This week’s podcast is about JD’s big moves in robotics and embodied AI. Specifically JoyAI and JoyInside.
You can listen to this podcast here, which has the slides and graphics mentioned. Also available at iTunes and Google Podcasts.
Here is the link to Agentic Org Now.
Here is the link to our Tech Tours.
Here are some photos from JD’s exhibit at the World Robotics Conference.
——-
Related articles:
From the Concept Library, concepts for this article are:
- Ecommerce / Retail
- Embodied AI
From the Company Library, companies for this article are:
- JD: JoyAI
- JD: JoyInside
—–transcription below
00:06
Welcome, welcome everybody. My name is Jeff Towson and this is the Tech Strategy Podcast from Agentic Org Now and the topic for today my first-hand look at embodied AI robotics at JD. Specifically Joy AI Joy Inside. I wasn’t really that aware of how much JD was doing in robotics and in this area and it’s a lot. Very aggressive which is really compelling.
00:35
So I got to look at it, I’m going to sort of go through my takeaways and what I think they’re doing and yeah, it’s clever. I think it’s really smart. So that’ll be the topic for today. Housekeeping stuff. I have a new email I’ve started sending out every weekend. It’s short, it’s about three minute read with lessons in sort of agentic organizations, agentic commerce, things like that.
00:59
I think it’s pretty good. Let me know what you think. If you’re interested in that one, it’s available really at any of the webpages, jefftausen.com, at jentacourtnow.com. You can’t miss it. It’s a free email sign up. You get a free copy of my book if you do it. If you’re curious, sign up and let me know what you think. It’s trying to be super quick and valuable, the kind of thing you can skim on a phone in about three minutes. And over time, makes you smarter about
01:26
agentic operations as sort of the core of a business as opposed to something ancillary. So let me know what you think about that if you’re curious. Other issue, we’ve just finished up our sort of first e-commerce training program with Alibaba, which was really fantastic, a lot of fun, five to six weeks worth of classes and then going out to Hanjo, spending a week there. We’re going to roll out another one of these and we’re starting to look for a partner.
01:56
Countries we’re thinking about are Turkey, Indonesia, Brazil, Mexico, but we’re kind of open to anything. So we’re looking for a partner who would sort of be an educational partner. know, someone who has a local presence who knows, you know, this class is tailored for entrepreneurs, people who are already selling on e-commerce. Usually they’re very small, you know, a couple of people, maybe 10 to 20 people. So it’s sort of for SME, small entrepreneurs.
02:25
You know, we train them in the Alibaba way, and my way as well, and then take them to the next level. So if you have sort of a client base or a network that would fit that profile, yes, send me a note. We’re going to start working on this over the next month or so, and maybe launch it early next year. Anyways, if that’s interesting, send me a note. Let’s see. Standard disclaimer. Nothing in this podcast from my writing is investment advice and numbers and information for me. And it gets me being correct. The views and opinions expressed may no longer be relevant or accurate.
02:54
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 Okay, so the two sort of concepts for today learning lessons, whatever you want to call it that you know Basically go to Jeff Towson comm you’ll see a whole tab Which is just concepts to understand and there’s I don’t know how many there are 50 or 100 somewhere in there But really when you start talking about digital
03:22
AI and now with agentic, you’re talking about 30 to 40 ideas. And if you can get those down, you’re going to understand most of what’s going on. So, you know, every one of these podcasts, I kind of like to point to two concepts that are worth thinking about. Now, this one is sort of more general. Often it’s things like network effects, economies of scale, switching costs, platform business models, whatever. And they can be digital concepts. They can also be business concepts. Today, okay, it’s a little bit.
03:50
higher level than that. We’re really just talking about embodied AI as a phenomenon, which embodies a lot of things. yeah, people generally equate that with robots, lot of humanoids this year. It’s a lot more than that. It’s anything that is becoming intelligent and autonomous. Now, we’ve had intelligent products for quite some time. A smartphone, it’s literally in the name. It was a phone.
