With self-executing smart contracts and tamper-proof data ledgers, autonomy has always been a core feature of blockchains. DeFi brought significant attention to the power of this technology by linking its autonomous infrastructure with financial products, such as decentralized exchanges and lending facilities. 

Since DeFi’s breakthrough in 2020, the power of on-chain AI has advanced significantly, and the most recent use case to capture the industry’s attention has been AI agents. Some of the most popular use cases of agents thus far has been to improve the efficiency and user experience of DeFi, leading to the emergence of “DeFAI.”

What Is DeFAI?

In 2020, the term “DeFi” was created to describe the nascent group of on-chain applications that enabled users to trade, lend, borrow, earn yield, and more in a permissionless manner. While this evolution has been seen as a major advancement in terms of user freedom and decentralization, DeFi apps have generally lacked the necessary UX to become adopted on a widespread, global scale.

Using DeFi applications as a foundation, “DeFAI” incorporates task automation via AI agents to abstract away user-facing complexities. Instead of the user having to navigate to the app/website, connect their wallet, sign requests, and confirm transactions, this is all handled by an AI agent. With the implementation of AI agents, users will be able to simply type in a command such as “borrow $100 of USDC.” In seconds, the agent will have scanned various lending markets to find the best deal, initiated the transaction, and received the borrowed funds, all within seconds. 

Compared to the current process of looking through and interacting with DeFi apps, DeFAI presents an infinitely more simple and intuitive approach, accelerating the adoption process and bringing a more efficient financial system to users around the world.

Another key feature of AI agents is their ability to evolve over time. Unlike bots, which are also primarily used for task automation, agents can be used to observe and “remember” vast amounts of data, and use that to improve their performance and efficiency.

DeFAI Use Cases

For example, if someone wants to test out a new trading strategy, they can program an agent to start out making trades based on certain criteria. However, over time, the agent can “learn” based on its trading history, external price data, trading history of other users/agents, and more. This process can potentially create more effective trading strategies to fit any risk profile.

Trading agents are one of the most common early applications within DeFAI, and have appeared already on Avalanche. In fact, one of the early successes within the budding Avalanche DeFAI ecosystem has been Ket, an AI agent that specializes in trading. You can find a record of Ket’s trading activity here.

In addition to trading strategies, DeFAI agents can usher in a new age of on-chain finance by using their skills in a plethora of ways while greatly reducing the need for people to perform repetitive tasks through their wallets. Some examples include:

• Actively hedging long-term/large positions

• Optimizing money market activity Managing liquidation riskMaintaining a certain LTV ratioManaging borrowed assetsPaying back debtEntering lend/borrow positions based on interest rate fluctuationsManaging leverage exposure via automated looping

• Managing liquidation risk

• Maintaining a certain LTV ratio

• Managing borrowed assets

• Paying back debt

• Entering lend/borrow positions based on interest rate fluctuations

• Managing leverage exposure via automated looping

• Optimizing yields Switching between lending positions/platforms to earn the highest yield for a certain asset/basket of assetsAllocating funds to specific yield opportunities based on rate fluctuationsManaging basis trades to earn funding rates across various DEXs

• Switching between lending positions/platforms to earn the highest yield for a certain asset/basket of assets

• Allocating funds to specific yield opportunities based on rate fluctuations

• Managing basis trades to earn funding rates across various DEXs

DeFAI and Avalanche

While the thought of having a bunch of customizable, efficient AI agents collecting data and performing tasks for users non-stop is appealing, this type of on-chain environment requires infrastructure that can scale massively. Thankfully, the Avalanche network was designed with scalability at its core, boasting features such as:

• Up to 4500 tps 

• Sub-second finality

• Fully customizable L1 chains with built-in, seamless interoperability

Avalanche is also an ideal training ground for agents. Its robust ecosystem of successful DeFi projects can be used to not only facilitate transactions for AI agents, but also provide comprehensive data histories that agents can use to refine their performance. 

As the AI agent space continues to develop, Avalanche's flexible and unique infrastructure can support the growth of a strong ecosystem of agents within DeFAI and beyond. With the implementation of Avalanche9000, the Avalanche network significantly lowered its barriers to entry (including reducing deployment costs by 99.9%), enabling more devs to build highly-customizable Avalanche L1 chains. This unlocks new possibilities for builders to design products for agent-based, as opposed to human-based, activity.

Kite AI

As the first AI-focused L1 in the Avalanche ecosystem, the unique infrastructure of Kite AI is bringing new efficiencies to Avalanche.

In short, Kite enables coordination between AI products such as agents, models, and data. Their customized Proof-of-AI consensus mechanism creates a positive environment for all participants by ensuring that they are acknowledged and compensated for their contributions. 

By aligning incentives in this manner, Kite creates an ecosystem for AI builders that fosters growth and development of AI-related products, including those focused on DeFAI. 

For example, let’s say there’s an AI agent whose responsibility is to collect comprehensive on-chain data history from DeFi apps. Using Kite AI, that agent can then allow access to that data for certain models within the Kite ecosystem. Those models can then be used by other agents to make financial transactions on behalf of their users for a fee. In this case, the data-collecting agent would receive a portion of that fee, creating a flywheel effect that results in an ecosystem of revenue-generating agents. 

The above example is one of many ways that Kite AI can serve as a hub for AI products on Avalanche, leading to tangible benefits within DeFAI and beyond. 

Kite is off to a red-hot start on Avalanche – their testnet went live on February 7th and saw activity from over 250,000 wallets within its first 3 days. This marks a major step for AI on Avalanche, and it will be exciting to see the resulting efficiencies for users, builders, and all parties involved.

‍