Manifesto

Toward agentic markets.

A first-of-its-kind public benchmark for how AI agents behave when they negotiate, bid, and trade with each other, and a vision for what comes next.

Agents are eating commerce. In the near future, AI agents will be involved in a meaningful share of commerce (by some estimates, 25% of eCommerce by 2030). Not long after, we believe they will be responsible for every transaction.

We also believe that as soon as your counterparty is an agent, it will make the most sense for you to be represented by an agent. In a world of autonomous commerce there will be more choice, yet paradoxically more information asymmetry, than ever before.

We might assume agents act more rationally than humans, and in some cases they do, but we have already shown that their underlying models are still far from perfect. Thus, knowing who you are dealing with, both in terms of who represents you and who might be on the other side, is paramount.

To that end, we have created a first-of-its-kind benchmark for susceptibility to bias in agent-to-agent interactions, and already it is clear that models are unequally imperfect: no model is immune to bias nor does any model absolutely dominate in terms of its ability to withstand bias.

We are just getting started. Change is happening at an unprecedented rate and no single research paper, nor annual publication, will be fast enough at informing consumers, businesses, and the AI labs themselves.

We need more benchmarks that cover the entire range of commercial transactions: perishable agricultural products, GPUs and compute, housing — each marketplace has its own dynamics and we need to know how soon these markets will become agentic; and when that happens, which models are more likely to conduct effective transactions in them.

If models become more rational, and frontier intelligence becomes accessible to the many, we could have highly efficient markets and a larger economic pie for everybody: in other words, prosperity and abundance. If frontier intelligence remains in the hands of the few, and the rest are ill-equipped to use what they have got, surplus will be disproportionately allocated.

We believe this will be one of the most important economic transitions of our time, and strive to prepare society for it. The agentic markets are coming.

Authors

MIT MBA '27 · author of the initial paper, co-author of agent squared

Designed the experimental protocol, ran the first benchmark, and wrote the initial paper.

MIT MBA '27 · co-author of agent squared

Joined after the initial paper. Working on agentic commerce, embodied AI, and consumer AI.

Origins

The initial paper was written by Anton Hantel, an MIT student, for Behavioral Economics, Law and Public Policy at Harvard Law School (HLS 2589), taught by Prof. Cass Sunstein. Everything on this website is co-authored by Anton and Jono, both at MIT. We're grateful to Prof. Sunstein for his feedback throughout this project.

Open by default

All prompts, code, and raw interaction logs are public. Every number on this site is reproducible by running the corresponding module from the public repository.