This week’s podcast is about digital and AI agents. This will lead to the emergence of non-human platforms and business models.
Here is the link to the TechMoat Consulting.
Here is the link to the China Tech Tour.
Here is the podcast mentioned.
Here are the 9 digital business models I see often.
- AutoGPT and Other Tech I Am Super Excited About (Tech Strategy – Podcast 162)
- The Winners and Losers in ChatGPT (Tech Strategy – Daily Article)
- Why ChatGPT and Generative AI Are a Mortal Threat to Disney, Netflix and Most Hollywood Studios (Tech Strategy – Podcast 150)
From the Concept Library, concepts for this article are:
- GPT and Generative AI: AutoGPT
- Digital Marathon: Zero human operations
- Digital vs. Human Agents
- Platform Business Models: Non-Human Platforms and Business Models
From the Company Library, companies for this article are:
- OpenAI / ChatGPT
Welcome, welcome everybody. My name is Jeff Towson and this is the Tech Strategy Podcast from Techmo Consulting, where we analyze the best digital businesses of the US, China and Asia. And the topic for today, the question for today, is life just AI agents plus gaming engines? That’s kind of a strange question, but I’ve actually been thinking a lot about it. And… I think it might be true. Anyways, that’s kind of a high level question, but within that, there’s a lot of good business questions which are smaller and more usable, which are really about AI agents, which are sometimes called digital agents, digital AI agents. I’ll talk a bit about them, but in terms of things I keep an eye on in technology, digital agents, AI agents, it’s in the top two. I think it might be the most important thing that’s happening. I think it’s gonna be massive. We’ll see, we’re still in the early stage. So I’m gonna talk about that today, what that means for business, and then a little bit higher level, like, you know, is this what life really is? Digital agents plus some good gaming engines. Anyways. useful stuff but also a little high level stuff. Okay so that’ll be the topic for today. I don’t really have any housekeeping stuff. I just got into Dubai a couple hours ago. I’m sitting in the Dubai Marina which is pretty awesome actually looking out at the sun going down, yachts in the harbor. It’s pretty fantastic. Anyways so I’m a little tired so I’m just gonna jump right into it. No real housekeeping or anything today. Disclaimer, of course, nothing in this podcast or my writing or website is investment advice. The numbers and information for me and any guests may be incorrect. The views and opinions expressed may no longer be relevant or accurate. Overall, investing is risky. This is not investment, legal or tax advice. Do your own research. And with that, let’s get into the topic. Now the concepts for today, big surprise. Digital agent, I gotta think of a better term. Let’s say digital AI agents. That’s concept number one, the whole point. And then second to that would be generative AI, general topic. Anyway, so no real frameworks per se today, but yeah, two big related topics and this will all go in the concept library under those two headings. Now, all right, so let’s sort of start at the beginning. Digital agents, I’ve actually been talking about these a little bit. for probably two years. And it came up when we were talking, well, I was talking about marketplaces and things like IGE and Quora and Google search and Baidu search. And these are all basically platform business models, right? Platform business models were helping two different user groups interact. Maybe they’re buying and selling, maybe they’re watching videos, maybe they are doing search. things like that, which is a connection learning. But I’ve always used the term users as opposed to people. Because the user could be an individual watching a video, it could be an individual buying some ZOX on a marketplace platform. It could be a company selling. Now, ultimately that’s still people, but okay, a company could be a user. But when you start to get into things like search, what would be the user group that is creating content that the search engine is then searching through? Well, it could be the person putting an SEO on a website. A lot of that is sort of more automated over and over. And what you start to see is this, it gets a little blurred between is it a person as a user or a company or is it a person plus technology? Maybe software is more and more putting items up on sale on a marketplace that you can buy. Well, in that case, it’s not really a human agent as the user, it’s more of a digital agent. And when I’ve talked about companies like Google Search, Baidu Search in the past, if you ever look at any of my frameworks, my graphics, I’ve always put on sort of the content creation side, human and digital agents. because it could they are both potentially users and but they’re kind of linked at the hip usually okay and not too long ago i did a podcast i’ll give you the number here podcast one Podcast 163, which was the rise of digital agents and non-human platforms and business models. And basically I started to talk about Auto GPT and a company called Agent GPT. And these