04:18
Steve Jobs made it smart, gave it intelligence, hence smartphone. Smart vacuum cleaner, smart refrigerator. We’ve been hearing about this for 15 years. Okay, when we take it up to the next level of embodied AI, we’re really talking about autonomous. You know, it’s your phone doing stuff without you. Okay, it’s smart, but it’s operating on its own. That’s robots. But in theory, you could do the same thing. You could put it into vacuum cleaners.
04:49
You can put it into blenders. You can put it into little toys. You can make anything smart and autonomous. Well, not anything, but you can make a lot of things. And they can be big cutting edge devices like autonomous vehicles, Tesla. They can be little things that sit on your desk and send you messages and turn the lights off and on and walk around or whatever you want. Could be toys, could be anything. So embodied AI is a whole big idea. And one of reasons I want to talk about JD
05:18
is because they’re playing across the board all types of products. Chairs, washing machines, little toys for children, things like that. Everything’s becoming embodied AI, which is smart and autonomous. Hence, important concept. The other concept for today is just e-commerce. OK, we all know what that is. But the combination of e-commerce and embodied intelligent products, which is really what JD’s going after.
05:47
That’s really interesting. So I’m going to sort of talk about that a little bit. Okay, so those are the two concepts today, but pretty high level generally, you know, usually we’re a little more specific than that. Okay, so why are we talking about JD? Back in August, I went out to the World Robot Conference in Beijing, which is fantastic, by the way, it was August 19th, the 23rd. It was amazing. I mean, it was just amazing. And
06:15
spent time there going booth by booth and then the next day I went to the World Robot Competition which was across town which was really fun. But within all of that, and I’ve been writing a lot about that, within that JD kind of surprised me. were two or three booths exhibits. They’re not really booths, they’re big. There were two or three that really sort of surprised me.
06:39
One of them was SOEs, state-owned enterprises, and all the stuff they’re doing in robotics. I’m going to write a lot about that. I think that’s super important. Another one was JD. I knew they’d be there because JD obviously has robots associated with their logistics division, and they have some industrial. But most of their robots are sort of in logistics. They got drones. They got warehouse robots. They got autonomous delivery vehicles.
07:07
I didn’t expect them to be nearly as involved in this as they were. They had one of the biggest exhibitions in the whole conference. It was like, I looked it up, 600 square meters. They were a strategic partner for the entire conference. It was massive. I probably spent an hour going through everything, writing down everything, looking at all. It was kind of crazy. Okay, so why?
07:32
They had products all over the place. I, for those of you who get my email list, I sent you a ton of photos. They had big robot horses you could ride on. I mean, they really, they don’t call them horses, but they look like robot horses. They’re huge. They got tiny little toys that sit on your desk or go around the floor. Little cute robots. They got little quadrupeds that are cute that you can put on a leash and sort of lead them around.
08:03
They got robotic hands. They got, you know, cutting edge humanoids. So Galbot, you know, fully autonomous smart humanoids. They also have that one from Unitree. I don’t if you know this one, the Unitree massive robot that basically is a huge exoskeleton you climb into, like out of the movie Aliens. They had that one there. I’ll put the picture in of that one. That one’s crazy. I do want to get into that. Like I want to climb into that and
08:33
and drive it like an exoskeleton like Ridley in Aliens, but I haven’t figured out how to do that. Anyways, they got products everywhere and that’s just on the robot side. You can go into like kitchen appliances, quadrupeds, robot dogs, exoskeletons that are just for your legs so you can walk upstairs easier, you can hike, people who have mobility issues can use them.
09:01
They have wheelchairs that are getting super smart. They have little machines that will throw the tennis ball to you so that you can practice hitting it back. I mean, it was just everything. There was stuff everywhere I hadn’t even thought of. And I thought that was really interesting. So why are they doing this? That’s kind of the question. And the first thing is the vast majority of these robots are not theirs. They’re operating as a retailer.
09:30
And, you know, they’re bringing in all these types of devices and they’re helping companies become smart. I’ll talk about how they’re doing that. And then they had their own products, which goes under the title of Joy Inside. So I’ll talk about Joy AI and Joy Inside. so what are they doing? They’re doing four things as far as I can tell. I think this is JD’s sort of approach for robotics and embodied intelligence AI.
