How AI Empowers Financial Services on Crypto Rails

TL;DR

• AI and crypto are fundamentally compatible from a data-stream perspective.
• There is a capability–security dilemma among DeFi agents. If AI agents want to perform more complex tasks, they need to have greater control over users’ assets.
• Maintaining a competitive edge is difficult for a data analysis AI agent, as valuable insights tend to become less effective when they are widely used.
• A good DeFi AI product need to fit the following criteria:

1. Can precisely understand users’ intentions and translate them into correct on-chain transactions.
2. Has the ability to find the best DeFi protocol for users, including finding the best yield and more.
3. Do not require users to deposit their assets, only trigger transactions for users to approve and they could use a smart contract wallet to simplify the process.
4. Has the ability to become an all-in-one asset management platform in the future.

1.What is the fundamental logic behind the collaboration between AI and Crypto

To start, we need to understand why and more importantly how AI and Crypto provide services to each other. . The logic of AI would be “Input data — AI — Output data”. As for input and output, it could be any form of data, including text, image, 3D models and more. For different purposes, we have different AI services to process the data. Blockchain is essentially a trustworthy, standardized, shared database updated in real time. These features allow the creation of a record of value transfer activities and establish a value exchange system on top of a blockchain.

Source: Inception Capital

This diagram briefly shows three ways AI and blockchain interact along the data flow between unstandardized off‑chain information (left) and standardized on‑chain data (right). In “Type 1,” AI ingests off‑chain inputs — like user instructions or external data feeds — and automatically trigger on‑chain transactions (e.g., personal AI assistants or oracle services). In “Type 2,” AI lives entirely on chain, reading and writing standardized on‑chain state to optimize yields or route trades in real time. And in “Type 3,” AI analyzes on‑chain data and pushes insights back off‑chain — powering tools like smart‑wallet trackers or on‑chain warning systems. Together, these three types of AI enhance the ability of dapps and provide great services for users.

2. The Key Problems, Capabilities and Dilemma of AI+Crypto financial services

2.1 The key problems to solve and what AI can do

First, the growing complexity of DeFi protocols has raised the entry barrier for users. DeFi protocols often expose dozens of knobs — leverage levels, collateral ratios, slippage tolerances, staking durations, etc. — that a user must configure correctly.

AI agents can dramatically improve the usability of complex DeFi protocols. In particular, LLMs can serve as natural language interfaces for DeFi protocols, wallets, and dApps. Users can simply ask an AI agent to perform financial tasks, lowering the barrier to entry for non-technical users. We are already seeing natural language command interfaces in crypto. Some DeFi “abstraction” platforms let users interact via conversational prompts and allow users to use AI agents to swap tokens, make deposits and even launch meme coins.

The interface difference between Jupiter and the AI agent for Jupiter in Griffain

Source: Jupiter, Griffain

Second, users cannot constantly monitor on-chain activity or take timely action. Markets and protocol conditions may shift when users are not convenient to do the transactions.

Users can set high‑level “if‑this‑then‑that” rules in plain English. An AI agent can watch key metrics — price feeds, liquidity pool ratios, interest‑rate changes — in real time and then submit transactions on users’ behalf, avoiding delays in manual monitoring.

An interesting component of this type of AI is that since on-chain data is standardized and utilizes smart contracts, blockchain can act as a unifying layer that allows multiple AI systems or agents to cooperate in asset management. On top of the asset standard and smart contract, different AI agents can interoperate, share data, and coordinate tasks autonomously on-chain and provide enriched financial service with the composability of smart contracts. For example, AI can use on-chain data to find the protocol that provides the highest yield and complete the on-chain transaction for users. AI can also find the best trading path for users.

Third, the complexity of on-chain data makes it difficult for newcomers to comprehend. Raw on‑chain data (transaction logs, contract events, LP token balances) means little to a newcomer. AI can aggregate and translate this into clear dashboards and provide simple reports like “Your total exposure across four pools is $12 k with average APR 8.7%”. AI‑powered tools can automatically plot your portfolio’s performance, protocol metrics, or historical yield curves, highlighting trends.