were all sort of digital agents, but not based on sort of software just running code, but based on AI, based on prediction, based on creating… let’s say generative images or something where it’s that same digital agent idea, but it’s AI. So hence AI agent, but it’s really kind of a blurred line. So basically when you start thinking about auto GPT, agent GPT, you start to think about digital agents, AI agents. Now what is sort of an, let’s say an AI agent? An AI agent’s just a GAN model, Generative Adversarial Network. It’s basically AI that’s testing against another AI, trying to fool each other, and that adversarial network is how they train and get smarter and smarter. But ultimately, we’re just talking about AI that can do prediction, whether it’s predicting the next word in a sentence, predicting text, natural language processing. whether it’s predicting entire paragraphs as an answer, hence large language model, where it’s predicting an image, things like that. What happens when you go beyond prediction to prediction plus decision and action? That’s when we go from talking about generative AI, or any type of AI really, to something we would call an agent. where it’s not just making a prediction, hey, here’s a potential answer to a question. We predict this is the answer you will like. We predict this image in the photo is a cat, but it’s actually predicting and then taking action itself without the human. So that’s Tesla. It looks at what’s going on the road, it makes a prediction, we should change lanes right now, and then the car takes the action without human involvement. It doesn’t prompt you. We recommend you take, you know, change lines right now. So it’s prediction plus decision and action. That’s when you start talking about an agent. Digital agent, AI agent. And that’s one of the reasons these things start to look a lot like humans. Human agent versus digital agent. Okay, let me give you some examples since that’s pretty vague, but that’s kind of theory. All right. We know… Certain digital business models are very common. I’ve outlined nine different digital business models I think we see all the time. You can look that up on the concept library. Just look for nine digital business models. There’s platform, there’s platform pipeline, there’s protocol ecosystems, there’s company ecosystems, there’s consumption ecosystems, there’s production ecosystems. There’s about nine that I see pretty often. But within those nine common business models, and the simplest business model would just be a linear business model, a traditional company like a factory that makes something and sells it. Linear business model versus platform versus protocol. What happens when you start to inject digital or AI agents into those existing business models? Now that gets kind of interesting. So I’ll give you two examples. One I’ve talked about before. Let’s say I go into a marketplace like Grab, Uber Eats, Meituan in China, something like that, Jumi of Food in Africa, Glovo for those that are familiar. And I go on there and I want to order a pizza. Okay. I go on as the buyer. I’m one user. human agent and I go and I look at all the websites. Oh, there’s Domino’s Pizza, there’s Pizza Hut, there’s Pizza Company, those are the other user group, the sellers, and I choose the one, I place the order. There’s a human on the other side or a human with some software, and then they deliver it. You know, they buy it, we do the transaction, someone delivers it. Okay, that’s a platform marketplace business model with mostly humans on both sides, human agents as users. What happens instead of me buying the pizza? I have an AI assistant, an AI agent, and I tell my agent, get me a pizza. And I give it a prompt. And the AI agent would then do, would take my prompt, and they would break it into a series of tasks. And this is what AutoGBT does. And AutoGBT is basically an AI agent based on GPT as the large language foundational model. But it can be based on. Image generation models, multi models, multi model, lots of things. But let’s say Auto GPT would break it down into tasks and say, okay, what are the most popular pizza joints within a three mile radius of my home? What do we know about Jeff? What kind of pizza does he like? What are the prices of these pizzas? And based on those three tasks, let’s say three investigations, it would then make a decision. Here is the pizza we recommend. Based on that, it could come back to me with a prompt, we recommend you buy pizza A at Domino’s. But in this case, it would just do it without me. It would go out, go to the website of Domino’s, it would interact with the website, it would place the order, place my credit card, do all of that, or do it through the app. And then the pizza gets ordered, the pizza gets made, it then gets delivered to me, it would track the delivery if there’s any issues, customer service, it would engage on that. So it basically is the person buying the pizza, not me. So the human agent as a buyer has been replaced by a digital agent on one side of the platform. And in theory, you could see that dominoes might have agents on the other side that take all their