What Are Autonomous AI Agents Doing In Crypto?

AI agents in crypto are one of the hottest narratives of the moment. They use advanced artificial intelligence algorithms to help traders and investors navigate the complex world of digital asset markets. 

Trailblazing early adopters already use AI agents to automate a surprisingly wide range of activities, such as smarter, autonomous trading, portfolio and risk management, yield farming optimization in decentralized finance, valuation of non-fungible tokens, fraud prevention, and more. 

What is a crypto AI agent?

Most crypto AI agents utilize natural language processing algorithms to understand their users. If you ask an agent, “What’s the current market cap of Bitcoin?”, it’ll be able to interpret that question and know what to do to find the answer. 

In addition, crypto AI agents use application programming interfaces that allow them to retrieve key market data in real-time so they can provide responses that are informed by the latest information. As with generative AI chatbots, their basic intelligence stems from the underlying LLMs that power them. These LLMs are trained to understand users’ questions and prompts, and by using APIs to obtain the latest real-world information, they can deliver an accurate response for crypto users. 

They also possess advanced “reasoning” skills, so a user could ask an agent: “Should I buy or sell my Bitcoin, based on current market conditions?”. The agent will understand the question, then leverage its APIs to find out the latest market sentiments, based on real-time prices, trading volumes, and social media posts, so it can generate an informed response. 

Going beyond traditional generative AI chatbots, crypto AI agents can perform actions on behalf of their users. So if a user gives it full access to their crypto wallet and permission to execute transactions, they can tell it to trade on his or her behalf, based on specifications such as their profit targets and risk tolerance. The user might instruct it to “buy SOL if the price drops to $150” or “sell SOL at $200”, and the agent will immediately undertake these actions if the price reaches those targets. 

How are people using crypto AI agents? 

Crypto AI agents have a surprising number of use cases, and they’re being used to automate just about anything someone can do with digital assets. 

1: Automated trading

Whereas traditional trading bots are structured around rigid, rule-based strategies, AI agents are far more capable. Their ability to reason and continuously learn, based on market movements, sentiments, and the available liquidity, means they can make more accurate trading decisions, take more risks, and increase profits for traders. 

An example of this is Giza’s ARMA, which is an AI agent that’s designed to track of the latest yields on hundreds of DeFi markets across the most popular blockchains and manage complex yield farming strategies for investors. Among other things, it can spot profitable opportunities, execute trades, and dynamically balance user’s portfolios based on the latest market data, helping them to maximize their DeFi returns. Although still experimental, ARMA already boasts more than $1 million in assets under management, having performed 77,809 transactions on behalf of its users. 

2: Portfolio management

Smart investors have complex portfolios of assets spread across multiple blockchains, taking advantage of liquidity pools and staking, lending, and borrowing. AI agents can reduce the cognitive burden for investors, automatically rebalancing their portfolios in line with their risk mitigation strategies, ensuring they’re protected against sudden, volatile market movements.   

Tools such as Griffain and Anon can respond to these happenings in real-time, making smart decisions to protect investors from these risks. Griffain helps users to allocate assets across multiple DeFi platforms and manage risk exposure in real-time, while Anon utilizes machine learning to dynamically adjust users’ positions based on the latest market data. 

3: Real-time fraud detection and monitoring 

Fraud and security vulnerabilities are very real risks in the crypto world, and AI agents can help mitigate these dangers by analyzing transactions and identifying anything suspicious. For instance, they can detect wash trading activities designed to artificially inflate trade volume, flag risky smart contract interactions, and scan social media feeds to alert users to any security incidents they need to be aware of. 

That explains the popularity of platforms like Numerai, which is an AI-powered hedge fund that uses data sourced from its community of data scientists. It applies machine learning to this data to help users identify risks and execute trades while avoiding unnecessary risks. 

4: NFT valuation and trading

AI agents are also expanding into the market for NFTs. They work by analyzing the metadata of NFTs and monitoring market trends and price movements to try and identify if an asset is overvalued or undervalued. They can also make intelligent predictions about future NFT prices and then take action on these insights, buying and selling these assets autonomously. 

SingularityNET is one of the best examples of this. It has created a decentralized AI marketplace where developers can create and deploy AI agents. It includes several that have been customized to monitor NFT assets, taking into account the current floor price for the specific collection, its rarity and traits, and trade volumes.  

Is there anything bad about AI agents? 

Crypto AI agents sound like a wonder to behold, but there are indeed some negatives to consider before diving in at the deep end. These include technical constraints and questions about ethics and regulations. 

One of the main technical challenges of crypto AI agents is the integration of off-chain data with blockchain-based systems, which is an issue that also affects the wider Web3 industry. The solution for agents will likely be similar to what we’ve already seen, namely, data oracles. For instance, projects like Ocean Protocol are working to create AI-powered oracles that can supply smart contracts with real-world data in a trusted and secure way. 

As for the ethical concerns, a number of mitigations have already been posited, such as AI explainability tools that provide greater transparency into how AI systems come to their decisions, enabling greater human oversight. 

Because AI crypto agents are such a new concept, there’s little by way of regulation around the technology. Most AI agents don’t sit well with existing regulations, which require some kind of accountability. That’s because most crypto AI agents are decentralized, which makes it hard to hold anyone responsible for their actions. However, as with any new technology, new regulations will inevitably follow to solve these issues. 

As far as issues around trust are concerned, these have largely been solved. For instance, Giza’s ARMA is built on a self-custodial architecture, ensuring that users always have exclusive control over their funds, even while the AI manages them. Neither Giza nor ARMA can ever access the user’s private keys, and the session keys that allow it to make trades on the user’s behalf operate under strict time and protocol limitations. Moreover, users can instantly revoke full permissions at any moment, giving them further reassurance that they’re still in control. 

So what’s next? 

Given what they’re capable of, it’s unlikely that AI agents in crypto are just a fad. As more users experiment with the concept, we can expect to see a new wave of DeFi projects emerge that come with their own, integrated AI agents, simplifying access and paving the way for everyone to autonomously trade, lend and manage assets. 

It’s not easy to predict anything relating to digital assets, but one thing we can say with certainty is that the crypto industry has always embraced innovation. AI automation is about as innovative as it’s possible to get right now, and that’s why we’re confident we’re going to hear a lot more about it in the months and years to come.