to help a team get specific about who are we targeting. It’s not just, is for everybody. This is for Debbie, 27 in Des Moines, who’s, you know, junior finance professional and dating, but not married and blah, blah, blah. So that’s a persona. Now let’s say that you’ve collected a bunch of data about.

your target audience and you’ve done a bunch of interviewing of real people. Okay. This is something I do a ton. And so you don’t just have Debbie in Des Moines. You also have, you know, Mallory in California, you know, who’s trying to save up. Like I was working with a FinTech app and we had a millennial female target. So we talked to probably 15 to 20 people within that target and we had all these interviews and they really fit our target. There’s an example. Let’s say you have that situation.

What you can do now is you can upload all of that information about those people, all the demographics and psychographics, plus interviews with real people answering questions. You can upload it and create a bot, create a synthetic persona that represents how that person would be likely.

to answer in certain situations or the job stories that they might be encountering in their life if they were to use product X, which is what we do. So let me now crystallize that in a simple way. Synthetic personas are a representation in a form that you can interact with as a product creator or an experience creator that can help you think through the impact of what you’re doing on your audience.

without needing to go and talk to specific people. Does it replace talking to people and showing concrete artifacts to people to find out if it’s what they want or not? No, it doesn’t replace that. When you try to use it to replace that, you get into trouble because of the way LLMs work. They can’t be novel. And I don’t know how much research you’ve done or hello audience people, how much research you’ve done. I’ve done just.

probably at this point, thousands of interviews, just so much research of ⁓ different clusters of people. And when you do that in a product setting, you learn novel things that you could never get out of synthetic personas, and you could never get out of just thinking about it with your smart friends in a room with a whiteboard. And know, when it’s really ideas hitting reality and there’s no substitute. But what you can get with AI is all the prep that goes up to that moment. You can have much better hypotheses.

which you then test against real world people and issues. Does that make sense?

Rob (17:26)
Absolutely, absolutely. It helps you accelerate a part of the process to get somewhere that you want to get.

We’ve been… Go ahead, go ahead.

Amy Jo Kim (17:35)
Right.

helps you ⁓ not need certain skills that are very tactical versus strategic. It takes care. I’m a long time UX designer. I made my living by doing that for decades. I don’t do it all the time now, but I have decades of experience and expertise. Right. I love it. I go into flow states when I’m doing UX design. Guess what?

I don’t have to do it anymore. And part of me is threatened, right? my God. I can actually, if I can describe it in a lot of detail, I can get something in a few minutes out of Claude. It’ll give me like 10 sketches. I just have to have a good, really good description. The really good description is where the strategy is. That’s what, that’s what LLMs, like they’re a, you have to figure out what part of the process to use them for and which tasks. But I just want to emphasize what you said at the very beginning of this interview.

There’s no replacement for strategy and judgment.

Rob (18:34)
Absolutely. I love that. Like you gave us an explanation of many of the things that it’s doing, but I feel like it has a little bit more of the the back end of what’s happening behind it, which is amazing. Like I love understanding all of these things. And I’m sure the audience is appreciating a lot of this because this is where you get actually go deeper into learning and understanding things to be able to apply them yourselves. how does a sort of a flow of one of these users of this journey maker look like? Like I come in.

And what am I going to be doing? What am I going to get out of that? And how does that flow look like? Can you guide us through a hypothetical or a real example?

Amy Jo Kim (19:10)
Absolutely. Amazingly, our current flow is very close to what Piyush envisioned a year and a half ago that I was dead set against. Let me just, and one thing about me is I’m willing to change my mind if new data comes in. In fact, eager. I mean, that’s really what the scientific mindset’s all about is make hypotheses, gather data, change your mind if the data tells you to. So this is how it goes. You come in, let’s say you want to start a new project.

There’s two situations, you can start a new project or you can tweak and edit and iterate on an existing project. So you come in, start a new project. The system’s going to ask you a few questions. What are you building? Who are you building it for? Tell us about your three to five year TAM market and now tell us about your initial much more narrow specific niche market. it guides you through. Yes. So one of the most important things for building.

Rob (20:02)
Total addressable market just in case.

Amy Jo Kim (20:08)
Anything useful with AI, but also as a product leader of any kind as understanding the difference between who you’re going to target in the next few months and who you’re going to target in the next few years, especially at the beginning of a project. So we make sure we’ve, we tried to build it without this. We tried to just say, who’s your audience? What are you building? We could not get good results when we actually got people to differentiate and say, no, tell me about your TAM.