orders for them and do all the interactions. Then this marketplace platform is really not human to company or human to company with the person looking at the screen. It’s really digital agent, in this case, AI agent, AI agent to AI agent. It’s a marketplace for non-humans. That’s kind of odd. That pretty much works now, although not with AutoGPT, but it’s getting kind of close to that. So, okay, that would be one way AI agents would change a business model we know very well, which is marketplace platforms. Another example, let’s say, you know, I want a lawyer. Okay, that’s not a platform business model, it’s just a standard linear business model, a value chain. sir, you know, like Michael Porter, and you would hire a lawyer. Well, what if I hire a digital agent as a lawyer? So it’s still a service I’m buying, but it’s not a human anymore. And I might have a lawyer I go to that’s digital. Maybe it’s a chat bot on my phone. and it gives me answers to basic questions. It reviews contracts for me. I mean, this already exists, like not as a service, but you can just do it with ChatGBT. I do this all the time for contracts now. Every contract I feed into a chat bot and I have it make suggestions. It’s very, very helpful. Now, I wouldn’t do that for anything serious yet, but I’m doing it for contracts where I would traditionally not have hired a lawyer. And I’m just, it’s free, sure, run it through there, why not? Okay, in that case, yeah, it’s a linear business model, but we’ve got an AI agent on the other side. So you could have a virtual lawyer. You could have a, let’s say, virtual assistant that does everything for you. You could have, what this really looks like to me, you could have a virtual best friend. I played with one of these the other day, which was like a… You could have a friend online, Gentosa, something like that. Gentosa AI, where you could have a friend online that’s virtual. You know what it reminds me of? It reminds me of how billionaires live. Because every billionaire I’ve ever interacted with always has a team of like, eh, five to 10 people that are just sort of their extenders and assistants. They have a lawyer, they have a banker. They have a couple personal assistants, one, two, three, four maybe. They have maybe one assistant who takes care of their family. They have one assistant who takes care of their house. Even some cases, there’s someone in their, almost like royal court, who is a jester, who is just there to plan fun. Well, not the jester, that would be to tell jokes. There’s another person who’s just there to plan fun and plan trips. It’s like they have a… like a team, like a royal court or an entourage, and each person has a role, and billionaires are rich so they can afford to have all those people on staff to just sort of handle aspects of their life. Well, you could see that AI agents, we could all basically have a royal court around us. You could have five AI agents that you deal with all the time, one that takes care of your house, one that sort of you interact for everything related to your family, one that just tells you jokes. One that is just for fun stuff in life, and what should I do this weekend? Oh, you should go to this movie, blah, and this restaurant. So when I think about sort of linear AI agent business models, I tend to think that we’re all gonna have that sort of little structure around us. Okay, so I think I’ve made my point. The idea is we’ve got digital agents, AI agents, we start to put them into the nine digital business models that are pretty common. And we can see that they start to be interesting. They start to change it. I’ve talked about YouTube before that I think generative AI in particular is gonna get very good at making videos either on its own or initially as a tool for content creators. So their marketplace platform, in this case, their audience builder platform for YouTube, there’s gonna be a ton more content. but they can also offer a direct service that’s not a marketplace, which is just a, anything you wanna see on demand, the generative AI will create. So in that case, a platform business model, an audience builder is probably gonna become an audience builder plus a pipeline, plus a linear business model. I think that’s probably what YouTube is gonna look like. I think that’s already what Google Search and Bing have become. They’ve complemented their platform model with a direct, answer service, which is BARD, right? Or Bing. Okay, so that’s kinda how I’ve been thinking about this. And before we get to sort of high level questions, let me talk, why are digital agents working AI agents? Kind of, not really. Most people who have tried these companies, these auto GPTs, agent GTPs, they don’t work that well. There’s some problems and there’s a couple problems with sort of where they are today. Number one problem is they have very limited context. So one, you can only put in a prompt of you know so much information to give tell it what you want to do but more importantly it doesn’t remember all your past interactions. So it can’t retain what it has learned from our past conversations. So all these digital agents, my