10:00
Number one, they are becoming a very big retailer for robots and any sort of product that is intelligent and autonomous. Now that actually makes a lot of sense. didn’t know, I didn’t think about this before. JD historically has been a powerhouse in electronics, consumer appliances, laptops, mobiles.
10:28
I mean, this has always been their core category. One, because that’s what they came out of, the three C’s. But two, when you buy products like that, the trust factor is much more important. If you’re buying clothing and you buy something from Shein and it turns out the socks are a little bit cheap, you’re not that angry. But if you buy a robot and you find out it’s not working or you find out that the inner components have been swapped with cheaper ones,
10:55
which totally happens by the way, I’ve had that happen to me, not robots, smartphones. When you’re going to buy something more complicated, luxury, expensive laptops, things like that, who you buy from matters and that’s always why JD has leaned into where the retailer you can trust when you buy things online. And if you have, we don’t have any fakes, which is a JD thing.
11:21
If you ever have an issue, you can call our customer service and we will take it back. You know, that’s very hard to do in Alibaba and these massive marketplaces. They don’t have armies of people that pick up the phones, generally speaking. JD does. So the idea that they’ve sort of expanded from their core product category into, you know, products that are becoming smart and autonomous, well, that makes…
11:48
That makes sense. It’s a natural extension of what they’ve always been doing to try and sort of dominate this space. And you know what? Thinking about it, when I buy laptops in China, I always buy from JD, always. If I was going to buy a robot, that is absolutely who I’d buy it from as a retailer. If I was going to buy a… Now, sometimes it’s about the complexity of the product, the price tag. Other times it’s about the logistics. If you’re going to get a washing machine,
12:17
or a refrigerator delivered and installed, you need someone who’s got really, really good logistics capabilities. Okay, that’s like JD again. They’re really good at the installation part. yeah, so number one, they’re becoming a big retailer instead of just living in the core three Cs, they’re expanding to robots and all of these products that are becoming smart and autonomous. Totally makes sense.
12:46
Okay, number two, what are they doing? They are becoming an embodied intelligence solutions provider for tons of hardware companies, really OEMs, Original Equipment Manufacturers. You know, they want to be the go-to partner for any traditional product or hardware company that wants to make their product smart and autonomous and doesn’t know how to do it.
13:15
They are offering an entire suite of services that let you do that very easily. And that’s Joy Insight, which I’ll talk about. Number three, they have an entire suite of foundation models focused on embodied intelligence. That’s called Joy AI. Now,
13:40
Tencent is doing something similar. I’ve talked about Tyros, which is Tencent’s of suite of embodied AI models that you can use to plug in. They want to be the Android of robot intelligence. So any robot maker can just sort of plug in Tyros the same way any smartphone maker can just plug in Android. Okay. So Joy AI is similar in the target, but
14:10
They’re not just going for high-end humanoids and quadrupeds. They’re going all the way down to things that have intelligence but not that much. Tyros is not necessarily what you would do if you wanted to make a blender smart or if you wanted to make a $20 children’s toy smart or glasses or something simple.
14:34
JD goes all the way down from highly advanced to pretty basic and simple. So if you’re a toy maker and you’re having a hard time selling your toys because toys are becoming sort of smart and autonomous, what do you do? Well, you call up JD and they give you everything you need and part of it’s the Joy Inside, but part of it is the suite of foundation models you would need. And both of those, Joy Inside and Joy AI.
15:04
If you’re getting my emails or you can just go to my webpage, I’ve done pretty deep dives on these. I’ve got two emails sent out. There’s another one coming out tonight. I suspect I have written more about Joy Inside and Joy AI than anybody. It took me quite a lot of time actually to of dig into how the embodied sort of models all work together. What’s on robot? What’s in the cloud?
