Agents are the future of finance: Enter the Agentic Revolution

Transcript

Hi everyone, my name is Sam Green. I am the uh founder of Cambrian and today I'm going to be telling you about the future of agents. Uh I'm going to start with a controvers controversial opinion. Crypto sucks. What I meant to say was that crypto UX sucks. So, I don't know about you, but I hate seeing this and this and this and this. Um, I've made thousands of transactions using these interfaces.

And, uh, every time I do it, I don't like it. And so, I know what you might be thinking that I'm going to say. AI solves this. Well, yes, it will, but saying this isn't specific enough. And specifically, I believe that agents solve this. So, and by this what I mean is that agents solve the crypto agents are going to be solving the crypto UX problem. And in addition to that, agents are going to be solving many more things which I'll be getting into in this talk.

So just in case that you're not familiar with the word agents or what what agents actually mean, I'm going to explain them briefly. So agents, I define agents as autonomous decision-making programs that take actions on your behalf. And so this is a very broad definition of what an agent is because many of you out there are developers and you might think hm I can make some sort of auton autonomous decision-m program that takes actions on users behalf and it's not using AI. Well, I would still classify that as potentially agentic. It has to be broad and I'll I'll explain why. So briefly going through this what agents actually do. agents specifically agents sense,

think and act. They have to sense their environment or get information about the decision-m process. They then have to do some sort of computation around the actions thinking about the actions they want to take and then they make an action and then this this loop goes on and on. So agentics agentic systems run continuously typically. So I I know that you're already deeply familiar with the thinking step and I have a question. Who who in the audience has used one of these models for financial advice? Just financial advice.

Can I get hands for people who have asked questions about money to chat GBT? Who? Okay. Now, who has actually made directly made money through direct advice from one of these models? Okay, I I saw one hand. Uh, [laughter] congratulations. Um well all of you are well aware of the limitations of the systems today of these LLMs today. They make mistakes. They hallucinate. Many of them can't even count the number of Rs and the word strawberry. And what I'm going to be going over the next few slides is basically explaining how these limitations are going to be a distant memory. You're going to be telling your children about these like errors these

early systems made. So for some historical context f first I'm going to give some historical context. So 60 years ago uh Gordon Moore who is the co-founder of Intel created or proposed Moore's law and he predicted that the basically computational power would double every two years and it did. It has for 60 years and it's going to continue doubling every 60 years or every two years. And it turns out that this rate of change has also been applying to AI. In fact, AI AI doubles in performance every single year and it's going to continue doubling in performance every single year.

So in 2022 after decades of AI progress basically this doubling of performance in AI systems combined with Moore's law progress basically we broke through uh the public perception or basically AI became part of public awareness when GPT3.5 came out and with GP23 uh GPT uh was able to achieve 100 million users in two months and you could compare that to 2 and2 It took Instagram who is second place two and a half years to get 100 million users.

So the point is AI is going to continue getting twice as good every year. And what this means is that the brains of agentic systems are going to continue doubling in performance every year. This is something you can just almost guarantee. So imagine if you got twice as smart every year. Imagine how powerful you would become if you could continue getting if you got twice as smart year over year. So one of the basically I would like to give some intuition of how far AI has come in the context of robotics basically using something tangible.

Robotic systems are actually much harder than just doing an LLM because robotics robotic systems have to basically move actuators. They have to see many of them have to see things. they have to hear things and so their job is much harder and they've come a long way and what we're going to be seeing is that many of the basically you can assume that you know as you see these robotic systems that progress have you see as you see AI progress you're going to see that this progression being applied to software agents as well especially in the financial setting so in the past robotic systems robots weren't much to be proud of so if you look back about 10 years. This is the state of robotics 10 years ago. You can

find many fail compilations online, but now 10 years of progress uh has has brought robotics a long way. Hey little guy, please don't fall. So here we can see this this robot is using state-of-the-art AI. Uh it's an open source project. Basically there's a technique called reinforcement learning that was used to to train its its system. You can see that it almost fell but it didn't. 5 years ago it would have

fallen at this moment. Um and um yeah so robotic systems are coming a long way. I'd like to uh take a moment to thank uh Bailey. Bailey right there. He is the founder of Prisma X for for doing this demo. Um, Prisma X is is operating at the intersection of crypto, AI, and robotics. And if you have any if you're curious about this this topic, find Bailey after the talk. Thanks, Bailey. Okay. So, just like Moore's law teaches us to continue expecting doubling in

capabilities every two years, we can we we're going to see the we're going to basically going to be seeing um we're going to basically be seeing the same thing in Agentic Finance like I said. And what I want to do is Oh, sorry. Thank you. Okay. So basically I wanted to bas I want to take a few minutes to talk about some history to make the argument that we're moving into a new era that is going to change our all of our lives. And specifically the era that we're moving into is the agentic revolution.

