Dr. Jeremy Weisz: 13:48
So they’ll use a kind of, you know, in conjunction with each other.
Raghu Bala: 13:52
Correct. So this left side column, this Agentic AI column, all these tools are for building an agent, but that part of it is for using these agents within an organization. So that’s where we started. But the bigger thing that we are sort of going towards, and we have already started to sign up networks, and you’ll see in a second what I mean by networks is the agent economy. So the agent economy is bigger than just building agents. The agent economy is to really change the way everything works. So in the agent economy, these agents are not only not confined to an enterprise, they can go beyond the enterprise. That means you can create a twin for yourself called Jeremy Twin. The twin top-level domain belongs to Synergetics, both on web two and web three. And Jeremy Twin can be given instructions, and Jeremy Twin could autonomously shop at various merchants. It has its own wallet. You would have assigned it some funds. It can negotiate with other agents. It can talk to websites. So that is the second column, this autonomous AI column that we are talking about.
And that is something that we are very excited about. And we are a first mover in the space because most of the companies in the space are talking about agents within the enterprise. We were the first mover in making agents sort of break out of the enterprise and go do things on the open internet. And that’s where sort of the collaboration with Project Nanda and MIT comes into play, because they had the same vision as us, except that we had already built it out. By the time they had published some of these, they had conducted some events and so on. And when I presented, presented on stage, they said, wow, you have already built out what we were thinking about. And so we have been sort of like a couple of years ahead. In fact, we started working at Google, and even they said, you guys are like a couple of years ahead of us in certain regard. So, the age. So I’ll just describe very briefly what the Asian economy is about. So let’s take the example of e-commerce.
So e-commerce, if you look at it, let’s say, let’s put 1995 as a starting point. So for about 30 years, e-commerce has been manual. You go to a website and view a human’s basically shop visually with point and click, and put things into a cart, and then check out, and then pay, and then get the stuff shipped to your home. So that process has been sort of the way every merchant site works, and that’s going to remain there for human-based shopping. But where we are headed next is in AgentConnect. It’s going to be the consumer might now become a may not be a human anymore. It might not be Jeremy anymore, but Jeremy dot twin. So Jeremy Dot twin is shopping, and Jeremy Dot twin, like all other agents, won’t navigate a website visually; it needs to communicate with the website, and for that, you need a protocol. So that’s where our pattern comes in.
We created something called AgentTalk And after we got our patent, Google came out with a protocol called A to a agent to agent protocol. And we told them we spoke with Google and we said, can we, you know, sort of not compete and sort of coalesce the two protocols. And they gave us a sort of a approach where we can extend their protocol and so on. But anyway, what’s important is when an agent sort of breaks out of the enterprise, it needs to now, first of all, prove its identity to websites or to other agents and so on. And so agent ID is a very core part of what we have built. That’s very important because just like just think about it this way, within an enterprise is like almost flying in an airport within America. You can just use your driver’s license and.
Dr. Jeremy Weisz: 18:23
Say the real ID, your real.
Raghu Bala: 18:25
Yes, real. You can use a real ID, and you can fly within America. Nobody is going to question you. But the moment you want to go to Europe or Asia or whatever, you need a passport. So that’s a decentralized ID, and that passport is what we have developed. So that agent is not confined to the enterprise. It can go worldwide. So think of the world as the internet and within America’s intranet. So it’s almost like that. So now with this decentralized ID, the agent can go anywhere. And we want to make sure that the agent is a good agent. That is, it’s not a bad actor. It’s not going to commit fraud and things like that. So we go through a process called key to know your agent. And for those who come from a banking background, they are very familiar with KYB, KYC, know your business, know your customer, and so on.
So these are just fraud prevention sort of processes that are there, even for those who have been building apps on the App Store. Apple won’t let you put an app on the App Store without going through a background check for your organization to make sure that you guys are not going to introduce something malicious. Same thing here. Okay, now that the registry in which our agent is registered is called the AgentRegistry. And what we have done with MIT is we have our own registry. But MIT, we have also helped them develop something called a federation of registries. That is, any registry in the world would be visible to the MIT registry. And so that way, it enables the discoverability of agents anywhere in the world. So that’s the part that’s sort of an extension of what we built. Okay.
