When AI Agents Need Their Own Bank Accounts: The Blockchain Solution

A deep conversation with Gautam Dhameja, founder of Block Deep Labs, exploring why autonomous AI agents will need blockchain for identity, trust, and transactions when they stop asking humans for permission and start making their own decisions

November 19, 2025

13 min read

By Rachit Magon

Listen on

Today we're talking about software that stops asking for permission. AI agents that make their own decisions, hold their own money, cut their own deals, and operate completely independently.

The question nobody's answering clearly: how do you keep autonomous software accountable when there's no human in the loop?

The answer, according to researchers and builders in the space, is blockchain. Not for the hype, not for the tokens, but because when agents start talking to other agents, they need something they can't fake: cryptographic proof of identity, ownership, and transaction history.

Meet Gautam Dhameja, founder of Block Deep Labs. He's spent nearly eight years in the web3 trenches, led solution engineering at Parity Technologies, architected blockchain solutions for enterprises, and before that built cloud native IoT systems at Microsoft. He's seen emerging tech cycles from the inside, from IoT hype to blockchain skepticism to AI explosion.

Today, he's watching these worlds collide and explaining why the boring infrastructure work happening now will matter when agents become economic actors.

Key Takeaways: AI Agents Meet Blockchain

The Core Problem:

• Agents need identity, ownership, and the ability to transact without human gatekeepers
• When two agents interact, they don't trust each other because they're machines
• Blockchain provides accountability without requiring trust in a central authority

What Actually Counts as an Agent:

• Conversational AI like ChatGPT: predicts text, no decisions, no actions
• Workflows like N8N: some automation, but human driven reasoning
• True agents: combine reasoning, decision making, and autonomous action

Why We're Still Early:

• Right now, only trading bots are using blockchain in any meaningful way
• Agent commerce, agent identity, and agent to agent transactions are still theoretical
• We're at the stage where agents are figuring out individual tasks, not society level coordination

Q: You've been building in emerging tech spaces for years, from IoT to blockchain to AI. What drives you to keep chasing the next wave?

Gautam: The short answer is that life keeps giving me chances to try new things.

The longer answer is the pursuit to solve problems with what is the best possible solution out there. To give you a very simple example, when I was at Microsoft back in the winter of 2016, we were trying to do a new system on the supply chain side. Suddenly there were some data inconsistency issues. Some devices suddenly went missing.

I started looking into what could be a solution to data inconsistency when there are multiple parties involved, when they don't trust each other, when there are these vendors and warehouses and all of these things. I stumbled on Ethereum and smart contracts. It was very new at that time.

It's about solving problems with something that has a more direct implementation instead of doing a workaround. That has been my approach since the very beginning. Sometimes it takes a lot of time to understand those new things and implement them. Sometimes it's rather straightforward. That just keeps me going.

🔥 ChaiNet's Hot Take: Most people pick a technology and find problems to solve with it. The best builders pick problems and hunt for whatever tech actually solves them, even if it means learning something completely new.

Q: You first encountered Ethereum when it was just heating up. You've seen the whole industry evolve from nascent to crazy bull run to somewhere in the middle. How has your perspective changed?

Gautam: I keep following the Gartner hype cycle since the very early days. When I started in IoT, it was at the top. I've been following the way they look at things, and it's the same for blockchain as well.

It started with skepticism, then there was hype, then fundamental technologies. First it was Bitcoin is good, what is blockchain? Then it became Bitcoin is bad but blockchain is good. Then blockchain doesn't scale, it's bad. Then okay, now it scales. Should we do an L1 or L2?

It keeps evolving. With those trends, there are two sides. There's tech crypto, there's money crypto. To be honest, I don't follow too much of the money crypto side of things. I don't comment on that. There are people posting charts and this is going up, down. That's not me. I don't understand those things as well. I do not have a finance background.

I like to keep following the tech side of things. It's evolving very organically, very naturally. We stumbled on something very interesting with solving the double spend problem. Then we generalized it to solve other problems. Then we started looking at the trilemma, then solving that trilemma using different techniques. Finally, right now where we are with programmable money and stable coins, things are evolving very organically, I would say.

But still, we are yet to go mainstream. I still believe that.

🔥 ChaiNet's Hot Take: Tech evolves in cycles, not straight lines. The people who survive are the ones who ignore the hype graphs and focus on solving actual technical problems, whether it's fashionable or not.

Q: There are voices saying blockchain's major use case might be found in AI. What are your thoughts?

Gautam: I wouldn't comment on whether it's the primary or major use case of blockchain. That record has been playing for quite some time now. Sometimes every year there is a primary use case of blockchain.

But I would say yes, there is definitely some convergence, some connection between the two. On one side, you're talking about autonomous agents and software that thinks. On the other side, you're talking about technology that keeps things accountable.

We started with making blockchain accountable for humans. The initial use case of Bitcoin was that banks were not being accountable, so we made Bitcoin so finance could be more transparent, more clear, and more fair to people.

But that has generalized now. We can make machines accountable as well. That's where I see there's this underlying layer that just makes sure everything is cryptographically secure. On top of that, whether you put humans or machines or banks or institutions, those are just use cases.

🔥 ChaiNet's Hot Take: On one side, software will start working on its own. On the other side, how do we keep it accountable? Blockchain already solved that problem for humans. Now we're just applying the same solution to machines.

Q: For non technical folks, what exactly is an AI agent? How is it different from ChatGPT or an AI tool?

Gautam: Just a small disclaimer. I'm very new to this. I've mostly been into web3, but now I'm learning AI. So I'm an AI beginner, I would say.

My understanding right now is that there are three parts.

First, you have a conversational AI tool like ChatGPT, which is pretty much just predicting the text you're looking for.

Second, you have something called a workflow, which is more like what N8N is offering. You can say, okay, this is my workflow. When this happens, do this, then that, then that. There is some sense of autonomous stuff happening behind the scenes.

Third, it comes to the point where we're looking at workflows driven by reasoning from the machine, not reasoning from the human. That is what I call an agent. If I'm having an agent, I want it to think. I want it to actually take a decision on my behalf and then do some action based on that decision.

The first conversational AI tools, they're doing the reasoning but they're not making a decision and they're not doing an action. The second one is workflows, which in some cases do reasoning but not in all cases. They're doing some action based on reasoning from the human. The third thing is where you're combining reasoning, decision, and action. That's an agent, essentially.

🔥 ChaiNet's Hot Take: ChatGPT is a smart parrot. Workflows are automated checklists. Agents are the first one that actually decides what to do next without asking you first.

Q: When does blockchain actually become necessary in this agent world?

Gautam: At least in my initial understanding, I think blockchain comes a little bit more downstream.

To give you an example, when a human is interacting with an agent, if the human knows that on the other side I'm talking to an agent, and there could be things going on because an agent is reasoning on a particular subject. There could be a coding agent, there could be an e commerce agent. They're goal oriented and they have limited reasoning skills on the other side. I pretty much do not think blockchain is needed because we can make agents accountable if we're interacting with them.

But when an agent is talking to another agent, they do not trust each other because they're machines. They don't know what the meaning of trust is. They need something that they can keep each other accountable for. But they cannot keep each other accountable. They need some third party.

That third party cannot be, and they don't even trust that third party essentially. Now this goes back to the initial use case of blockchain. You have multiple parties. They don't trust each other. They want agreement. They need a layer that can get them that agreement.

These multiple parties now become multiple agents. So when two agents or more than two agents are talking to each other, I would say blockchain becomes essential in that scenario. But when humans are involved, I still do not see the need for a blockchain.

But I could be wrong because I'm very early into this.

🔥 ChaiNet's Hot Take: Humans can judge if an agent is lying. Agents can't. When machines start negotiating with other machines, they need math, not trust. That's where blockchain stops being optional.

Q: Agent commerce is being talked about as the next big thing. But are we actually there yet?

Gautam: We're all very early. The only agents using blockchain right now is very nascent. There are trading use cases being solved, but that's about it. The more deeper use cases, when there are two agents actually transacting, that's still theoretical.

Commerce is a very, like, as society developed, first we started to communicate with each other, try to address our feelings or communicate our feelings. Then the barter system sort of came in.

If we compare it to human society, we're still at that stage where there are humans independently trying to figure life out, but the society itself has not developed. So first that needs to develop, then communication happens, and commerce happens, and all of that.

🔥 ChaiNet's Hot Take: Everyone's talking about agent economies like they're around the corner. Reality check: we're still at the stage where agents are learning to not crash when given basic tasks. Commerce requires society. We don't have agent society yet.

Q: When did you first realize blockchain and AI agents could converge?

Gautam: My understanding has been evolving. I still remember back in 2017 when I just entered into web3. This was the major ICO era at that time. I literally started web3 in the middle of ICOs.

One of my friends and colleagues back at Big Chain DB, we used to sit next to each other, and he was building bots for trading. In his free time, he would just create some bots. This was not an AI bot, this was an if else condition bot. If the price goes up, sell it. If the price goes down, buy it. Something like that.

That kind of stuck in my head. Okay, this guy is just sleeping and his bot is just running on an AWS server and making money for him.

Then ChatGPT happened in 2022. Then I connected that idea. If this was an agent, he wouldn't have to write so many if else conditions. He could just give a prompt. So instead of that bot being created by an engineer and used by an engineer, that could be used by anyone.

That's where it kind of clicked that you can have AI do things for you. In this case, it was trading because it's more straightforward for us in crypto. But that's when it clicked. You can have another brain which is more like a smaller subset of goal oriented stuff.

🔥 ChaiNet's Hot Take: The lightbulb moment wasn't about technology. It was about accessibility. When your trading bot stops requiring an engineer to code every decision, suddenly everyone can have one. That's when things get interesting.

Q: A lot of people still don't understand why you need blockchain for identity or trust. Why can't it be done by a simple database or cloud system?

Gautam: Let me give you the broader understanding of this. I did a lot of work in the decentralized identity space in the beginning of my web3 days.

It's not that you're offering your identity and you need a blockchain for that. You have your passport, you have whatever. There's an angle that you don't want to show all that information. You want to say you're more than 18 years, then you should just have a claim on your identity anchored. That's all fine. That's on the user side.

But now think of the organization. Let's say you're a bank or a government. You essentially want them to trust that this is your identity. This is why they store it. If Facebook is asking for your identity or if a bank is asking for your identity, they want to store it on their side because they keep going back to it.

But if they get a layer or an external thing that they can trust, and they will not have to store your identity, that's where blockchain comes in. It's not for storage or something. It's for making sure whoever needs to trust it gets to trust it without storing it.

Now the same thing comes to the agent side as well. The other agent either will make sure who you are, which agent you are, because there could be multiple agents. In the cases of agents, it's very easy to copy some identity. If you're putting it in a database and you're not securing it with a private key or with a fingerprint or something like that, what are the chances that it gets hacked and copied?

You want to anchor it on something that cannot be hacked. If you're talking to an agent that I have for doing something and you have an agent for doing something, these two agents are talking to each other, they should verify each other's identity without storing it. Because storing identity again makes it more malicious.

To verify, you need some external third party that is not centralized, that cannot be hacked.

Rachit: So for someone who doesn't know much, tomorrow when there are thousands of agents talking to each other, we would need a way that all the agents, we would need an Aadhaar of the world of agents. And that Aadhaar cannot be run by a Meta or a Google. That Aadhaar needs to be something which is run by everyone together. That's where identity comes in and that's why it cannot be done by a database.

Gautam: Exactly. The other thing is also, when we're talking about Aadhaar, we're talking about people living in India. We can trust each other's Aadhaar because we are affiliated towards the same institution, the government of India. There's a central authority which is verifying, and we trust that authority.

But the government of Germany will not trust the Aadhaar. So there has to be something more. That's why we have passports, then, that have international acceptability.

Similarly, when agents are there, they are across boundaries, across countries. They don't know what a country is. So they need something more global and, most importantly, something more secure. That's where blockchain comes into the picture.

🔥 ChaiNet's Hot Take: Databases work when everyone agrees on who's in charge. Agents don't have governments. They need a global identity layer that no single entity controls. That's not a feature. That's the entire point.

Q: Walk us through a simple real world example. What is an AI agent on the blockchain actually doing in practice today?

Gautam: In practice, we're still very early. Yesterday I was just reading a Reddit thread about why do we need a blockchain for an AI agent. There are two angles to it.

One is some people are saying we don't need one because if you trust an API, you should just go with it. Some people are saying why should you trust an API when you've already seen some disadvantages of that side.

Keeping all that aside, practically, I think right now we only have trading agents. People are experimenting. It's more tactical today. But it's the first step of tomorrow.

The moment agentic commerce hits, payments would become very obvious. Stable coins would become very obvious. The moment agents start communicating to each other, identity would become very obvious.

These are things that aren't happening yet, but we're nowhere close to what can happen. That also gives a very interesting perspective.

🔥 ChaiNet's Hot Take: We're at the "trading bots on blockchain" stage. Like saying "the internet is for email" in 1995. Technically true, massively understating where it's going.

Q: On one side, thousands of people are losing their jobs because software development has changed. On the other side, there's so much to be done in AI that there's still tons of work. What's your advice for developers and entrepreneurs?

Gautam: I would say there are two things.

First, think of this as an extension of you instead of as a replacement of you. If you're worried that AI will take your job, then you're thinking AI is going to replace you. But think of what can you do that it can become an extension of you.

If you become a programmer that takes advantage of AI, then it becomes an extension. But if you become a programmer that only does stuff that AI can also do, then it is replacing you.

So move up the ladder. Become a smarter version of yourself. AI can still be dumber than you. It's a catch up race, in my opinion. AI wants to become like humans, but it has not become yet. So there is always a time or gap where we can be and AI can always be dumber than us, at least for the time being.

Eventually it might catch up, and then we'll all go back to making more music, making more cool stuff that we should be doing.

Second, I think it's a good time to come out of the comfort zone as well. It's an exciting time to try new things, get more knowledge out of things. At least I'm taking this as an opportunity that okay, there's another thing I have to think about and come out of the comfort zone and see what can be done.

That's what I would suggest others as well.

🔥 ChaiNet's Hot Take: The developers who survive aren't the ones who code faster than AI. They're the ones who use AI to do things they couldn't do before. Upgrade or get replaced. Simple as that.

Final Thoughts: The Infrastructure Before the Revolution

Gautam's perspective: "We're at the stage where agents are figuring out individual tasks, not society level coordination. But when that coordination starts, blockchain stops being optional."

The bottom line: Right now, we're prompting ChatGPT to write LinkedIn posts while talking about agent economies. The gap between conversation and reality is massive. But that gap is closing.

When agents start needing to prove who they are, hold their own money, and make deals with other agents, all the boring blockchain infrastructure being built right now suddenly becomes critical. Identity protocols. Stable coin rails. Cryptographic verification systems.

This isn't hype. It's infrastructure. The kind of infrastructure that looks useless until suddenly everything depends on it.

For developers and entrepreneurs, the opportunity isn't in betting everything on agents or blockchain individually. It's in understanding how they fit together. The stack that emerges when software needs to be both intelligent and accountable.

We're early. Trading bots are just the beginning. But the pieces are coming together. And the teams building at the intersection, solving boring problems like agent identity verification and cross agent payment rails, those are the ones building the foundations of whatever comes next.

Q: How can people connect with you and continue learning?

Gautam: I'm pretty active on social media. You can find me on Instagram at Geekbaat, where I break down blockchain and emerging tech concepts in layman language. I also wrote a book called Beginning Blockchain, and I host a podcast on Spotify called Blockshots.

I'm always reachable if you want to chat about blockchain, AI agents, or where these technologies are headed. Feel free to reach out.

Final words: We're moving from a world where we use apps to a world where agents work for us. Blockchain isn't just a technology choice. It's the foundation that gives these agents identity, ownership, and trust. Whether this becomes the next internet or not, we don't know yet. But the good thing about being early is there's a lot of work to be done. If you upskill, if you learn this technology, if you get out of your comfort zone, you can definitely build for that future and survive the next 10 years.