Artificial intelligence agents integrated into DeFi protocols are not a passing fad or a mere narrative, but the beginning of a new class of autonomous economic actors capable of executing financial strategies, reallocating capital, and optimizing risk without direct human intervention.
While the dominant narrative argues that this is a speculative cycle similar to the “DeFi summer tokens” of 2020 or the AI tokens of 2023, there is no doubt that we are witnessing a structural shift in the market architecture.
To frame these transformations, three variables currently at play in the market stand out: the maturity of on-chain infrastructure, the availability of AI models with decreasing costs, and the growing institutional participation in digital assets.
The question is no longer whether AI can operate in DeFi, but rather what happens when it does so persistently, with significant capital and programmable rules.
Financial automation in a liquidity scarcity environment
One of the key developments is that since 2022, the global macroeconomic cycle has been characterized by positive real interest rates in the US. The Federal Reserve raised the benchmark interest rate from 0% to the 5%–5.50% range in 2023, making the most aggressive adjustment since the 1980s. This change increased the cost of capital and reduced systemic liquidity, affecting both traditional and crypto markets.
In environments of abundant liquidity, marginal capital efficiency is secondary, so in restrictive environments, it becomes central. AI agents applied to DeFi emerge as a microstructural response to this macroeconomic context: they optimize yield, rotate collateral, execute statistical arbitrage, and manage risk in real time with less friction than human managers. When capital is more expensive, automated optimization gains economic value.
The closest historical comparison is the expansion of algorithmic trading after the 2008 financial crisis. In traditional markets, the percentage of volume traded by algorithms in US stocks now exceeds 60%, according to estimates from exchanges and market research firms. In crypto, this process is just beginning. The structural difference is that in DeFi, the agent not only executes orders: it can also custody, lend, provide liquidity, and govern protocols programmatically.
On-chain evidence: the key data supporting the role of AI agents
The rise of automated agents is clearly supported by metrics that reflect their current state. For example, in 2020, DeFi’s TVL (Total Value Locked) increased from less than $1 billion to more than $15 billion in a single year, according to historical data from on-chain aggregators. Following the 2022 collapse, TVL contracted by more than 70%, reflecting systemic deleveraging.
From 2023 to 2025, TVL partially recovered, but with a different composition: greater participation in liquid staking, decentralized derivatives markets, and overcollateralized stablecoins. The structural difference is that current capital is more concentrated in complex strategies that require dynamic management of collateral and exposure. Protocols such as Aave, MakerDAO, and Uniswap enable carry, leverage, and market-making strategies that can be automated.
An AI agent can, for example:
- Monitor fluctuating rates across multiple pools.
- Realize collateral based on implied volatility.
- Adjust stablecoin exposure based on macro metrics (Treasury yields).
- Execute derivatives hedges on DEXs if the funding rate becomes misaligned.
These tasks are already performed by bots. The novelty lies in the incorporation of predictive models and contextual adaptability, not simply fixed rules.
Regulatory Context for AI Agents
On the other hand, regulators have addressed DeFi under two frameworks: unregistered securities and anti-money laundering. The Securities and Exchange Commission has stated in multiple communications that many tokens could be considered securities under the Howey Test. In parallel, the Financial Stability Board published recommendations on systemic risks in crypto assets.
However, autonomous agents introduce a third dimension: algorithmic responsibility. If an agent optimizes yield and triggers a cascade of liquidations, who is liable? The developer? The operator who deployed it? The underlying protocol?
The evolution of AI agents in DeFi
The structural difference compared to 2020 is that back then the risk was human (excessive manual leverage). Now it could be systemic and automated. This could lead to demands for algorithmic transparency or auditing standards for agents managing third-party capital.
Since the approval of Bitcoin spot ETFs in the US, traditional managers have increased their indirect exposure to crypto infrastructure. Institutional participation shifts incentives: they seek efficiency, reporting, and risk control.
If a fund tokenizes real-world assets and deposits them in DeFi protocols to generate returns, automation through AI agents reduces operating costs. It’s no coincidence that infrastructure firms are investing in automation tooling and programmable custody.
Historically, each institutional entry into crypto (CME futures in 2017, corporate stablecoins in 2019, ETFs in 2024) has generated a leap in product sophistication. AI agents could be the next step, but focused on operational efficiency rather than retail storytelling.
On the other hand, critics of the growth of AI agents in DeFi argue that they don’t add anything substantial compared to existing bots. They also claim that they increase the risk of automated herd behavior. In fact, if the agents replicate similar models trained on correlated data, they could amplify volatility during stressful events.
Furthermore, if regulations impose restrictions requiring direct human identification for every financial decision, the fully autonomous model would lose practical viability.
