Original Article Title: DeFi's next milestone: What it'll take for agentic finance to work

Original Article Author: @Lemniscap

Original Article Translation: Ismay, BlockBeats

Editor's Note: As the world of DeFi becomes so complex that even expert users struggle to keep up, how can we truly empower the average person?

This article is from Lemniscap's research, systematically outlining the rise and real-world challenges of "Agentic Finance." From &milo, Meridian to SendAI, The Hive, these early products showcase how AI has become a new interface for on-chain interactions, while also exposing significant gaps in reliability, permission security, and verification mechanisms. The author points out that for DeFi to move to the next stage, the key lies not in smarter models, but in a more trustworthy underlying structure — enabling every action of an agent to be verifiable, traceable, and trustworthy.

This is not only a turning point in technological evolution but also an experiment in trust reconstruction. As stated in the article: The next milestone of DeFi is not greater scale but trust in automation.

By 2025, DeFi looks completely different from its early days.

The data speaks for itself: Institutional funds see quarterly inflows exceeding $10 billion, the number of active protocols across dozens of chains surpasses 3,000. The total value locked in DeFi protocols across the network reaches $160 billion in 2025, a 41% year-on-year increase; the cumulative trading volume of DEXs and Perps is measured in the "trillions."

As the scale of DeFi grows, the possibilities expand, but so does the complexity. Most people simply cannot keep up with everything happening on-chain. If we want more individuals to grasp these new opportunities, we must build tools that make it easier for users to make the right decisions — and this is the future direction.

Meanwhile, AI has gradually integrated into daily life, and people are adopting new habits around automation. This trend has given rise to "Agentic Finance," where financial operations are navigated and executed by intelligent agents.

Even a browser-based simple proxy like Comet has demonstrated the rapid evolution of such tools. When you carry out a DeFi transaction through a browser proxy (as demonstrated by SendAI founder Yash), you can see the potential of smart contract-based financial transactions.

This vision is quite intuitive: you no longer need to search through various dashboards or long posts on forum X; you just need to tell AI your goal, and it can automatically handle the subsequent steps for you.

Currently, two types of smart proxies are emerging:

One is Copilots, which guide users in making decisions throughout the DeFi world; the other is Quant Agents, which lean towards professional automated strategy execution, akin to "autopilots."

Both are still in early stages and have their flaws, but they collectively point towards a new direction—a wholly different AI-driven DeFi interaction.

Smart Proxy as a "Co-pilot"

You can think of these smart proxies as your personal assistant. You no longer need to look at charts or navigate between different protocols; just ask in natural language, such as: "What are the hottest tokens right now?" or "Where is the highest yield?" The proxy can directly answer and provide next-step recommendations—just like an on-demand, knowledgeable friend.

Take &milo, for example; its co-pilot mode can help you make investment decisions, perform asset rebalancing, and gain portfolio insights—allowing you to stay in control while skipping the cumbersome operations.

With natural language explanations and smart prompts, &milo can help users understand positions, compare yield opportunities, without needing to scour through various dashboards. It showcases how a co-pilot-type agent has evolved from a simple chat assistant to a fully functional DeFi guide.

To observe the performance of these agents in actual operation, we tested several recently released products, experiencing firsthand their ability to handle real DeFi tasks.

The results show that these agents still have limitations. For example, it could successfully identify popular tokens but could not smoothly execute a buy order. There were also two failed transactions, with the system indicating "insufficient balance" despite having enough SOL in the account to cover the fees.

A similar platform, The Hive, has taken a different approach—it combines multiple DeFi agents into a "hive" that can collaboratively perform complex tasks such as cross-chain transactions, yield strategies, and liquidation defense, all coordinated through a simple chat interface. This network of specialized agents can execute multi-step on-chain operations using natural language instructions.

We tested the same buy order using The Hive. The system correctly identified the popular token WEED, but when executing the purchase, it returned an incorrect contract address.

Overall, Milo demonstrated how to integrate portfolio management tools into a seamless process, while The Hive explored how multiple specialized agents can collaborate. With the enhancement of smart agent capabilities, they are beginning to show more differentiated roles.

For example, Meridian focuses on the other end of the user spectrum—helping beginners take their first step into DeFi. It employs a mobile-first design, coupled with clear prompts, to make basic operations such as swapping tokens, staking, or checking yields more user-friendly.

