Why AI agents could be the next revolution in DeFi

Source: Bitcoinist

5 Sept 2025 17:37

By nature, new technologies arrive in waves that reinforce each other. Mobile, social, and cloud shaped the previous era. The next will be marked by AI, crypto, and agents - where “architecture is destined” and the user's intention becomes the main interface.

AI is taking over Web3, and it's happening fast.

According to DappRadar, over the past 18 months, AI has transitioned from a novelty to a foundation in crypto: LLMs summarize governance, agents rebalance portfolios, and bots execute on-chain strategies in real time. Investors vote with their capital: by June 26, 2025, AI agent projects had raised 1.39 billion dollars since the beginning of the year, already surpassing the pace of 2024.

Chris Dixon summarizes the macro situation well: AI and crypto are complementary. Blockchains provide ownership, credible commitments, and identity, primitive elements that AI lacks but desperately needs to create open markets for computation, data, and content. According to him, “AI needs blockchain-based computing.” - a16z crypto

More broadly, the industrial impact of AI supports this shift towards agency. Jensen Huang of NVIDIA views AI as the beginning of a “new industrial revolution”, involving new layers of users and new models of automation in finance - Nasdaq

From applications to agents: the infrastructure fades away

The emerging final state is simple to describe but difficult to construct: you express an intention; an autonomous agent composes the stack - data, liquidity, risk checks, settlement - and then executes. Research on agent systems and “Agentic Web” outlines this world where agents pay other agents for data and services, coordinate via smart contracts, and perform transactions without human supervision. The IKANGAI development tools are catching up: frameworks like elizaOS show how to connect LLM agents to DeFi wallets and actions ( “transfer” and “exchange” in natural language ), suggesting a future where the application is an orchestrator of agents.

The data problem: Web3 remains fragmented

Agents thrive on reliable low-latency data. Web3, however, is fragmented by chains, schemas, and sources. Indexing posts and provider documents converge on the same point: raw chain data is chronological and scattered; meaningful queries require specialized indexing, subgraphs, replication, and ETL pipelines - often repeated by chain.

Providers like Goldsky and The Graph help, but even they emphasize the need for inter-chain mirroring, real-time streaming, and composable subgraphs to serve complex applications - exactly what agents will continuously demand. Independent analyses echo the cost of fragmentation for risk and UX in DeFi.

Conclusion: if the UI becomes a box of intent, the heavy lifting shifts to a programmable data layer that standardizes the on-chain/off-chain context, exposes deterministic APIs to agents, and supports low-latency calculations (alerts, scoring, routing) across chains.

Why AI agents are perfect for DeFi

DeFi is native to machines: transparent ledgers, programmable liquidity, and composable contracts. It's an ideal playground for autonomous agents to:

Trade and rebalance via structured prompts (“sell niche assets for ETH if volatility exceeds X”).

Scan the risks (contractual anomalies, oracle drift ) continuously and integrate them into execution.

Arbitrage and market-making through AMMs/CEXs without interface friction.

Govern (draft proposals, simulate results) using on-chain data and forum.

I have always thought that DeFi would be revolutionized by intelligent automation. Academic work on autonomous AI agents in DeFi precisely envisions these roles, linking the decision-making of agents to market microstructure and governance design. Buterin also suggests that the most viable role is AI “as a player” in crypto games, which fits perfectly with the markets.

The emerging landscape: chat-based DeFi platforms

Here are six products based on chat or centered around agents that illustrate the spectrum, from consumer bots to intention-driven execution.

HeyElsa: AI crypto copilot with natural language/voice, aimed at routing, bridging, swapping, lending across chains with secure MPC wallets and safeguards. Projected USP: unified control via chat/voice plus (MPC) custody model for public UX.

Kuvi.ai: Presents itself as Finance Agentique; “Don't trade, hoot.” Text-to-trade execution through DeFi, positioning agents as problem solvers that connect user intent to settlement. Projected USP: end-to-end intent pipeline and cross-domain ambition (finance, identity, gaming).

Igris.bot: Focused on destination-based swaps: you specify the desired outcome ( “end with 2 ETH on Base” ), and the system determines the wallet source, the route, and the cross-chain fees. Projected USP: Focused on the destination rather than the source - reducing the decision-making burden on the user.

Defi App: Intent-based swaps through solver/relayers; routes through multiple aggregators/DEXs; comprehensive documentation. Projected USP: Native execution based on intent (solver model): Users specify the outcomes; off-chain solvers/relayers compete to route through multiple liquidity sources.

AskGina.ai: AI Wallet Companion that can analyze assets and execute on-chain transactions from chat; exists as a web application/mini-app Farcaster. Projected USP: AI Wallet Companion (analysis → action): chat interface that understands your wallet and provides personalized insights.

What the agentic user layer requires as infrastructure

If agents are the new UI, the infrastructure must be refactored for the machines:

Programmable data layer: cross-chain ingestion → standardized schemas → real-time replication/mirroring → deterministic APIs consumable by agents.

Conscious latency calculation: triggers for price/volatility/MEV risk, agent policy evaluation, and pre-trade checks.

Identity & Permissions: wallet-related permissions, cryptographic attestations ( “proof of personhood/humanity” ), and policy guards around agent autonomy.

Safeguards: Vitalik's warnings: restricted APIs, ( “kill switches” ), and alignment layers must be prioritized.

Why it is important ( and why now )

The intent-driven model is gaining traction: users type in goals; agents manage the plumbing. The status quo - clicking through bridges, DEX, and dashboards - cannot accommodate the next 100 million users. Architecturally, the solution is not just a better interface; it is open rails for ownership and programmable data so that many agents - not just a few closed super-apps - can compete on user value.

When the big waves come, they “complement and work together.” AI brings creativity and automation; crypto offers open ownership and incentives; the new devices ( from phones to wallets to wearables ) complete distribution - together forming a user stack that reads like agents by default.

Final reflection

If “read-write-own” was the previous era, the next one introduces “act”: software that acts on behalf of the user. In DeFi, this means agents that understand your intent, assess risks, and settle across fragmented markets - securely and instantly. The winners will not simply provide nice chat interfaces; they will view architecture as destiny and invest in layers of programmable data and incentives that allow agents to thrive at scale.

Warning: For informational purposes only. Past performance is not indicative of future results.