How agents unlock new markets

Source: Sangeet Paul Choudary; Platforms, AI, and the Economics of BigTech (Substack) · 2026-10-11

The problem was the cost of discovering and negotiating such arrangements - particularly with smaller suppliers, where the potential benefits often didn’t justify the time and effort required to negotiate individually.

Walmart now uses AI to conduct these negotiations - engaging suppliers individually, exploring trade-offs involving prices, payment terms, and other commercial conditions, and identifying arrangements that work for both parties.

But there’s an important shift here - a previously uneconomical activity now becomes economical.

This is the larger opportunity in the emerging agent economy. As I mentioned in part 1 of this series, much of the excitement around AI agents focuses on helping buyers find better offers and negotiate better prices.

What if the bigger opportunity is not simply making existing transactions more efficient, but making transactions possible that would otherwise never happen?

No matter how intelligent the buyer’s agent becomes, it remains largely limited to choosing among the offers that sellers have already made available.

When sellers are also represented by agents, they can understand what a buyer is trying to achieve, evaluate what they are capable of providing, and explore different ways of reaching an agreement.

When both sides can understand each other’s needs and constraints, they are no longer limited to negotiating over what has already been offered. They can begin to construct offers that didn’t exist before the interaction.

Consider a manufacturer looking to purchase 5,000 units of a specialized component. The company specifies its requirements, suppliers respond, but none can meet all the requirements. The manufacturer concludes that a suitable supplier isn’t available.

The transaction never happens because neither side has fully explored what the other is willing or able to do.

As a result, many potentially valuable transactions never happen - not because buyers and sellers are fundamentally incompatible, but because discovering and constructing an agreement is too expensive.

Agents could make a much bigger difference here. By reducing the cost of exploring alternatives, negotiating trade-offs, and coordinating agreements, they could make transactions viable that are unrealised under conventional market mechanisms.

Like Walmart, Honeywell has also used AI in procurement to negotiate with suppliers, completing more than 2,500 autonomous negotiations covering over $500 million in spending. These negotiations involve different combinations of payment terms, rebates, and discounts.

In Anthropic’s Project Deal experiment, 69 employees were represented by AI agents in a small marketplace. The agents identified opportunities to buy and sell, proposed prices, negotiated with one another, and concluded 186 real agreements involving goods worth more than $4,000. The experiment was small, but it demonstrated that agents representing different individuals could negotiate directly and arrive at agreements without requiring their principals to manage the process.

Taken together, these developments suggest that three constraints that have historically shaped market activity are beginning to weaken.

With that, a larger opportunity emerges when these capabilities allow buyers and sellers to discover and construct agreements that would otherwise never have been reached.

For this larger opportunity to materialize, agents need to move beyond working with the standard offers to uncover what each party would actually be willing and able to do to make a transaction possible.

On the buyer’s side, an agent needs to understand not just what the buyer is asking for, but what the buyer is trying to achieve. A job seeker looking for a particular role may be willing to consider a different position if it offers greater responsibility. A traveller searching for a particular flight may be willing to change departure dates in exchange for greater flexibility.

On the seller’s side, the transformation is even more significant. A supplier’s agent needs to understand not just what the company currently offers, but what it is capable of providing. A hotel may be able to offer a different combination of services and conditions. An employer may be willing to reshape a role around an exceptional candidate.

The larger transformation requires connecting agents to the operational capabilities and commercial decisions of the organizations they represent.

And these capabilities must be able to interact across organizational boundaries. Buyer and seller agents need mechanisms to discover one another, exchange proposals, verify what is being offered, and arrive at agreements that both parties can execute.

Imagine a manufacturer that needs a specialized component delivered within a narrow window. One supplier can produce the component but cannot meet the delivery requirement. A logistics provider could solve the delivery problem, while another company could provide specialized processing or certification.

Agents could potentially make such coordination economical across a much broader range of transactions. Instead of asking which individual supplier can meet a requirement, the market could discover which combination of suppliers can jointly deliver the outcome.

An agentic participant becomes more effective at finding, negotiating, and executing opportunities within an existing market.

An agentic market can potentially expand the opportunities that become available for exchange in the first place.