Next comes a wallet. So just like you and I carry our driver’s license in our wallet, this agent will carry its identity in its wallet. But that wallet contains not only the ID, but also money. Similar to, you know, you and I are carrying money in our wallets or purses or whatnot. And that money can then be used to spend on e-commerce or other purposes, and so on. So there’s, there’s money with the agent, and just to sort of validate what we had done last week, Coinbase introduced AgentWallet. And I had to remind a lot of people online that that is not the first AgentWallet, because we have had ours for about a year and a half. But anyway, now agents have got ID that’s registered, they have got the wallet, they’ve got money, and now agents can operate in the physical world or virtual world, and they can talk to websites. And there’s a protocol for communicating. Now, while all of this was being built, Google also continued innovations.
And then they introduced something called Universal Commerce Protocol UCP. And so that’s actually very useful. What it does is I mentioned earlier that in Agentic commerce, all of the merchant sites now need to adapt, or they have to go through a retrofitting period where they need to be able to accept transactions from non-human agents. And so that would be a very difficult exercise if each and every merchant had to invent a method of doing that. So the universal commerce protocol, what it does is it creates a standardized way in which all e-commerce sites can employ the same protocol, and so that agents have a standardized manner by which to, you know, add to cart, remove from cart, checkout, pay, and all of that. And then another innovation that’s being that that’s taking place in the industry is that a number of companies have glommed on to this agentic commerce and have started to build agent payment protocols.
So there’s a set of them that have come out. So AP two is from Google. There’s, I think, ACP, I think, is the open AI one. And then Vic is from visa and then 402 is from Coinbase, and so on. So we, as a company, synergetics what we’ve done, what we have done is we have tried to standardize and embrace all of these protocols, including the ones that we have patented and created. So we don’t want to be the Switzerland of this space in the sense that. So when you go to a typical merchant, let’s say you go to Macy’s, and you want to shop for a shirt, and you have to go to the Cashier. All they ask for is a card, and you put it into the machine. And whether it’s Amex, Visa or Mastercard, it just takes the money, and the exchange of value happens with the merchant. And now that that box is actually figuring out, okay, this person put in a visa card, let me send the transaction to Visa or Mastercard, or this goes to Amex.
Even though these three companies compete with each other, the box gives the merchant what they want. And because merchants just want to get paid, they don’t care which card the consumer uses. So the same way, we want to be released from the payment side of things to be the equalizer in the sense that as long as the agent is able to pay through a multitude of protocols, we want to make it simple for the merchant to say, you don’t care which way it came. Just make sure that it gets settled and cleared in the banks, and the money ends up in your bank account. So we do all of that plumbing in the attic space. So anyway, I didn’t want to get more deeper than this, but this is.
Dr. Jeremy Weisz: 24:22
No, this is super interesting. And if people are listening to the audio as he’s talking, I’m kind of following through on the different pages on the website. So you can go to Synergetics.ai, and if you go, click on AgentWorks. There’s you can see I’m just clicking on the autonomous AI area. But I’m curious what it’s like, maybe a specific use case, right? Because I can see okay, there’s a twin. My twin has to have a wallet. You help power that. You have to register the person. Right. Because you want to validate that this is not an evil twin. It’s a, it’s a, it’s an angel twin. And then also ways to connect and talk in the marketplace. What would someone use this for? Sure. This twin for typically.
Raghu Bala: 25:07
Yeah. So let me give you an example. So that is a kind of customer of ours who is a fairly big e-commerce vendor, e-commerce platform similar to Shopify, not Shopify, similar to Shopify. They’ve got 400 merchants, 10 million consumers on their platform. And what they are doing with us is retrofitting all of their e-commerce merchant sites with e-commerce, such that the current 10 million consumers will continue to shop because they are all humans.