Total addressable market. The thing that you put on a slide that gets the VCs excited, your tail. Now let’s get specific. Who are the first 20 people who need this yesterday, right? If you’re building a community of any kind, same issue. So we ask you about that. We ask you to describe your product, who it’s for, what they expect to get out of it. And then we ask you a few questions about your recruiting demographics, location, age range, and the traits.

that differentiate them, that make them ready to use your product, what like they’re looking for a solution. And then you push a button. It gives you a screener, which you can edit and tweak or regenerate if you want to get more choices. And then you continue. It shows you job stories and the habit stories that your target customers might be likely to have. You get a chance to look at those, say great, or tweak them.

if you have more information, then you create and it, I mean, it does that automatically from the using that agentic workflow that I started to describe. I’m not going to go into the whole backend. We’re preparing a video for any, you know, crazy nerds like me who really want to get into that stuff and learn about agentic workflows and why it’s so interesting. We’re preparing a video. So that’ll be up on our channel soon, but from the front end, you might be, you know,

five minutes into this project at this point. Okay. You’ve got, you know, you’ve written a screener, you could run, you’ve used that data to create hypothetical job stories, meaning needs and habits and patterns you might see in your customer. It’s a, maybe it would look like this. You can look at it, you can tweak it. Then you answer a few questions about who you want to create a persona, you know, what gender.

what age, and you click a button and the system creates your persona in a black and white sketch. gives you a choice of who you want them to be. It gives you a visual for the personas that we’ve created under the sheets already. It then places that persona in a setting in their real life with an access device and fills in hypothetical story beats for you.

For discovery, onboarding, habit building, and mastery, the four stages of the player journey, it fills all that out and illustrates it with a picture of your persona in the setting, whether it’s an office or a coffee shop or their kitchen or looking at their phone in bed at night. Shows them there. And it also shows you sketches on the access device, which might be a phone or a laptop or a game console. It shows you sketches of what

the product might look like as well. So it fully fills out the first draft of end-to-end storyboards showing your customer using a product along with the interface of that, what that product is over time.

Rob (23:40)
Lots of it packed into that.

Amy Jo Kim (23:42)
Insane. Yeah. So, and at each step you can tweak, right? What we’re doing, let me tell you what we’re doing in a nutshell. This is very close to being instant end-to-end storyboards. You answer a few questions, tweak a few things, boom. You can then iterate from there. What we’re doing is making it radically easier for someone who’s got product ideas or experience ideas. Or let’s say that you want to build and launch

a community and you want to use gamification in your community and you want it to really work for retention. This is something you probably run into a lot. And often simple gamification backfires and it doesn’t work. Let’s say you’re in that situation and you want to untangle it. You can describe a few things and you can tell it more about your product and then click this button and it’ll give you storyboards of ideas. It gives you a whole bunch of ideas of something that might work.

that’s been trained on decades of real data for projects that shipped and worked in the marketplace.

Rob (24:50)
I want to highlight that it’s real and private because it’s not the same thing to ask chat GPT about something that is all that’s available on the internet and so on and so forth. Then to ask something that is from the information of somebody like Amy Jo Kim has been at it for decades as she’s mentioned several times and has input all this data in there. It might not be as massive as the whole internet but it’s a lot better focused, right? So you’re using both all that information out there in the internet which is already available through these LLMs.

plus a narrowing down of the specific audiences that Amy, Jo Kim has been working on. Amy, this has been fantastic, very illuminating. think at least a few people in the audience are now very excited about what AI can do for them, what even JourneyMaker in particular can actually help them build, validate, and so on. So in a nutshell.

Can you tell us how this both integrates with your game thinking framework, which I’ve gotten the clues. Of course, I have, you know, I’m a certified coach and so on and so forth. But for somebody who’s never heard of game thinking in a nutshell, how does it integrate to your game thinking and, you know, any further insights you want to give us before we take off?

Amy Jo Kim (26:00)
Great question. The way that this integrates with game thinking is it lets, first of all, anyone who’s familiar with game thinking, who’s read the book or learned from you will find it much, much easier to execute the recommendations, execute what’s taught in game thinking. The screening process is made faster and easier. The storyboarding process, as I described, is faster and easier.