personal assistant, all of, well that doesn’t work that well if they don’t remember what you talked about yesterday. Let alone over years. What you really want is a personal assistant, an AI assistant that remembers every conversation you’ve ever had. And what you like, and what movies you like, and what your responses are. So one, there’s not a retention of knowledge. And that’s mostly because of context limitations. Number two, interacting with the real world. is pretty chaotic. You know, most AI robots have a hard time navigating downtown. This is why Tesla, you know, when you hear Elon Musk talk about Tesla, what he’s really talking about, he’ll say this openly, is to make Tesla work, we had to solve the real world AI problem, which basically you had to train AI that could negotiate the real world, which is chaotic and confusing. And that’s a very different world than let’s say, training a robo taxi just to work in one business park. Or a robo bus that only goes on two highways back and forth. Or AI that only works, let’s say, within a video game or a marketplace. But no, real world AI is pretty chaotic, so that’s limited it. Another problem, large language models, which are text-based, aren’t enough. You really need multimodal. You need to feed in lots of type of data beyond just text, video, sounds, cameras, sensors, radar, all of it, to really do most things we’d want it to do. And then there’s always sort of the question is, how much can it improve? Can it remember what we’ve talked about? Can it actually improve itself? Can these digital AI agents write their own code? Which it turns out they’re pretty good at doing. They can write code, very well, but could an AI, which they don’t have the power to do yet, start to write and rewrite and debug their own code and improve, which is kind of a form of evolution. So can digital agents evolve on their own? So there’s sort of problems, and that’s kind of what I was thinking about two or three months ago when I was talking about this before. Let me tell you why I think this has gotten a lot more interesting recently. which is kind of why I’ve been thinking about it. Recently, I was listening to a podcast about this whole idea of AI agents that are basically being taught to play Minecraft. And for those of you who aren’t familiar, Minecraft is maybe the most successful video game of all time, at least it’s gotta be in the top five. You know, you build islands, you know, you play with all your friends, it’s kind of blocky and you can build farms and work together and go to war and all this. It’s kind of like this online world multiplayer where people can really, there isn’t a set like this is how you win the game, like a shoot-em. No, you do various things together. You can build and create in advance and all of that. Super popular, but basically, I was listening to someone talk about how they’ve been training AI agents to basically play this game. And more and more we see video games start to incorporate what really are AI agents. It used to be, let’s say you play Red Dead Redemption, you run around, it’s multiplayer, so there’s a bunch of humans playing and they’re all running around. But then there’s lots of characters you can interact with in the video game, townsfolk, and depending on what you do, they will respond. differently. You know, if you start killing people, you know, they all start chasing you, they start to recognize you, the sheriffs come after you. Well, I mean that’s not programmed. I mean those in the old days that was programmed like, you know, software. If I do A, this player, this, you know, non-player character NPC does B. Well more and more the NPCs are starting to sort of become AI agents that sort of act on their own, to some degree, and it’s getting more and more sophisticated. So when you go into Minecraft and you have basically AI agents playing the game, kind of like humans, that’s interesting for a couple reasons. Like, first of all, it turns out when you have an environment with set rules, as opposed to say the real world, which is chaotic, hence real world AI, but when there are set rules on what you can do, the AI does much better. simpler. It’s like the robotaxi that only operates in the business park, which has been designed to be easy for it, versus the robotaxi that’s going everywhere. Right? So one, it’s a simpler environment, it’s created that way. There’s tons of data. They get created by others and get created by interactions with others. So the more you interact with other humans or non-humans, generative, That part helps. So there’s a ton of data, and the data’s in fairly understood form. It’s not like the real world. It’s fairly structured. We know what it is, mostly video, obviously. And what I think is sort of most intriguing about that the AIs are doing this is it’s not like playing chess, which AI is very good at. There’s not necessarily a clear goal. It’s not like we have to win the game. You know, it’s, you can kind of create your own goals. It’s a little bit more free form. You can work with others together collaboratively on things. So it’s, in that sense, these digital agents, these AI