That will still go on. But now, with our capability, digital twins can also shop on their sites and be able to pay through one of these payment protocols. And we also have a relationship whereby we have some customers, such as cell phone carriers and so on, and these customers have apps on their phones where the twin can be created, and this twin looks like you. Even the body measurements are up to one centimeter accuracy. The reason why we did such things is because we know that once you create a twin, you might want to use it for apparel, shopping and things like that.
But anyway, so this twin can now be given instructions to go shopping. And one of the destination sites is our partner, who’s got these 400 merchants that you can shop from and so on. So the Agentic economy is a very thriving sort of marketplace of activity where people and enterprises can rent agents, consumers can create twins and ask those students to go to destination sites to either do restaurant reservations or travel bookings.
Dr. Jeremy Weisz: 27:18
Let me ask you a question. Let me see if I’m understanding this. And this is correct. Right. Because I feel like these platforms let’s just take Instacart, right? Let’s say Instacart, like, hey, I want to put the Synergetics.ai technology on my platform, right? So, like, I have a grocery list. It’s literally a list on the counter, right? And my wife’s like, hey, why didn’t you put the peanut butter on the grocery list? Whatever.
So, like technically, let’s say Instacart integrated with this. I had a twin. I could literally just probably go, okay, can you just purchase eggs, get milk, get peanut butter? And it kind of goes in. I don’t need to go to the site, find the peanut butter, click the thing, and add it to the cart. I could just be like, hey, here’s what I want. Go on and do it. And would that be an accurate use case of this?
Raghu Bala: 28:11
Yes, it’ll go and do it on its own, and it’ll also know.
Dr. Jeremy Weisz: 28:15
So why don’t we contact Instacart and get them on this? So I don’t know.
Raghu Bala: 28:19
Yeah. That’ll be an ideal customer because this could be inserted into the Instacart application. And all consumers can create twins if they don’t want to shop on their own. And also learn from your experience over time that you like Smucker’s peanut butter or Jif peanut butter. I’ll figure it out on its own because it knows your purchasing habits, and it knows whether you’re a Pepsi or Coke drinker, things like that.
Dr. Jeremy Weisz: 28:47
So I like Kombucha, but yes. This is interesting. So the platforms can integrate these digital twins at some point. Correct. And then that allows consumer the more, I mean, it’s kind of like the infrastructure if you take like electric cars, right? They had to eventually put the infrastructure in for these charging stations because, like, then people are supposed to drive. The infrastructure is not in place, so these platforms have to put this. Have to adopt the infrastructure so that I, as the consumer, can use it on their platform.
Raghu Bala: 29:19
Correct. So that’s what we are doing. So we are always B2B, B2C, or B2B. So our customers are always a business, and we give them the rails and the plumbing so that they can then empower their users to do twins, to do agentic payments, to do all these new and exciting things.
Dr. Jeremy Weisz: 29:39
So I could see from a consumer perspective other commercial uses. I’m just looking through your website, and it shows like a student loan agent, and it says a therapist agent. Can you talk about what are some of the commercial use cases?
Raghu Bala: 29:55
So one of our customers is a sort of mental health platform, and it provides addiction counseling for drug and alcohol, and gambling addiction. And these agents are augmenting, I would say, the best way to say it, counseling agents at various gambling sites, at various casino sites, at various sorts of healthcare facilities, and so on. So let’s let me give you a real-world example. So let’s say that someone with an addiction problem has got an acute compulsion to want to, you know, consume some, let’s say, alcohol or drug or whatnot at 3 a.m. in the morning on a Thursday. And you cannot tell that person that the counselor will be in on Monday morning, and you can talk to them. That’s not timely. So now, if you have a support system through an AI agent, and this particular one actually is an avatar, which is almost like a personal companion on a phone.