The way that it integrates with it one step down is in game thinking, there’s a core design framework, which is a mastery path with a habit loop embedded in it. And those two things go together. This is one of the hardest things for people to wrap their mind around, especially people that don’t think in systems. But that model is embedded into this tool. one of the things.

that we do is you define whether you’re doing it through synthetic data or through your own data, because you can also upload all your own research data, surveys, interviews. You can upload it to the tool and the tool can use that as additional data. Just like you can upload stuff to ChatGPT and Claude and say, use this as additional. You can do that with your own research. So you can make it even better. It is the tool, whether it’s.

Synthetically with what do you think of this or getting you to input stuff forces you to identify an existing habit that your customer might have that your habit loop can stack on. So there’s a concept called habit stacking. Have you shared that with your audience, Rob?

Rob (27:43)
talked

about it before, again, in a golden nugget, what would that be?

Amy Jo Kim (27:47)
Habit stacking is attaching a new habit to an existing habit. That’s really all it is. It is.

Rob (27:56)
The

example I like to use is, you know, when you have toothpaste and brushing your teeth with mouthwash. Like, it’s brilliant to attach mouthwash to brushing your teeth because you already do that. And getting you to do the mouthwash is easier if you stack it on top of doing that rather than creating an entirely new habit out of

Amy Jo Kim (28:12)
Right, so we’ve incorporated habit stacking as a key cornerstone fundamental element, much more so than when you and I were working together. And part of why we’ve done that is I’ve seen, because I’ve worked with so many clients, I’ve seen what happens if you habit stack versus you don’t. it’s just, I think it is the best retention hack I’ve ever seen. Like, people should do it more. So habit stacking, know, flossing or mouthwash after toothpaste.

Walk the dog, listen to podcasts, sit down with a cup of coffee for work, boot up Slack. All of these are habits stacking. A lot of tools have it stacked. So our tool now will use all this synthetic data to suggest habits that might exist that you might want to stack on. And then it uses those and builds the habit loop around that habit for you when it storyboards your end-to-end journey.

Rob (29:11)
can

validate with real users, right?

Amy Jo Kim (29:13)
And then, right, exactly, exactly. And that’s what we do. The end to end player journey, discovery onboarding habit building mastery, that’s what you’re storyboarding. And then within that, we use best practices of what should the main beats be. So for discovery, it’s like, where are they discovering it? What were they doing when they did it? You know, what happens? How did they learn more? What gets them to convert? You know, that’s all discovery. It’s, it’s, it’s like.

the core of strong product design embedded into those beads. What’s a habit loop? Well, there’s gonna be a trigger, there’s gonna be some activity, there’s gonna be feedback. There might be rewards, but feedback’s more important. There’s gonna, cause it’s a feedback loop, now you’re building a system. And there’s gonna be some sense of progress. And then the whole thing has a progression model where when you get to mastery, what’s different?

And so we have all these suggested beats. The system now generates a version of it from answering a few questions. You can look at that and then you can start changing it and say, okay, well, it’s saying they discovered it on Instagram, but my people have moved to TikTok, so I’m going to change that. And you can edit it based on what you know. So that’s, that’s what it does. And I have to tell you.

Rob (30:34)
Where can we find out more about this? Like if people want like they’re excited about the tool, they want to find out more, see a video demo or something. Where can they go?

Amy Jo Kim (30:42)
Go to gamethinking.io slash journeymaker.

Rob (30:46)
Amy Jo, thank you so much for investing this time in the Engagers and revealing so much of that that you’re doing right now. How that’s changing the way you work, how that’s making you a lot more productive and how, you know, you can actually dedicate more time to some of the things that you probably wanted to do already. Just focus on that, add a lot more value because that’s where the big stuff is actually happening. However, Amy Jo and Engagers, as you know, at least for now and for today, it is time to say that it’s game over.

Hey, Engagers, and thank you for listening to the Professor Game Podcast. And since you’re interested in this world of creating motivation, engagement, loyalty, using game-inspired solutions, how about you join us on our free online community at Professor Game on School? You can find the link right below in the description. But the main thing is to click there. Join us. It’s a platform called School. It’s for free, and you will find plenty of resources there.

We’ll be up to date with everything that we’re doing, any opportunities that we might have for you. And of course, before you go into your next mission, before you click continue, please remember to subscribe using your favorite podcast app and listen to the next episode of Professor Game. See you there.

 

End of transcription