15:30
Where’s the SIM running versus what is the robot controller, the robot action model? Took me a long time to take it apart. I think I’ve got it. Interesting. And if you’re curious, go to those. So that’s number three is they have a foundation model suite that these robots draw on, whether they’re highly advanced or whether they’re pretty simple. And number four, they are creating and deploying logistics robots in drones at an increasing pace.
16:01
If you have a warehouse and you’re looking to put robots in your warehouse, okay, most of the robot companies I’ve talked about, I don’t know, Limx, Galbot is kind of doing some of that, Agibot is doing some of that, Unitree talks about it, they’re not really doing it yet, their robots aren’t really industrial strength. Okay, you could look at all of those, but JD Logistics has been doing this for a decade. mean, they are arguably one of, if not the best for
16:31
industrial robots that go into warehouses and on the logistics side of the world because that’s all they do. They’re not doing manufacturing plants. There’s JD industrial but they’re not really robot wise. They’re logistics focused. So that’s a good bucket to look at. I’m not going to really talk about that one today. That’s a whole other subject. But yeah, that’s I think that’s what they’re doing in this whole space. They’re becoming the big retailer.
16:59
for robots and any product that is sort of embodied AI. Two, they’re providing embodied intelligence as a service, as a solution to hardware makers, OEMs, and really product companies across the board that might do this. Three, they’re building a whole suite of foundation models focused on embodied AI. Four, they’re already deep into robotics in logistics. That’s kind of the four things.
17:25
I really want to just talk about number one and number two today. That’ll be kind of, if you want to know more about the Joy AI foundation models, which is kind of deep into the weeds, just go over to jefftowson.com and look at Joy AI. You can’t miss it. There’ll be at least two articles about that this week. Okay, for the retailer role, let me look at my pictures here, see if there were any examples I didn’t give you. They’ve got…
17:53
brands all over the place that they’re carrying as a retailer. So, UbTech, OnePlus, Honor. I mean, just across the board, they’ve got kind of all the big names. Honor is actually interesting. People don’t talk about it very much. Honor is the smartphone company. It used to be a smartphone brand inside of Huawei. Huawei was the sort of high-end smartphones. Honor was their sort of mid-tier economy brand.
18:23
for smartphones that was competing with OPPO and Vivo. Huawei as a brand for smartphones was competing more with Samsung and Apple. Huawei got, I’m sorry, Honor got spun out of the company after the Entity List ban, because when the Entity List ban happened in 2019, the area arguably they got hit the hardest was they couldn’t get high-end GPUs, CPUs, which is exactly what you need for high-end smartphones.
18:52
You can’t get 3nm. Why would you want a 3nm or 2nm chip for the smartphone? You could do 7nm and other and more for things like base stations and other. But to do a high-end smartphone, you really do need those high-end chips. you know, that entity list ban, I’m sorry, the Entity list ban hit them pretty hard. So they spun out Honor as a separate company, which is based in Shenzhen. Okay.
19:21
They’ve started to do robots, humanoids, which people don’t really talk about except for if you ever hear these little news factoids about how a Chinese robot just beat the speed record for running. It’s the honor, um, humanoid robot. It’s always the one that like beat the Usain Bolt world record for running.
19:48
It’s a really funny looking robot. It’s kind of red and you know, it looks a little different than your typical human road bike. But anytime you hear about, a robot just beat the human running record, it’s almost always the honor robot. It’s kind of funny that they’re making a mark in this space, then people don’t really realize it’s the smartphone company. What else are they selling? Agibot, which I’ve talked about before. I think they’re a really important company to study in terms of, if I’m going to place my bet,
20:17
in terms of which humanoid robot company, specifically humanoids, is going to be the leader. I would put my money on Agibot. I think they’re still number one for me. There’s others that are obviously very good, but their strategy makes sense. Their robots are top notch and their management team is really impressive. Anyways, take a look at Agibot. I’ve written about them quite a bit actually. Okay, so there’s lots of stuff all over the place.
