Alex Svanevik, chief executive of the blockchain analytics firm Nansen, has predicted that AI trading agents will surpass human traders in both numbers and effectiveness within about two years. The forecast underpins a significant strategic redesign at the company he co-founded, as it transitions from a focus on data insights to enabling direct trade execution.
During an appearance on a daily program, Svanevik described the core ambition in simple terms: allowing agents to handle trading of every asset class entirely on-chain.
Achieving this, he noted, requires three simultaneous changes—moving from analysis tools to execution capabilities, expanding beyond cryptocurrencies into all asset types, and transferring the role of selecting positions from individual humans to autonomous agents.
The platform has already made progress on this path.
Since rolling out trading features, Nansen has handled just over $500 million in volume.
Activity is increasingly diversified; roughly ten of the top 15 perpetual futures contracts by volume on the service involve non-crypto assets, including SpaceX shares, the S&P 500, gold, silver, crude oil and Brent crude.
The firm also recently added Robinhood support, with trading expected to go live shortly.Svanevik contends that agents hold structural advantages over typical retail participants.
He characterized human traders as often “quite sheep-like,” tending to crowd into identical positions.
Agents, by drawing on a wider range of inputs, models and tools, generate greater diversity in approaches.
That same adaptability creates vulnerabilities, however.
Because the systems rely on inference rather than hardcoded rules, their data feeds can be contaminated or deliberately manipulated.
Nansen is working to address these exposure risks before releasing more independent agent features to the public.
Current economics remain challenging.
One internal agent produced only $23 in gains while incurring $700 in inference expenses, illustrating the early-stage cost imbalance.
Still, the company points to a key differentiator: more than 500 million labeled blockchain addresses compiled over six years.
This dataset enables agents to identify on-chain activity that would otherwise remain hidden.
In Svanevik’s view, centralized exchanges face obstacles in adapting.
Licensing constraints and established, profitable business models limit their flexibility with agent-based systems.
Nansen is taking a measured approach itself, withholding fully autonomous agents until they complete backtesting and simulated trading rather than releasing incomplete tools.
He expects meaningful performance edges to emerge in the coming months, even as the broader shift toward agents dominating trading activity is projected within two years.The remarks point to a wider evolution in markets, where on-chain transparency combines with machine-driven decision-making across traditional and digital assets.