Autonomous AI agents now drive 19% of on-chain activity, according to a new report from DWF Ventures. While these agents excel at narrow, predictable tasks like yield optimization, they lose to humans by up to a five-to-one margin in dynamic environments such as open-ended trading. Industry leaders like Coinbase CEO Brian Armstrong believe the “agentic economy” could surpass the human economy, but researchers caution that significant infrastructure development is still needed for agents to scale.
New research found autonomous software agents now drive more than 19% of on-chain activity. These systems plan, decide, and execute transactions without direct human input.
However, they lose to humans by up to a 5-to-1 margin in open-ended trading, according to a DWF Ventures report. In decentralized finance, agents manage yield strategies, liquidity, and portfolios.
Total value locked in agent-managed positions exceeds $39 million. Most deployments are still in early testing stages.
For example, the Giza agent earns users 9.75% annually by moving stablecoins between lending platforms. This outperforms yields on protocols like Aave and Morpho.
Agents struggle when objectives are not clearly defined. “Agents thrive when the objective is narrow and the parameters don’t move often,” said Xin Yi Lim of DWF Labs.
This makes yield optimization an early proving ground. They cannot yet reason and adapt to real-time market changes.
MoonPay chief engineer Neeraj Prasad warned agents can be as capable as humans with proper context. He also stated they can be more competent, harder working, and sometimes malicious.
Ethereum developers are working to ease complex on-chain tasks for agents. A new standard allowing multiple simultaneous actions was recently backed by the Ethereum Foundation.
Coinbase CEO Brian Armstrong predicted the agentic economy could surpass the human one. He tweeted it could drive unprecedented demand for stablecoins.
Most current agentic activity involves bots performing narrow work. True agentic activity remains a minority share of the 19% figure.
“A realistic timeline is five to seven years before agentic volume meaningfully rivals human volume in any major financial vertical,” Lim explained. On-chain may get there first due to its permissionless infrastructure.
Agents fall short in areas requiring contextual reasoning and narrative awareness. “Where they fall short is open-ended trading,” said 0G Labs chief growth officer Aytunc Yildizli.
Closing the gap requires more than better AI models. Yildizli emphasized the need for cryptographic proof, trusted execution environments, and decentralized infrastructure.