agents, that’s kind of how human beings are. You know, we don’t get up with a clear, you know, we’re playing chess and it’s all about winning the chess. No, we come up with our own goals and we work together. So, it starts to look really kind of interesting. If you’re training AI agents in a video game. where it’s sort of an open-ended game where you play forever and the goals aren’t necessarily clear. And maybe it’s an open world like Red Dead Redemption, where it’s really like you’re exploring a world that has been created by the gaming engine, by the game company. And you sort of explore this thing and you go on various adventures. So that’s where I was starting to think about Is life a bunch of AI agents living in a gaming engine? Where you have multimodal AI agents that have their own agency to sort of explore and to learn this world and to adopt projects and goals and work together. I mean, think about a very large open world like Red Dead Redemption or Ghost of Tsushima. where it’s full of AI agents who have by definition their own agency to explore and to learn. That’s the more I think about it. I’ve been playing with this idea in the back of my head, like, is that what we are at the end of the day? Are we human agents that are living in some gigantic gaming engine that’s been created by somebody or something or who knows what? on these open-ended quests with our own agency to do whatever we want and to learn as we go? I don’t know. That’s kind of a strange idea I suppose, but I’ve been thinking about it a lot. Like it’s a bit strange to think about because like I came out of a physics background and one of the fascinating things about physics is there are clearly rules to the universe, right? Like discovered a couple of them. James Clark Maxwell discovered, you know, the rules of electromagnetism. Newton discovered the laws of motion. I mean there are clearly rules that govern things, which is kind of how gaming engines work. Like you create the gaming engine for whatever game you’re building and there’s a lot of room for sort of chaos and creation and people get, but underneath that there are rules like for how things have to work in the game. is there gravity or not, right? So, you know, physics looks like, it’s one of the strange things about physics is like why are there natural laws that seem to be true everywhere? And it’s like, I don’t know, but there are. And anyway, so that’s kind of the gaming engine side. And then if you start to think about the world and reality as a giant gaming engine with certain things coded into the game like gravity. and then you start to think about humans as somewhat similar to AI agency with agency exploring and collaborating doing this. That’s a bit of a crazy idea but I have been thinking about it a bit. So hence the title of the podcast is life mostly AI agents plus a really big gaming engine. I don’t know. Anyways that’s just a fun idea but let’s get back to something a little bit more useful. Okay, now within all of that, I think there are three things that matter right now if you’re in business. Number one, think about the nine digital business models. You can find them in the concept library. which how are they gonna change as digital and AI agents more and more are players within the existing model and change the model. Like I think YouTube’s model has to change. I could say the same thing about DD and Grab, this idea that there’s a guy in a taxi and I need a ride and there’s a marketplace that connects us. Well, very soon there’s also gonna be robo taxis. So the digital agent, the AI agent, the robotaxi, that is going to change the marketplace model for transportation. So we can definitely see it hitting these business models right now. This podcast I was listening to, who’s a real expert in this stuff, he basically argued three types of AI agents appear to be working quite well right now. Number one, software agents. So these are AI agents that can write software, that can test software, that can debug it. It turns out they’re very, very good at that. So that looks like a lot of this is theory, a lot of this is me just speculating, that’s real. The same way human agents write software, AI agents appear to be able to write fairly good software. As mentioned, gaming agents, it looks like AI agents really work well within games. as non-player characters, you know, so you’re interacting with them as you’re walking around, you know, ghosts of Tsushima, but also as players. They can play video games the same way AI is very good at chess. That seems to be the real thing. And then physical agents. So three types, and I’ll put the link to the podcast because I’m not citing it well, but software agents, gaming agents, and physical agents. Physical agents are… you put an AI agent in a robot and it starts to go down the street on its own, it starts to walk around. And I heard Elon Musk refer to human beings as meat computers. So this would be a digital computer in a robotic body. We may be, I guess, considered meat computers in a kind of a meat body. Oh, that’s not the greatest language, but yeah. So robotic agents. they definitely seem to be moving. So