Artificial intelligence and blockchain are among the most promising technologies of our time. Yet their intersection has not always been a good match. In 2017, during the first wave of ICOs, many projects surfed on this technological convergence, promising wonders to investors.
Heated by these experiences, the market remained sceptical for a long time about initiatives combining AI and crypto. But today, a new generation of more concrete projects is emerging, notably with the advent of AI agents capable of interacting directly on the blockchain via wallets.
These AI agents, veritable intelligent automatons, can perform a variety of tasks autonomously: publishing on X (ex-Twitter), hosting live chats on TikTok, or even executing transactions. Their field of action extends well beyond blockchain, even if this technology is opening up new prospects for them.
In contrast to traditional bots, which are confined to predefined tasks, these new-generation agents have advanced cognitive capabilities. They can analyse complex situations, learn from their mistakes and adapt in real time to achieve their objectives.
More impressively, these agents can be organised into veritable virtual teams, where each of them contributes their specific expertise to accomplish a common mission.
In this survey, we will examine how blockchain can serve as a playground for these AI agents, what concrete benefits they bring to users, and what the current limitations of this emerging technology are.
Blockchain: a strategic tool for AI agents Blockchain technology is emerging as a fundamental pillar in the AI agent ecosystem. With its distributed and transparent networks, it guarantees real-time traceability of transactions while ensuring their irreversibility - barring exceptional cases of hard fork. This architecture revolutionises coordination between network players, eliminating the need for mutual trust.
To finance their operations and monetise their services, AI agents require efficient payment solutions. Blockchain perfectly meets these needs with four major advantages:
Instant and unlimited creation of wallets Virtually instantaneous and secure transactions Execution of smart contracts for complex operations The possibility of making micropayments at lower cost Although not indispensable, blockchain is establishing itself as the preferred infrastructure for value transfers by AI agents. The latter, greedy for computing power, operate mainly off-chain, publishing their results only on the blockchain.
A major challenge is emerging: verification of outsourced computations. Two innovative projects, Ritual and ORA, are currently developing solutions to certify on the blockchain the accuracy of AI operations performed off-chain.
Concrete use cases for crypto AI agents In the field, AI agents are already demonstrating their potential through a host of innovative applications. At the forefront are "influencer" agents that enliven social networks to promote their tokens, while others specialise in optimising DeFi returns.
Several platforms are vying for this emerging market. Virtuals stands out with its AI agent tokenisation solution and revenue redistribution model. For its part, Olas is banking on the synergy between multiple agents, while Spectral facilitates the deployment of smart contracts via a conversational interface.
Improving the user experience is another major area of development. As an example, layer 2 Mode has launched a veritable AI agent "app store", centralising various automation solutions for DeFi strategies.
Even more ambitious, some projects such as ai16z and Sekoia aspire to revolutionise asset management by creating fully autonomous investment funds on the blockchain.
The integration of AI agents also extends to DeFi infrastructures themselves. The Edge oracle, developed by Chaos Labs, is a perfect example of this trend: by harnessing AI to analyse off-chain data in real time, it dynamically optimises the parameters of the DeFi protocols that adopt it and enables them to adapt in real time.
For more details on oracles, see our analysis of them.
Future use cases: between innovation and caution The integration of AI agents into DeFi is paving the way for promising innovations. Edge, a pioneer in this field, is already foreshadowing what could be the next generation of decentralised financial protocols.
Among the most eagerly awaited developments is the automation of MAs on Uniswap V4, where AI could adjust parameters in real time. Lending protocols such as Morpho could also benefit from automatic rebalancing of their vaults. However, industry experts are calling for vigilance: a malfunction in these automated systems could have significant financial consequences for users.
A more secure avenue is emerging with the emergence of AI "solvers". These specialised agents promise to revolutionise the user experience by drastically simplifying transactions on the blockchain.
The principle is simple: the user expresses their intention, and the solver takes care of the execution, assuming all operational risks. Payment is only made once the transaction has been successfully completed, guaranteeing optimum protection for the user.
In this context, we could see the emergence of an ecosystem of specialist AI solvers, collaborating to orchestrate complex operations. The Khalani Network project perfectly illustrates this trend by developing an infrastructure dedicated to cooperation between solvers.
Current challenges and limitations Despite the enthusiasm generated by AI agents in the crypto ecosystem, several major obstacles persist. First and foremost, there are legitimate concerns about the reliability of these emerging technologies. A recent incident, documented on X (formerly Twitter), highlighted the potential financial risks associated with delegating portfolio management to these agents.
While their usefulness is proven in certain areas, notably the animation of communities on social networks and the automation of simple tasks, experts recommend extreme caution when it comes to asset management. The question of the real autonomy of these agents also remains unresolved: how can we distinguish a truly autonomous agent from a system assisted by human operators?
One notable advance, however, comes from the collaboration between Flashbots and NOUS Research, who have developed an innovative solution: an agent operating in a secure execution environment (TEE), thus guaranteeing its integrity and imperviousness to external intervention.
In this context of effervescence around AI agent-related projects, caution is called for. The sector, marked by fierce competition, is seeing a proliferation of initiatives whose relevance has yet to be demonstrated. Between the race for technological innovation and the battle for users' attention, the market is still struggling to identify the truly promising players.
VIDEO