20:48
Lots of small items that are actually kind of fun. Little toys you can put in your room. Little robots that go along the floor. Wearables. Exoskeletons. The wheelchairs are actually pretty cool. These wheelchairs you can sit in, they can go upstairs now because they’ve gotten smart and their robotics are pretty impressive. let me move on to sort of the more interesting part, which is…
21:16
JD providing basically trying to be a one-stop shop for any OEM hardware maker or product maker that wants to wants to basically go AI and wants to become embodied, intelligent, and autonomous. Okay, that’s Joy Inside. The other thing I mentioned, Joy AI, Joy AI is the brain. Joy Inside is the service and the solution that puts it in your product
21:46
and let you, know, end-to-end solution for you to do this. That’s kind of the distinction. JoyAI is sort of the cloud brain, really. Okay, so let me talk about JoyAI’s side, JoyInside. You know, JD is, the part of JD that people don’t talk about very much, people talk about the logistics, because it’s very impressive. They talk about, you know, sort of how they differentiate on the buyer side.
22:13
high quality retailer operating as a first person retailer as well as a marketplace. A lot of innovation in the physical side of that, which I just talked about with 101 Home. What people don’t talk about as much is how they work with various brands and merchants. They provide for them supply chain solutions. They provide for them logistics solutions. That’s what JD Logistics does. But you know,
22:42
They have had long-standing relationships with both Chinese and international merchants and brands, providing them various solutions. And this started out with, we can help you with payment, cross-border. We can help you with logistics. But it’s of expanded into more of the supply chain and particularly within the three Cs, which are computers, communications, and consumer electronics. So computers, personal computers, laptops, tablets.
23:11
peripherals, components, communications as mobile devices, networking equipment, smartphones, consumer electronics, digital cameras, audio equipment, wearables, things like that. Okay, they’ve been helping them with their supply chain in various forms for a decade plus. If you’re a, let’s say you’re looking for some components, sort of a B2B buyer supply chain,
23:39
and you’re making, I don’t know, let’s say you’re making gaming laptops, it makes all the sense in the world to work with JD to find the components you might need and then to work with them on the supply chain because they have that as well. So there’s a lot of that that’s been going on for a long, long time. JD Industrial does the same thing. So they’re very knowledgeable about the problems that these types of manufacturers have.
24:05
And they’re very experienced in providing them solutions, which could be marketing, it could be distribution, obviously they’re an e-commerce player, it could be financing, it could be logistics. They have solutions for all of these things. Okay, so this is a new problem that these businesses have. If I’m a toy maker, or I’m a, let’s say, wearables maker, or I’m an appliance maker, how do I make my product
24:34
smart and autonomous. Maybe I make a vacuum. Maybe I make a refrigerator. Maybe I’ve got, I mean, this is a major problem for these companies and they for the most part don’t have people inside that can do this. know, toy makers don’t have a lot of AI engineers that work there, generally speaking. Refrigerator companies actually can, but you know, this is a new skill set and new capability. They’re going to probably build this internally over time.
25:03
But in the near term, it makes sense to partner with someone like a company like JD. So what do they need? Well, they need robotic components for their products, sensors, cameras, actuators, batteries, all of that stuff that we would see in a robot. Yeah, they got to do all that. They would need some degree of brain power on robot.
25:33
on device, on product. Okay, that’s going to be GPUs, usually fairly low power ones. It’s going to be CPUs and you’re going to need a pretty tight feedback loop between those GPUs and the various sensors and actuators, right? That’s how you, a robot holds something in its hand. There’s a real tight feedback loop between the sensors and the actuators and the GPUs and CPUs on robot, which are pretty low power. Now you could put,
26:03
higher power GPUs in those robots. The problem is high power GPUs take a lot of energy, which means the more brain power you put on robot, the bigger your battery has to be. The bigger the battery is, the bigger the robot becomes, the heavier it becomes. So typically, you’re going to have fairly limited brain power on robot, on product, on toy, on washing machine, and…
26:29
you’re going to do a lot of the brain power up in the cloud. OK, that’s where you’re going to do your sensory processing. That’s when you’re going to do your spatial awareness. That’s when you’re going to do your higher intelligence. Most of that is going to be off robot, off toy, off refrigerator, and so on. So you got to kind of give them all of that. Now, in addition to that, you’re going to need to give these companies tools to both develop their products
26:59
programming, training, and then you’re also going to need to give them tools so that the customers can control their products when they use them, whether that’s a controller, device, whatever. So there’s four or five levels here that a typical hardware OEM would have to have to launch a product. Joy Inside is pretty much a solution to all of that.
27:30
And I’ll put some pictures of their robots and you know, they’re smart. I don’t know what these are rice makers, cookers, kitchen appliances, and they all say Joy Inside right on the device, which is kind of clever cooking machines, things like that. Okay. And you can, they would work with a hardware OEM to do this or JD does launch some things themselves.
27:56
which is pretty much how they always work, right? They will partner with certain products and then certain they will launch themselves. So they have a robot dog called J.Zao, which is a little robot dog for kids. It’s pretty cute actually. I’ll put pictures in the show notes. That one is actually by JD itself. So what do they actually provide? What is their solution to all the problems I just sort of described? Well, it’s pretty much four things. One, you get the AI models, right?
28:24
part of its on-device, most of the advanced stuff is going to be off-robot. And if you’re curious, look up Joy AI. I basically go through eight different models that they’ve built and that they’re advancing and how they work in basically four layers to handle all this. it’s…
28:47
They’ve got them for voice, which is called Joy AI Talker. They’ve got them for vision. They’ve got them for robot action autonomous called Joy AI RA. They’ve got simulation. They’ve got training. They’ve got runtime robots, blah, blah, blah. It’s I think it’s super interesting. It’s complicated. I actually spent quite a lot of time going through to make sure I understood it, how it’s working. I think if you’re interested in this space, you really do have to dig into the various types of models.
29:17
that you would have in a robotic device and how they all work together. In that sense, it’s much more complicated than just running something on a laptop. There’s a lot more going on. So if you want, go look at the models. That’s all under Joy AI. Number two, they provide them with, and if you’re an OEM doing this, you’re going to need ongoing access to these models. You could eventually at a certain point,
29:46
sort of bring it on premise and run it yourself. But, you know, these things are advancing so fast right now, you’re probably doing this as an ongoing contract with them for a while, depending on how complicated what you’re doing is. Okay, number two, JD’s going to provide a whole lot of developer tools. So these are the software toolkits, the SDKs, the APIs, all the external hardware makers can start to, you know,
30:16
put in all these capabilities into their robots and put all the micro controls and all that stuff. Generally speaking, these software development kits, they look to me to be mostly focused on sort of lightweight, to be sort of low cost, lightweight micro controllers, setting up your arm chips, your robotic edge boards, all of that.
30:46
You know, it doesn’t look too complicated. I’m actually sort of toying with the idea of setting up my own robot. Someone I follow who writes about equities in Asia, he just built his own robot doing something like this. It’s actually not that hard and I think there’s probably a lot of value in just buying the hardware, putting, using the SDKs and setting up your own robot and getting a little experience that way. I may do that. If I have a free weekend, I may try and…
31:15
You know, build my own sort of robot. Maybe something simple like a robot dog. Anyway, so the developer tools, they’re all there. They look pretty straightforward. Number three, this is more business. The first two are kind of tech. This one’s more business. JD provides the supply chain. Yeah, I mean, let’s say you want to ramp up a robot. You’re a toy maker or whatever. You don’t even know where to get the component parts. Well.
31:42
Okay, JD has a massive B2B industrial procurement and supply chain business. So all these hardware partners, it can basically match them with qualified and high quality component makers. And yeah, the hardware designs. Yeah, you can get those sort of pre-validated, checked, the board design, the component lists for adding all these capabilities. Yeah, you can sort of put this together within your legacy appliance.
32:12
you load that in, they will sort of do the supplier matching. They’ll connect. And this is very helpful if you’re sort of a startup or a small robotics company like I would be doing. Yeah, they can sort of help them match up with qualified manufacturers, sensor vendors, chip suppliers, all of that. Put all the parts together. And for some, know, in some cases there might be some co-development. You know, the larger partners, let’s say like a hire or something like that.
32:42
they could probably co-develop with JD. So I think the supply chain component is actually pretty powerful as a service. Yeah, I like that one a lot. And the fourth bucket, which is what JD is known for, is yeah, they can help you go to market. They can give you distribution and accelerate your go-to-market. So this is where their big online retail footprint comes into play. You know, they can put you…
33:10
in their Joy Inside store. They will promote and highlight. If you sign up and you build your robotic device with Joy Inside, I assume they’re going to promote you pretty good. They’re going to give you channel placement. They’re going to do co-branding. You can basically be within, let’s call it the Joy Inside ecosystem. That’s pretty compelling. And you can show up with the other devices.
33:37
Now on top of that you can do all the standard tools like e-commerce marketing, live commerce, all of that stuff, and you know these digital human live streamers, okay they have those. Those are pretty compelling. They might make sense in a space like that. Those aren’t specifically for Joy Insight and Robots. Those are for everyone selling anything on JD. They will have you know digital human live streaming systems which they call Joy Avatar. They’re pretty great.
34:07
They’re pretty high quality. you haven’t taken a look at Joy Avatar, take a look at it. If you’re a merchant or brand selling on JD, or you could sell in other places, yeah, it’s pretty impressive, the sort of the digital humans you can ramp up. Alibaba has something similar. Their digital avatars for creating videos or running live stream are getting really impressive. I did some articles on that.
34:32
Anyways, that’s it. They’re kind of offering you the brain, the supply chain, the storefront all in one package. I’m really tempted to try and ramp this up to tell you the truth and do this from front. Anyways, that’s kind of what I wanted to cover. I’ll give you a quick rundown of Joy AI just before I finish up here. Joy AI is their suite, but their suite of foundation models. But unlike, let’s say, Alibaba or Tencent, they are not playing across the board.
35:02
They are not building models for everything, video generation, image generation, all the types of LLMs, reasoning, coding, all of that. No, they’re pretty focused. They have their main LLM, is Joy AI LLM Flash. They seem to be mainly using one reasoning model. This is not just for embodied AI. This is for everything in JD, as far as I can tell.
35:30
And it sort of operates in a dual hybrid form. It can go between quick responses, flash, and longer form reasoning. It can kind of alternate back and forth. So it’s interesting the strategic choices they’ve made in terms of what they’re building and where they’re actually competing. But I said, look, they have like eight models. You can put them into four, let’s say five layers. And here’s how I think about them. The brain, Joy AI, the
35:59
The reasoning models, Joy, AI, LLM, Flash. Okay, that’s the brain. Most of the high-powered stuff, the high-powered intelligence is happening there. Fine, easy to think about. The next layer is kind of the eyes and the voice. This is listening and speaking. And within there they have two, Joy AI voice, Joy AI talker. So they’ve segmented out just processing incoming auditory information and then speaking. So this is important for voice interaction.
36:29
intelligence, but also sort of emotional back and forth with people. So a second layer, if the first layer is the brain, second layer is eyes and voice. Third layer, is pretty big layer, is the eyes. This is Joy AI VL interaction. Basically, this is actually four or five models within there. Long form video generation, simulation, video editing, image editing.
36:58
multimodal interaction. This whole eyes thing. This is basically, I mean, what’s the difference between AI on your laptop and AI on a robot? AI on your laptop, you tell it what to do and that starts the process. AI on a robot, it has to sort of perceive the outside world, which is mostly visual, and then based on what it sees, it triggers its own action. If a ball is
37:27
coming towards it, it sees the ball, it triggers itself, it responds. So you can put all of that under multimodal interaction. I just think about it as the eyes, you the eyes and then you respond to it. So brain, eyes and voice, level three, I’m sorry, ears and voice, level three, eyes, level four is body. They call this Joy AI RA, which is, I think it’s robot autonomous. I don’t know what the RA stands for, but.
37:56
That’s most of the stuff that happens on robot, right? This rapid fire robot action. It’s holding something in its hand. It starts to slip. It tightens its grip. Well, that happens very much on robot. Not entirely, but mostly. OK, we can call that body. All the stuff having on body, that’s one, two, three, four layers. The fifth one, which I actually think is pretty interesting, is another product, well, service.
38:24
capability called Joy Agent, which I’m not going really talk about it. But Joy Agent is sort of another level of this that lets one model, which is the Joy Agent, orchestrate and control other models. So this is how you go from one thing to 20. So I call this like the army. So Joy AI, it’s the brain. Ears and voice level two. Eyes, multimodal interaction level three.
38:54
The body, level four. The army to me is level five. That’s how you have one autonomous agent spinning up and orchestrating 10 other agents, which are probably specialized. So to me, all their models with Enjoy AI fit in kind of those five layers or categories. Now the last one, army, I don’t think that one’s on the Embodied AI playbook yet.
39:22
I think when they talk about Joy Agent, they’re mostly talking about software. But you could see in theory, that would one day lead to something where you have one main agent in your house and it commands the other things. So in theory, one agent could control your refrigerator, your robot, your vacuum cleaner, all of that. You talk to one, it tells everything else what to do and it does it all in a coordinated fashion.
39:48
You can see it might lead there one day, but right now I think it’s mostly one software agent commanding other software agents. Anyways, that’s kind of a simple way to think about Joy AI. I’m still digging into this. I think I’ve got it right. We’ll see if JD corrects me on any of this stuff. I’ve sent copies of my summary to them, so we’ll see if they come back and say, no, no, you’re totally off base on this. I don’t think so. I think I’m pretty close, but you never know.
40:16
Okay, that is kind of it for the content for today. Joy inside, Joy AI, and a little bit of Joy Agent. And that’s the content for today. I hope that’s helpful. I think it’s important. I think 2026 is really the year of agents, and I think 2027, we’re going to start to put these things, not just in products, which is what I’m kind of talking a lot about here today.
40:41
but I think they’re going to become part of business operations. That’s my working belief, which is, you know, businesses are going to be combinations of humans, agents, and robots in combined operations, which I’ve been calling Harko, human, agent, robot, combined operations. We’ll see. But robots are coming soon, but not this year, probably next year. That is it for me. I don’t know, not much going on here. I’m just working like crazy, having a really good time.
41:11
I’m flying out to Tokyo in a couple days, which I’m looking forward to. I really love Japan. I don’t do really much business there. I’m going to a conference, a robot conference in Tokyo, and I’m going to be on a panel. But this is a lot about just having a couple days to relax in Japan, which I absolutely love. There are certain places in the world I will take almost any excuse to go and just spend a couple days and relax. Japan’s on that list.
41:40
Spain, Italy, France are on that list. you know, my joke is the threshold to get me on a plane is so low. So anyways, Japan’s one of my places. So I’m going to just relax a little bit. And I spend a lot of time just sort of reading about Japanese history, which I really enjoy. So I’m going to see some stuff about that. Restaurants, I really do like going to Japanese restaurants a lot, which is interesting because I’m not really much of a food person, but this is kind of an exception to the rule there.
42:09
I don’t do shopping or any of stuff other people do. They go out for the nightlife and they go for shopping and I don’t really do any of those things. I’m more likely just to be sitting in an ice park reading for a couple days and then going and having a little bit of some good food. That’s pretty much the extent of my vacationing. This is just a couple days, so not really a vacation, just a little bit of a break I suppose. Anyways, that is it for me.
42:37
I hope everyone is doing well and I’ll talk to you next week. Bye bye.
——-
I am a founder, consultant and keynote speaker specialized in digital and AI agentic strategy.
I am the founder of Agentic Org Now, a firm that helps leadership teams launch AI and agentic initiatives fast.
I write (a lot) about digital and digital AI strategy (3 best selling books, +2.9M followers on LinkedIn). You can get a free book at the email sign-up below.
My Moats and Marathons book series is a framework for building and measuring competitive advantages in digital and AI 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.