And this avatar will show up on your phone and know everything about you because it knows your case history, knows all of your sort of like, you know, times you have sort of come off the rails and things like that, you know, and had setbacks and so on. And the, the timeliness is very important. And also, it has the 12-step program and all that programmed in so that it’s a rehabilitation process, and it’s able to graduate you from step one to step two, up to step 12, and so on, and counseling and things like that. So, that’s a very useful real-world use case. And we know that human counselors are great, but they’re not going to be at your beck and call at all times of the day. And this is a very, very good augmentation of the support system that a person needs when they’re going through these types of episodes and problems.
Dr. Jeremy Weisz: 32:07
So it’s similar, like okay, a telehealth company, like, hey, instead of connecting you with an actual human, we can have these twins, and they can act, or let’s say Chase Bank, like instead of calling Chase Bank, you could go, hey, I need a loan or something. You go on, and I just talk to the loan agent here, and they’ll kind of communicate back with whatever. Hey, I need to get underwriting. I need to get this. I need to get that. And they can communicate, which is more efficient for the bank itself. Is that accurate?
Raghu Bala: 32:38
Yeah. Accurate. Another one, I’ll tell you. In healthcare, which is also kind of useful, is ambient listening. So let’s say that there’s a conversation between a patient and a clinician, a doctor, a nurse, or whatnot. And typically, what happens today is that as they talk to you, they are taking a lot of notes, copious amounts of notes, and then those notes get transcribed into an electronic medical record system. And the amount of data entry or time spent on data entry is significant at these medical facilities. So with ambient listening, what happens is it’s all HIPAA compliant and so on. And with the patient’s knowledge, a piece of software is listening to the conversation and is able to, first of all, summarize and transcribe it and enter all the fields in the electronic medical record system automatically, so that the clinician saves a lot of time.
Instead of doing data entry, they can focus on the patient. And that’s number one. Number two, not only does it do that, but it also can look up the ICD ten CPT codes for medical billing and things like that. Number three It can also, depending on the target EMR system, it can come back with a protocol. So protocol, in a sort of healthcare parlance, is a treatment process. So let’s say that you have a knee injury and you go to a physical therapist. They have a set of exercises to improve your range of motion over a period of time. So it may begin with certain types of exercises and then certain other types of exercises. Because when you have a knee surgery, it’s sort of like it becomes like glue. It doesn’t move after a while. So you’ve got to slowly sort of break all of those, you know, adhesions. Yeah. That’s a word. It becomes like fibroids inside. So you have to slowly sort of ease them out and loosen them before your knee gets back to normal.
But there’s a set of procedures you have to go through. I know it firsthand because I had to, unfortunately, go through that myself. But, but but that sort of process, it can even describe it, can tell you, like, okay, for this kind of injury, because let’s say the knee is a very complex sort of body part, and there are many types of injuries. The ACL or your kneecap breaks, or you have other types of injuries. Each one has a set of things you have to do. It’s not all one size fits all. So, it can tell you, okay, I think, based on what I heard, this would be the protocol to follow. And then the human expert can come and say nah, I don’t think we need step three. You have to replace it. Step seven or something like that. And you can modify it. But it’s a real helper to, to the clinician in these facilities.
Dr. Jeremy Weisz: 35:54
Really interesting. I’m curious, from your standpoint, what are some of the apps, tools, and software you’re playing with? You could be personally or professionally, obviously, Synergetics, are there any other apps or tools?
Raghu Bala: 36:12
Yeah, sure. I mean, like everyone else, the developers are always, you know, they use various coding tools to generate code and things like that. Claude is now the one in vogue, but others are very good as well. There’s one which we use, Cursor.ai, which you know is very popular. Then on my end, I like one called Eraser. So Eraser is very good when you want to draw diagrams, you know, like box diagrams, flowcharts, things like that, process diagrams, it’s very good. It’s really capable. And in the old days, I used to use Visio, and then later on, I started to use, you know, like some people use Draw.io and even PowerPoint and so on. This just generates diagrams based on prompts. So it’s very useful.
Dr. Jeremy Weisz: 37:10
What do you use it for? So what kind of like is this what you’re talking about right here?