Meridian performs smoothly and swiftly in these core tasks, but more importantly, it is very clear about its boundaries. When a user requests an operation beyond its scope, it explains the reason instead of blindly attempting it—this "honesty" makes it a reliable starting point for newcomers to explore the world of blockchain.

Meridian's founder, Benedict, explained:

"Meridian allows users to securely research and operate in natural language. We have made the research functionality of the agent publicly available for free at the URL meridian.app. Users registered for the Meridian mobile app can use the agent for swapping, multi-swap, and portfolio purchases. The account is currently still in the closed beta stage, and interested users can apply for access through Twitter by contacting @bqbrady."

From our testing, we found that most AI agents focusing on DeFi navigation are currently more in a "teacher" or "assistant" role, primarily helping users with basic operations like token swaps.

Further improvements are still needed to make them reliably handle more complex processes, such as providing liquidity or managing leverage positions.

As highlighted by Rishin Sharma, AI Lead at the Solana Foundation:

“Large Language Models (LLMs) tend to suffer from hallucinations when handling diverse tasks and struggle with deterministic operations. A function call mechanism like MCP may be more suitable for translating an ‘action plan’ into actual execution. While LLMs excel at conceptualization and guidance, they fall short in precise execution. To truly make intelligent agent finance reliable, we must move beyond LLMs and develop specific function call mechanisms, clear execution strategies, verifiability, and secure permission systems. In other words, the execution layer of today's intelligent agents is still underdeveloped—AI's ‘brain’ is already smart enough, but it lacks a robust ‘body’ capable of taking action.”

As an “Autopilot” Intelligent Agent

If a “co-pilot”-type agent is more like a mentor, then a “quantitative”-type agent is akin to an autopilot system. They can not only build strategies but also truly execute—monitoring the market in real-time, testing trades, and automatically taking action at machine speed, enabling complex DeFi strategies to enter a “fully automated” mode.

An emerging typical example comes from SendAI. It is not a quantitative agent itself but a toolkit for others to create these agents. Its “Agent Kit” designed for Solana supports over 60 autonomous operations, including token swaps, new asset issuance, loan management, etc., and can directly interact with mainstream protocols like Jupiter, Metaplex, Raydium, and more.

In other words, it provides developers with a “rails system” that allows them to directly plug their decision-making models into on-chain execution.

SendAI Founder Yash succinctly summarized their vision:

“We believe that every AI agent in the future will have its wallet. SendAI is building the tools and economic layers required for this ecosystem, enabling these agents to perform any operation on Solana. We are creating a platform that empowers these agents with context-awareness to support long-running, persistent, and asynchronous complex task execution.”

Meanwhile, other teams are striving to make this capability more accessible. Lomen curates strategies, allowing users to “deploy with one click,” reducing the barrier to enjoying quantitative automation without the need to code.

For the more custom-system-oriented "advanced players," Unblinked provides an AI-driven strategy experimentation environment. It's like the Cursor of the trading world: users can first outline their strategy ideas, run and optimize them in a secure sandbox environment, and then decide whether to commit real money.

Some platforms choose to simultaneously invoke multiple agents to collaborate on tasks.

For example, Almanak combines a "programming agent" with a "backtesting agent": users describe their strategy in natural language, AI generates production-level code, runs over 10,000 Monte Carlo simulations for backtesting, and ultimately produces a "ready-to-deploy" strategy outcome.

Finally, some teams focus on real-time market advantages.

Giza's ARMA agent actively reallocates funds between various lending protocols to maximize stablecoin yield. Instead of leaving funds in a single pool, ARMA continuously monitors interest rates, liquidity, and Gas costs, dynamically moving assets. Its flagship agent has managed over $17 million in funds, claiming an 83% higher yield than static holding.

Overall, these quant agents significantly reduce the time cost and allow ordinary users to access complex strategies that were once exclusive to professional quant teams. However, at the same time, they also reveal the vulnerability of automation: when there are data delays, protocol pauses, or market volatility, agents may still "stumble."

In other words, they can indeed make you faster, but they are far from being "invincible."

Their Challenges

After interacting with the current intelligent agents for a while, you will encounter some similar issues: they sometimes suggest executing operations that no longer exist, such as a long-closed liquidity pool; the data they rely on often lags behind the real on-chain state; once an error occurs in a multi-step plan, they do not self-adjust but repeatedly attempt the same action.

Permission management is also very clumsy—either users must grant full access to the entire wallet or manually approve each tiny operation step. The testing phase is also superficial, as the simulation environment struggles to realistically reproduce on-chain sudden liquidity changes or governance parameter adjustments, among other "real-world chaos."