Source: Inception Capital

2.2 The dilemma in AI+Crypto financial services

There is a dilemma in AI financial services, that is the “capability-security” dilemma.

Since some AI services involve changing on-chain data, access to users’ assets is necessary, which opens up a potential attack vector. If the services want to provide more services to users, they need to have higher permissions of the user’s asset. For example, some AI projects may require users to deposit their asset into the platform’s address. Some services don’t require users’ permission, but they can only provide data analysis functions.

For type 1 and 2 AI services, there are two groups of products. One is with “medium security and medium capability,” the other with “low security and high capability.”

“Medium security and medium capability” means the services require users to connect their wallet to the dapp and use AI to trigger on-chain transactions. These transactions are approved by users so they have control over whether to execute it or not. However, it can be difficult for users to verify the transaction and there is a chance that the dapp will take control of users’ assets.

“Low security and high capability” means users need to deposit their asset into the protocol and all asset management transactions will occur within the services’ address. This approach requires users to fully trust the service providers. However, these types of services can conduct more complex transactions and even can provide greater active management of the user’s assets.

For type 3 AI services, they are “high security and low capability.” Since they only provide information and they don‘t involve in any on-chain activities, it is unlikely that user’s assets will be at risk because of the AI services. Below is a brief mapping of some AI crypto financial services.

In the following graph, we group AI + crypto financial products according to three types of AI and the extent of their control over users’ assets.

Source: Inception Capital

3. In-depth analysis of Three AI financial services

3.1 Griffain

Griffain is an AI agent platform on the Solana which focuses on the DeFi sector. It coordinates a network of autonomous agents that can execute complex on-chain actions like token trading, NFT minting, portfolio queries, and even social tasks (e.g. tweeting) on behalf of the user.

Griffain acts as an AI agent marketplace. Users can create personal AI agents by minting them as soulbound NFTs, which serve as the agent’s identity and configuration. Griffain offers both “personal agents” (tailored to individual users) and “special agents” (pre-built agents for specific tasks) to enhance the user experience.

Griffain’s key advantage is the already established, enriched AI agents’ choices. There are bots for main solana projects including pump.fun, Jupiter and Backpack. There are also bots for research, including deep research and editions.

Sources: Griffain

The biggest disadvantage of Griffain is the unsmooth user experience. Users need to deposit SOL into Griffain and then swap the SOL into “Energy” and use the Energy to pay for the chatting. Moreover, users need to deposit money into the protocols to do all the asset management operations. These two boundaries will dramatically decrease users’ interest. They can charge a monthly subscription fee like every other AI service and trigger users’ wallets for the on-chain activities.

3.2 EnsoFi

EnsoFi provides AI agents for users to earn yield and lending across multiple blockchains. The core idea is that users will be able to deposit assets and earn yield seamlessly across chains. EnsoFi will handle routing the funds to where they earn the best returns, whether that’s lending protocols on Solana or liquidity pools on Sui, etc. EnsoFi also has a product called EnsoFi DeFi Agent System (EDAS). It is an agent that continuously scans DeFi markets to discover optimal yield strategies and can execute those strategies automatically on behalf of users.

EnsoFi’s advantage is its combination of cross-chain flexibility. By being multi-chain from the outset, it can tap into yield opportunities wherever they arise, not limiting users to one ecosystem. If executed well, EnsoFi could significantly simplify DeFi investing. Users would effectively have a personal AI fund manager handling their lending and farming for them.

However, the strategy that EnsoFi’s AI agent has is not appealing enough and it may be limited by the AI’s ability. One of the strategies is Sol LP with 5% APY. They also have incoming BTC yield but we can’t see the APY for now. Sol LP could be a too simple strategy and we need a better strategy for users. It may be because it still doesn’t have enough data for training to find the best yield and it requires longer time to optimize.