Raghu Bala: 37:16
Yeah this. So if you give it a block of text, it can create a process diagram that describes the text and break it down into steps. And then you can see it draws lines between the boxes and things like that. It’s very good if you’re doing any sort of process flow and things like that.
Dr. Jeremy Weisz: 37:39
Nice.
Raghu Bala: 37:41
And then I mean the usual ChatGPT, and I don’t know whether you guys have tried Grok.
Dr. Jeremy Weisz: 37:47
So, I haven’t played around with Grok no.
Raghu Bala: 37:51
Grok is very interesting. You can take a still image. Okay. Any still image. And it can create an animation around the still image. So, for example, I think you have to click imagine and that box there. And then you can upload your own any image that you have.
Dr. Jeremy Weisz: 38:12
That’s pretty cool.
Raghu Bala: 38:13
You can upload a still image at the bottom. At the bottom. Yeah. You can upload any image on your phone or whatever it is, and then it will animate it as though it were a short film. Just I don’t know how it does it, but it is just able to do it. It’s amazing. So I should play with that.
Dr. Jeremy Weisz: 38:36
That’s amazing. Any other productivity tools software you just use just to be more productive in general could be in your email. It could be in your phone.
Raghu Bala: 38:49
These are the primary ones I’ve been playing with. I’m sure you know, we have 25% team, and I’m sure these guys are always screwing around with something or other.
Dr. Jeremy Weisz: 39:02
Exactly. I’m curious, you know, you’ve had several other successful companies. What are some of the things you learned from an acquisition perspective? Because I know, you know, you’ve been a, you know, a company, a founder, and you’ve been acquired.
Raghu Bala: 39:23
Yeah. So that’s probably why I’ve been on both sides, actually, when I was working at Infospace, and even a little bit of my function at Yahoo was looking at companies to purchase and things like that. Infospace, I did a lot of that, but so on. I would say its acquisitions have to be done carefully because it’s not only. A kind of fit in terms of the technology companies, the product set has to sort of coalesce nicely. That’s one part of it, but the culture needs to fit in properly. So I’ve had acquisitions where. I, the acquisition after being acquired. I’ve worked for those companies for years, and not, you know, I felt normal, but then in some places I felt out of place because the culture was very different. So when I sold, one of my companies got acquired by Infospace, and it was okay. I mean, they had their own giant.
Dr. Jeremy Weisz: 40:36
Giant infospace. So that was okay because I sort of felt at home at Infospace, it was okay. But then one of the companies got acquired by a small public company. It was a services company. So when a product company gets acquired by a services company. Service companies don’t understand product companies well. So product companies have a different sort of like a mindset on how to generate revenues and how they scale the company, and so on. Service companies tend to look at billable rates and hours and things like that. So it’s a very, very different mindset. And so the two sort of never come together. So I felt sort of out of place, and it’s a difficult sort of thing to marry these two types of companies. But culture is very important. I think in, in doing acquisitions.
Dr. Jeremy Weisz: 41:38
Yeah. It’s interesting. I’m curious, who do you follow in the AI space that you consider thought leaders that you’re following to see where people are innovating and what they’re doing?
Raghu Bala: 41:51
So there are, you know, I won’t mention specific names, but there are, I think, at least about 20 or 30 people that I subscribe to on Substack, Medium. And I get inundated by the incoming email every day. But, you know, because they all have got this way of bringing you the latest articles, and then you can either read it in the email or you can just click to go back into Substack or Medium or wherever. And I find that very useful because the information comes to me as opposed to me going out there and looking for it.
And so I’m, I’m, I don’t mind being bombarded a little bit, but I, I am a good scanner of information. So a filter of information. So I quickly go through and say up? That sounds interesting. And then ignore the rest of it. So. And then I pick up those pieces and then try to figure out how it fits within sort of my own thought process. So one of the things I tell folks is there’s so much going on in technology that, you know, the periodic table in chemistry, where every element has.
Dr. Jeremy Weisz: 43:07
To have it in my wallet. I’m a biochemistry major, so I hate to admit that, but I’m very geeky. Yeah.