And the the agent the the agent revolution is going to have more impact than all of the previous revolutions in the past. And what happened with every revolution in the past is we had an explosion in productivity. That's a key word, productivity. And each revolution started with what's called a big bang moment. And we're going to start I'm going to start this brief history with covering the first revolution which happened 300 years ago and that was the uh industrial revolution. And that was caused by having reliable steam power. So reliable steam power basically led to uh an explosion in uh physical productivity

and we got factories, we got mass production and it resulted in an increase in efficiency and how fast humans could make physical things and this meant that that many people could get clothing, many many people could get new tools, many people could get furniture where otherwise before it was out of reach for them. And of course, we didn't stop there as a society. And every every single revolution, industrial revolution, or sorry, uh uh technological revolution lays the foundation for the next revolution.

And the next revolution came 120 years later. And that was the that was the age of electricity. And this revolution brought humanity power amplification. And a big bang moment for the the the uh the the electricity revolution was when the first central power plant was created in Manhattan in New York City. And all of a sudden we had industrial level power go from being concentrated in steam engines at plants to being able to be distributed to the users at at the end points. So it became much more efficient to generate power and distribute power. And this led to things like dishwashers, refrigerators, heating and cooling, and many people's homes. We

had light bulbs instead of candles at that time. And so once we could reliably distribute power, that was then used as an infrastructure for the next revolution, which was the information age. It started 75 years ago, and it was a direct result of having power conveniently available within laboratories and and uh homes. And the big bang for that moment happened in 1947 with the invention of the transistor. And all of a sudden within a period of a few generations every many people had telephones, telephones, TVs in their homes and eventually computers.

And when everybody had computers in the home, of course, that laid the groundwork for the internet. And so this revolution amplified our ability to create, communicate, and store collective human knowledge as a spec as a species. And this the the information revolution primed us for the next revolution. And I believe that the revolution that you're going to experience in your lifetime is going to be the agent revolution. And I think it could begin any day. the the big bang moment for that could begin any day. So remember how we all agreed that it's probably not a good idea today to let chat GBT manage your wallet and but also remember that AI performance continues to double every

single year and will continue to double every single year. And so the the Gentic Revolution will be the moment when that moment happens when we have the big bang bang moment for for this one. It's going to be the transition moment when AI leaves the stage where it's just giving advice about whatever to actually taking actions on your behalf. And because we're at a cryp crypto conference, I'm going to focus the rest of the talk on on crypto and finance. So, I believe that crypto and blockchains are the perfect substrate for the agentic revolution. Crypto is programmable money and when you combine that with agents that can make autonomous decisions on your behalf, you

have the ingredients for Agentic Finance. So next, I'll explain exactly what Agentic Finance is and I'll also go into the projects that you should be paying attention to that are live today. So agent fi refers to the market segment of crypto products that have various degrees of autonomy and various degrees of intelligence. And so the way that the way that we do our taxonomy here is that we we place all of the agentic products on this chart. We can place all agentic products on this chart. On the y- axis you have autonomy where at the top you have things that are making direct

transactions actually interacting with the chain moving money. On the bottom on the y- axis you have things that are giving advice not taking action. So that's autonomy. On the xy you have intelligence. On the left hand side you have what we call algorithmic agents. These are agents that are using if then type logic or maybe traditional methods from quantitative finance. And on the right hand of the right hand side of the plot you have AI agents that could be using models that are well beyond what we have available today. So I believe that all products all autonomous products software products especially uh and focusing on this talk agent five products can be placed somewhere on this chart and we've done that. So why pay

attention to this topic now? Um well money is moving into agent 5. At the beginning of the year, there was almost zero assets under management in the space. People were still people became excited as many of you know at the end of 2024. There was a lot of hype. But in 2025, people actually started building real products. And as of October of this year, we now had we had $20 million in assets under management. And now I think today it's around 500 million assets under management.

So during the past year, my team has investigated all of the agentic projects that are out there and we've used most of them and we've curated the ones that we think are actually uh legitimate, the most legitimate and the most useful and we've created the a report called the 2025 agent fi landscape which I'll link to at the end of this at the end of this um presentation. So now I'm going to break down what each of these projects do. Each of these segments are about. So first we have agents that are doing prediction and betting and they're they're partic participating in uh betting markets and or they're doing things like sports prediction or

election outcomes. And these agents need onchain and off-chain data. And of course they're participating in onchain markets. Next we have trading and portfolio optimization agents. And these agents manage users funds or manage users portfolios and pick assets on behalf of their users. And they need onchain pricing data, risk preferences, and potentially things like sentiment and news for their decision-m. Now, we also have information agents that perform sentiment analysis, technical analysis. They can read financial news, financial reports. These agents don't direct directly take actions. they more offer they're more like co-pilots today. So this would be

on that uh that that axis that I that I showed earlier. These would not be autonomous systems today but of course they're well positioned to move into autonomy. Next we have liquidity provisioning agents and these agents allocate liquidity across different dexes and so they need to track returns available across dexes and they need to help their users avoid impermanent loss. And what they're trying to do is basically generate fees, collect fe trading fees for the liquidity providers for their users who who become liquidity providers. And finally, we have lending lending agents. And these these agents are actually this is the most popular segment of all of them. And this is this is a segment that's received most of the assets under management today. and

they're actively uh dynamically allocating collateral, their users collateral and lending protocols so that their users can collect yield uh from interest rate payments. And so these agents need to be aware of what's the yield being generated by different lending protocols, what's the risk of different lending protocols uh for them to do their their actions. So remember I I introduced this sense, think act loop and all of the different agents I just showed you, they need to sense their environment which could use various either onchain or off-chain inputs. They need they then need to think about what they're going to be doing. Um they need to be thinking about their financial strategy they're going

to do and then they execute and they may use algorithms to do their thinking and they may use AI to do their thinking. Most of them are all of the ones that have had product market fit fit are algorithmic agents if they're taking actions. So I'd like to take a moment to to highlight some of the work that my team at Cambrian does and where we fit in the sense act loop. So just like you and I agents need information and not all information is useful. So for example, if you go if you were to open up Twitter right now, most of the stuff that you see in your feed is just trash. It's very low signal to noise information.

Similarly, for agents, if they were to look at all blockchain translation transactions, most of those transactions have nothing to do with their decision-making process. And so, the agents need only specific information. So at a bare minimum, agents need organized and transformed data for their decision-m things like protocol volumes, revenues so that they can get the most yield that that they're trying to go after or maybe whale activity so that they can do copy trading. That's something that we provide at Cambrian, but we also go a step further. So financial intelligence is the biggest superpower power that you can provide an agent because the most well-informed agents can become the most profitable agents and financial

intelligence allows agents to discover alpha and do in make insights faster than a human ever could. So imagine an agent that can scan hundreds of chains simultaneously looking for yield opportunities and going after those yield opportunities without you having to be involved at all. So this brings me to some forward-looking statements and this isn't financial advice. So where do I see things going next? Out of all the agents, it's worth noting that stable coins are the agents that are getting I've already noted that uh the agents that are working on lending protocols are getting the most PMF. And the agents that are working with stable coins, those are also the agents that are working with lending protocols. And so agents that I believe the agents that

can put stable coins to work to basically generate yield are going to have high PMF. And these agents are going to be automatically discovering opportunities across protocols and across chains. And they're basically not you're not going to have to be involved at all for them to go after them. Stable coin adoption is increasing a lot. So in the past 5 years, we went from 20 billion assets under management to 300 billion assets under management under under over the past five years. stable coin transactions have outpaced Visa transaction volume as as of this year and they could replace AC payments in the near future. And the what's really what's driving this is that stable coins have the least risk of all DeFi

opportunities. And so that's what what we're seeing is that Trady is coming on chain through this stable coin through the stable coin market segment because it's the lowest risk. But that's just going to be their initial entry point. Um basically what we're going to be seeing over the next few years according to a survey from uh Ernston Young is that over the next two years we expect to see 24 currently financial 24% of financial institutions are interested in DeFi and they expect that over 75% of financial institutions traditional financial institutions are going to be become interested in DeFi over the next two years. So they're they're starting to experiment with things outside of stable coins.

So, um, I'm out of time, but I'll I'll I'll leave you with this. So, the crypto experience of yesterday was not very good. It was annoying to use, and it was prone prone to failure, but things are about to become much easier and much more profitable for everyone. So, thank you so much. I appreciate you being here today. If you'd like to see our agent 5 report, you can go to this URL here and you can follow us at Cambria Network and you can follow me at zeroxam green. Thank you. >> [applause]