Source: EnsoFi

3.3 AgentXYZ

AgentXYZ is an AI trading platform and terminal that deploys a suite of 28 AI agents to assist cryptocurrency traders. It offers real-time technical analysis, on-chain data monitoring, social sentiment tracking, and fundamental market insights — all integrated into one interface. In practice, AgentXYZ’s agents scan a wide array of indicators and deliver actionable insights or alerts to the user.

AgentXYZ’s top advantage is its comprehensive and unified approach to AI trading assistance. By combining technical analysis, on-chain analytics, and social sentiment into a single feed, it gives traders a 360-degree view of the market without needing multiple tools.

Source: AgentXYZ

The biggest challenge for AgentXYZ is proving the reliability and performance of its AI strategies in live trading. Moreover, users could prefer an all-in-one platform more compared to this kind of data service. Traders will adopt it only if the agents’ recommendations consistently lead to profitable or at least sound decisions. However, it is hard for AI platforms to keep providing valid information. If the target users are people who barely know how to use data to make trading decisions (which are the majority), they would like to use services that can do the decisions for them.

Source: Inception Capital

4.Conclusion

First, the core competitive advantage of AI financial services lies in the precision of an AI model’s ability to understand users’ intentions. It is a matter of how well the AI model can understand users’ needs and identify the most suitable on-chain operations. It requires the companies to provide enough trading data on users’ words about the related on-chain activities. The algorithm that can find the best trading path or do other tasks is not the most important thing because a blockchain is an open database and everyone can access the same data. The competition of algorithms is the competition of data. In a world with the same trading data, many companies can have the algorithm with the same ability.

The response of Griffain and the Hive for the same question

Source: Griffain, the Hive

Second, another core competitive advantage is the ability to find the best DeFi protocol for users. This requires judgment from DeFi experts and it is hard for algorithms to make the best choice. It is because the yield of DeFi protocols are sometimes fake due to unstable token distribution and even and sometimes can’t be delivered. Moreover, choosing the right DeFi protocol is always hard and it requires a lot of experience.

Third, projects that require users to deposit assets may not be able to compete with those that only require users to connect their wallets. Depositing assets is an extra step and requires significant trust from users. Most importantly, smart contract wallets can improve the ability of AI agents and the advantages of depositing assets will disappear. The EIP-7702 in the latest Pectra Upgrade has allowed externally owned accounts (EOAs) to be temporarily converted into smart contract accounts. The process of smart contracts wallet adoption has sped up.

Fourth, the project that provides all-in-one services can be the final winner in this section. By all-in-one, it refers to the services that include personal assistant, on-chain optimization and data analysis. It is easier for users to do everything on the same platform. Some projects provide the bot marketplaces for users to create and use different AI bots for different tasks. It also fits the thesis of intent-centric and lowers the barrier for users in crypto.

A good AI+Crypto financial service should fit the following criteria: first, it can precisely understand users’ intentions and translate them into correct on-chain transactions; second, it has the ability to find the best DeFi protocol for users, including find the best yield and morel; third, it does not require users to deposit their assets, only trigger transactions for users to approve and they could use a smart contract wallet to simplify the process; fourth, it has the ability to become an all-in-one asset management platform in the future.

REFERENCE

• IOSG,DeFi + Ai = DeFai, https://medium.com/iosg-ventures/defi-ai-defai-5764e3d3a74e
• Griffain, Griffain Doc, https://griffain.com/docs/getting-started
• EnsoFi, EnsoFi Doc, https://ensofi.gitbook.io/ensofi-1
• AgentXYZ, AgentXYZ Litepaper, https://docsend.com/view/jhpwgepzx6c4atzq

Learn more about Inception Capital at our website & by following us on X @_inceptioncap

Disclaimer: This post is for general information purposes only. It does not constitute investment advice or a recommendation or solicitation to buy or sell any investment and should not be used in the evaluation of the merits of making any investment decision. It should not be relied upon for accounting, legal or tax advice, or investment recommendations. This post reflects the current opinions of the authors and does not necessarily reflect the opinions of Inception Capital, its affiliates, or individuals associated with Inception Capital. The opinions reflected herein are subject to change without being updated.