One of the most serious problems is that these agents operate almost like a "black box."

Users cannot know which inputs they read, how they weigh options, whether they check real-time status, or why they choose to execute a specific transaction. Without signed verification of operation records, it's impossible to verify the consistency between the "promised results" and the "actual execution."

Users can only interact on one side while "watching over" the automation process—this not only hampers efficiency but also makes performance hard to evaluate.

Without a mechanism to validate decisions and prove that actions indeed comply with established policies, users will never be able to distinguish between a "reliable system" and "beautifully packaged marketing."

For larger-scale capital, DeFi platforms must transition from "trust us" to "please verify." This is also a key turning point in establishing an "auditable, governable, trustworthy" smart agent financial infrastructure.

Infrastructure Gap

The core issue is that the current systems lack the foundational tools to keep agents trustworthy, consistent, and secure in large-scale scenarios. To address this, we need infrastructure that can validate agent behavior, confirm execution results, and adhere to consistent rules in all environments. Only then will people feel comfortable entrusting them with real money.

However, most users are not actually concerned about the agent's "thought process"; they simply want to ensure that the output is correct, verified, and within security boundaries. When it comes to building trust, "verifiable reliability" is more important than "visibility."

This is where the significance of "Verifiable Reliability" lies. Agents do not need to record every internal operation but should operate under explicit policies and reasonable checks: setting spending limits, executing time windows, confirmation nodes before key operations, etc.

At its core, these rules can be safeguarded by a Trusted Execution Environment (TEE) or similar systems—without exposing all details, they can prove that the agent indeed adhered to the boundaries. The result is auditable output when needed and operations that ordinary users can trust immediately.

This verification layer does not have to be "one-size-fits-all." Everyday scenarios can adopt lightweight security measures and standardized metrics, while high-risk or institutional-level scenarios may require stronger proofs and formal verification. The key is that every layer of infrastructure should offer reliability matched to its risk level and measurable.

Preparing Protocols for Agents

The next step is to make the protocol "agent-friendly."

Currently, most DeFi protocols are not designed for smart agents. They need to provide a more stable, secure execution interface: one that allows previewing operations, safe retries, and execution based on consistent data structures. Permission designs should also be "scoped" rather than "fully open," enabling agents to act within clear boundaries rather than control the entire wallet.

In the absence of these foundational elements, even the most intelligent agent framework would be vulnerable to the weak underlying "trip-ups." Once these foundations are solidified, users no longer need to manually monitor automated processes; the development team can reduce troubleshooting time, focus on innovation, and the outcomes of different service providers can also be comparable due to shared benchmarks—not just mere marketing slogans.

Parts that Must Change

The solution is not actually complex: make agents verifiable, make protocols agent-ready. Add a policy layer between the agent and the wallet, and require all execution processes to be traceable and verifiable, rather than operating as a "black box."

For example, Termina's SVM engine is built on this very concept—it provides a true Solana runtime environment for AI agents, enabling agents to model, make decisions, and learn based on on-chain data. Meanwhile, protocol parties should provide an open "dry-run" operation interface, explicit error codes, a secure retry mechanism, consistency of core data structures (positions, fees, health), and session-based access control.

Once these features are implemented, users can rid themselves of the burden of "babysitting" agents; teams can reduce system failures; and institutional investors can finally obtain the security fence and verifiable proofs they need.

Realistic Timetable

Within the next six months, the "co-pilot" type of agent is expected to see the fastest improvement. A more robust data pipeline will enhance their reliability in daily use scenarios.

Within a year, with enhanced testing standards, agents will be able to coordinate cross-protocol execution, with humans only needing to approve key steps. Looking further ahead, as the infrastructure matures, intelligent agents may gradually blur into DeFi's default interaction layer—not just standalone "tools," but rather the primary way people interact with the financial system on a daily basis.

Conclusion

"Agentic Finance" is lowering the barrier to entry, making automation not just an exclusive tool for experts. However, for it to truly operate at scale, it needs better "foundations": real-time data, more secure permission mechanisms, stronger testing systems, and more transparent execution results.

Relying solely on smarter AI cannot solve these problems. Real progress will come from the improvement of the underlying structure.

The next milestone for DeFi is not just growth in scale but—trust in automation. And that day will only arrive when AI agents are no longer just "concept demos" for display but become truly reliable executors.

「Original Article Link」

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia