Imagine having an employee who could trade on your behalf on a decentralized exchange. This employee doesn’t need coffee breaks, doesn’t sleep, and doesn’t expect to be paid. They can crunch inhuman amounts of data, execute trades, and refine strategies on the fly, without you having to lift a finger.

One day, this employee won’t be imaginary. AI agents – algorithms that act independently and learn from data – are coming for DeFi. These helpful bots could soon be embedded into protocols, executing complex strategies with minimal input.

However, while the idea of set-and-forget trading bots is compelling, AI agent tech is still in its infancy, and DeFi itself isn’t yet ready. Right now, platforms lack the tools to distinguish between AI agents and malicious actors.

As the tech matures, we’ll need new ways to reliably distinguish agents apart from one another and from scams. To understand how this might work, it’s worth taking a look at the current state of AI agent tech.

Dawn of Hype

Interest in AI agents has surged in the broader tech world, thanks partly to generative AI – think ChatGPT – which enables bots to interact with humans in natural language.

DeFi specifically hasn’t seen much in the way of working use cases yet, though projects did start popping up last year in the so-called “DeFAI” sector, which seeks to integrate AI into portfolio management, trading, and other protocol functions.

These promoters say AI agents could analyze data, make trades, and manage assets with superhuman speed and precision. In practice, they could:

Interact with smart contracts to automate yield farming, arbitrage, or governance.

Optimize portfolios, reallocating assets based on factors like market conditions or investors’ risk/reward appetites.

Find trading signals in on-chain activity.

Comb through social media posts and produce sentiment analysis.

One prominent early project is ai16z, not affiliated with venture capitalist Marc Andreessen’s fund, and recently rebranded to Eliza Labs, which offers a platform for building AI agents.

Degens’ interest really kicked off, however, in mid-2024, when Truth Terminal, an AI agent with a dark sense of humor, caught the attention of Andreessen of actual a16z fame, who promptly sent it $50,000 in Bitcoin.

That interaction supercharged a Truth Terminal-linked meme coin and sparked flows into AI agent tokens. The category’s market cap ballooned 222% by the end of 2024 to $15 billion.

However, Truth Terminal turned out to be less an AI agent breakthrough and more of a performance art project-turned-meme coin cult. The episode underscored how speculative and meme-driven the DeFAI space is.

The hype has deflated since, with AI token prices cratering. CoinGecko now has the sector’s market cap sitting at around $5.4 billion, and the sector’s performance was down 81% in early April.

Riding the Cycle

However, though some are now predicting that AI agents will never take off in DeFi, that is by no means true. AI agents are a natural evolution of this space and offer major efficiency gains, so they will absolutely see huge adoption in the very near future. The current hiatus is just the calm before the storm.

One need only look outside of crypto to see the idea of AI agents is becoming commonplace. Car companies like Mercedes Benz use agents to power in-car assistants and enhance navigation systems. Banks use them to improve customer service bots. Wall Street deploys them for real-time financial analysis.

Like all novel tech, AI agents are riding a hype wave so predictable it’s been modeled and copyrighted. The Gartner Hype Cycle – named for the consultancy that came up with the rubric – maps out the fraught route that all emerging technology takes to success, from inflated expectations to troughs of disillusionment. DeFI AI agents may be in that trough right now, but this phase is a natural slump on the road to widespread adoption. This is the time when smart builders work on improvement.

A Lack of Identity

One major challenge that DeFi needs to figure out before it’s ready for AI agents is identity.

Right now, it’s difficult to know whether an AI agent is legitimate or malicious. That opens the door to scams where agents impersonate others to extract funds, manipulate markets, or influence governance, a problem that magnifies as the tech scales.

In traditional finance, identity is solved with know-your-customer (KYC) checks. If you’ve ever opened a bank account, you’ve almost certainly handed over your passport number or scanned your face for a biometric verification.

That kind of system wouldn’t work in DeFi, where pseudonymity is a feature, not a bug. There are a range of tools that will come in handy once AI agents take hold in DeFi, and they won’t require handing sensitive information to a centralized third party.

One option is Web3 domain names. Rather like Web 2.0 URLs, these domains could assign unique, trackable identities to AI agents, allowing users to see an agent’s transaction history and verify they’re interacting with a legitimate entity.

Since these domains aren’t human-readable, you’re not dealing with a string of numbers and letters – just something like “alicestradingrobot.”

Imagine again that you’ve got an AI agent, or more likely, many of them, running your trades. So will everyone else. There’ll be millions of them! We need to know who’s behind the bots executing high-stakes transactions at high speed.

Of course, Web3 domains aren’t the only solution. Users will probably stack ways to track AI agent verification, from whitelists at the protocol level to zero-knowledge proofs that could allow agents to prove their credentials without revealing their underlying code.

The promise of AI in DeFi is too good to ignore. AI agents will keep evolving, and the infrastructure must evolve too. There is much that still needs improvement, but a simple place to start is by giving AI agents a verifiable name.

This article does not contain investment advice or recommendations. Every investment and trading move involves risk, and readers should conduct their own research when making a decision.