there’s at least three categories of AI agents that look like they’re the real thing right now. I think that I would pay a lot of attention to them. So anyway, so two takeaways that are relevant to business apart from me just speculating all over the place. Keep an eye on these business models. Look at these three types of agents. And I think that is the point for today. Anyways, a little longer than I thought. I’m kind of tired. I’m sorry about this, Mike. brain’s a little fuzzy right now but it’s been a crazy week so I did want to finish this up and I’ve got a sort of a dinner here in a couple hours in Dubai. There’s a conference and some other things here I’m going to. So anyways that is it for the content for today. Let’s see any fun stuff? I’m actually I was thinking about this. I first came to Dubai it must have been 2001. I was Back then, I used to work a lot for a Saudi prince, so we used to do a lot of stuff in the Middle East, somewhat into Europe, somewhat into the US. But we used to stay in Beirut. Office was in Saudi Arabia, but Saudi is not the most fun place to be, so flying in and out of Beirut, that was kind of the place to be. So if you, you know, a lot of professional type folks, bankers, consultants, lawyers, they would commute in and out of Beirut back then. And then Dubai was really taking off at that time, and more and more people started shifting from Beirut to Dubai as their home base, which is really what happened. And then Beirut obviously had some problems eight to 10 years later. Now most people are in Abu Dhabi and Dubai, and they commute in and out, or they’re based in Saudi. But anyways, I had come out here about 2001. I was looking at a company called Arabia.com, which was the Yahoo for the Middle East. and it was invested in by Al Waleed. So it was basically, it was in trouble. It was having some difficulty. The dot com bust had happened a year prior. And it was a funded startup based in Dubai at Dubai Internet City. And it was basically a portal business model, just like Yahoo back then, but it was in Arabic and it was for Arab speakers and mostly the Middle East. But. the business model they had based on was digital marketing, digital advertising. Well, that was pretty anemic for Arabic back then anyways. And then the dot com bust really cratered that business model, so it wasn’t clear what to do with it. But when a company like that gets in trouble, usually the person who’s the largest investor in the company, in that case was Al Waleed, it kind of ends up on their desk. What do you want to do? And so then I had gotten tasked to fly out to Dubai and take a look at the company and I brought in a consulting team and we sort of took it apart. And you know, there wasn’t really a great strategy piece to be done, it was basically a take it back to the garage strategy, which is you’ve scaled up, you’ve built a world-class portal like Yahoo, well done. Turns out. that’s not working anymore, we got to take it back to a garage business, which we ended up moving most of the operations back to Jordan and Amman and closing down. So it’s not really a strategy gig, that sort of thing. It’s more about getting all the people who put equity in there together and everyone agreeing what to do next. But yeah, there wasn’t a great turnaround strategy for that one, unfortunately. That was kind of the first project I did here in Dubai. And then I don’t know how many I did after that, quite a few. Lots and lots of work out here in the early days. That was long before Dubai sort of became a big deal. Anyways, yeah, it’s fun to be back, but back then we used to go to Dubai Internet City for those of you who are familiar. You know, that’s about halfway across Dubai now, but back then it was the boonies. I mean, we were way out in the middle of nowhere, and it was just the Internet City and hanging out there with the team. It was a lot of fun. Anyways, that’s kind of what I’ve been thinking about. I’m sitting looking at the yachts in the Dubai Marina right now. It’s pretty great. Anyways, that’s it for me. It’s fun to be back. I’m just here a couple days and I’ll bug out. But that’s it. I hope this is helpful. I know it’s a little bit hand waving in terms of a discussion today. But yeah, I hope that’s helpful. The net of all this is keep a close eye on digital and AI agents. I think it’s… like it may end up being one of the most important things that’s ever happened. We’ll see, or maybe not. Anyways, that’s it for me. I hope this is helpful. Talk to you next week. Bye.
I write, speak and consult about how to win (and not lose) in digital strategy and transformation.
I am the founder of TechMoat Consulting, a boutique consulting firm that helps retailers, brands, and technology companies exploit digital change to grow faster, innovate better and build digital moats. Get in touch here.
My book series Moats and Marathons is one-of-a-kind framework for building and measuring competitive advantages in digital businesses.
Note: This content (articles, podcasts, website info) is not investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. Investing is risky. Do your own research.