Raghu Bala: 43:14
So, so you understand that every element has its own cubby hole. So the same thing comes to technology. Whenever I see a new piece of technology, I always try to figure out where, in the greater scheme of things, it sort of fits in, because otherwise it just becomes noise. It’s very hard to link things together, so you need to always connect the dots. So this is my way of keeping order in a world of chaos, which is, you know, there’s so much going on. So I have this kind of mental map, as to like, okay, this fits in here. That fits in there. And if I cannot fit it in, I try to then read further to figure out, oh, this is a new thing. We need to make a new cubby hole for it. But anyway, I follow a lot of.
Dr. Jeremy Weisz: 44:04
Are there any specifically? I’m just looking right now.
Raghu Bala: 44:06
There’s one guy called Cobus Greyling. He’s on Medium. He puts out a lot of stuff. There’s one guy called Sebastian Raschka. He’s on Substack. There are a bunch of people who I have on X, actually, Twitter has got a lot of good AI if you really are into AI, and I actually use the save feature in Twitter to save their particular posts, it goes into like a bookmark list, and I can refer to it. So that’s, but there are so many. I mean, there are literally dozens of these guys. I’m sure that I just, but not everyone, you know, every word they say is a pearl of wisdom. They all put out good pieces. Then I filter and say, okay, what really makes sense for me? And also, there are a lot of overlaps and repeating technologies right now.
You see, if you look at the world of AI, it’s sort of slowly we are all coming to a realization that there are three teams playing. There’s the anthropic team that is OpenAI team, and there’s the Google team. And you need to figure out which camp you’re in. And the same technology is repeated in each of these teams. Okay. So that means you, if you know which team you are on, then you don’t want to learn three of those things. You understand? It’s like. It’s like too much overload. Pretty much doing the same thing. So. So in this way, there are technology stack elements that work within the stack of one of these teams. So again, it’s repeated a lot of times in this way.
Dr. Jeremy Weisz: 45:56
So anyway, yeah, I can go overwhelming if you’re like I need to create an agent in those. Are you talking about doing this in OpenAI? Are you talking about Anthropic? Are you doing it in Gemini? What? You know, so. But I don’t know. I like to hear repetitiveness. It means something’s important. Like you’re maybe. Who knows the seventh person who mentioned Cursor? I’m like, okay, maybe I should pay attention to Cursor because Raghu is pretty smart, and he’s mentioned it, and he’s like the ninth person to mention it at this point. So for me, maybe it takes a couple of times to get hit over the head with it before I look into it, too.
Raghu Bala: 46:30
Yeah, no, all of us are like that. So, so that’s, you know, even like what this one called this cloth bot. But before that, there’s one called Molt Bot or something like that, which came along at the same time. And it was it’s got a lot of malicious things that people found. Actually, I brought it to my team. I said, you know, people are talking about this, you know, want to check it out. And then they did some search and they say that it’s sort of like exposing things on your hardware. And we said, you know, I think maybe let’s.
Dr. Jeremy Weisz: 47:01
I read about that. Yeah.
Raghu Bala: 47:03
Yeah. So, we do some research, and actually, there’s no hierarchy. We are a startup. So it’s like I’m one of the guys in the team and we all behave as equals, almost. And that sort of like a camaraderie is good. And also, I share things, and I ask them to share things so that we sort of figure out things together. And so you actually multiply your brain by the power of the number of people you have this way, as opposed to a hierarchical type of setup?
Dr. Jeremy Weisz: 47:38
Yeah. No. First of all, I just want to thank you. Thanks for sharing the journey. Thanks for sharing. I love thinking about this in a different way than what most people are, because it gives people ideas on what they could be doing. You know, even if it’s, you know, something not related to an industry, I think it expands the thoughts and minds. So I want to encourage people to check out Synergetics.ai and check out more episodes of the podcast. And we’ll see you next time. Raghu, thanks so much.
Raghu Bala: 48:11
Thank you so much.
