a16z Frontier Partner: Agent should rebuild a machine-driven Uber-style transaction network

Author: [Guy Wuollet](https://x.com/guywuolletjr/status/2120200572020437283?s=46), a16z Frontier General Partner

Compiled by: Gu Yu, ChainCatcher

Uber provides you with a simple way to express what you want: a ride service from one place to another. The network finds drivers willing to provide that service at a price you can accept. More drivers increase the availability of rides for passengers, and more passengers give drivers a reason to join.

I believe that agents will make it possible to build such networks around a wider range of goods and services. They can help people express more complex preferences and find suppliers willing to meet those preferences. Some of these exchanges are already happening, but prices or complexity keep most people out. Others are simply not arranged due to the workload involved. Agents will also make exchanges that we have yet to think to request possible. All of this presents opportunities for founders.

Consider a simple and familiar example: organizing dinner with a group of friends. Everyone has different schedules, budgets, and opinions on where to eat. Today, one friend usually ends up managing the group chat: collecting these preferences, searching for restaurants, and going back to the group when available options don’t work. Meanwhile, restaurants have tables to fill on Tuesday nights, but their ways of reaching a group that is still flexible in planning are limited.

Your personal agent coordinates with your friends or their agents and comes to an agreement on a request: arrange dinner for six at Mission, under $60 per person, anytime between 5:30 and 8:00, anything but sushi. The restaurant's agents respond with quotes based on their available tables. One restaurant offers a better price for a 5:30 seating. Another puts together a family-style menu for the group. Your agent compares these offers based on everyone’s preferences and completes the reservation within the limits you’ve agreed upon.

A restaurant with available tables at six can directly send an offer to a group willing to eat earlier, before they decide where to go. The group has preferences and budgets. The restaurant has capacity and a price it is willing to accept. What’s missing is a network that allows their agents to find each other and come to an agreement on a purchase.

Uber demonstrates how a network can bring buyers and suppliers together around a relatively simple request. Agents can extend this model to requests for more complex purchases, just like arranging dinner for a group. Most networks are built for humans. We need to build them for machines. Founders can create venues where agents bring these requests, and suppliers compete to fulfill them.

### **Building Networks**

Crypto has already provided a model for organizing these requests using intents: users specify an outcome, and providers compete to deliver it. For example, on [UniswapX](https://developers.uniswap.org/docs/liquidity/uniswapx/overview), users sign an order specifying the tokens they want to trade and the terms they will accept. Providers compete to fulfill that order, seeking liquidity from different sources. Agents can help extend the intents model to everyday commerce. A large language model can turn a conversation into requests that suppliers can act on, and agents can search for quotes and negotiate details. This could make even small purchases, like a dinner for six, cheap enough to arrange in this way.

The same network effects that make Uber useful could also develop around these more complex purchases. Each restaurant that joins adds another potential match for diners, especially when a group has specific requirements. More groups give restaurants more opportunities to fill tables, making it worthwhile to connect them with a booking system that responds with tailored offers. As participation grows in a city, the network becomes useful for a wider variety of requests.

A founder can start with dining in one city and build from there. The initial product needs to save people enough time or money to bring their friends. Restaurants need to see bookings they wouldn’t receive otherwise. These results keep both sides coming back and inviting others.

I believe a consumer-facing agent can be free to use and make money through a model similar to payment for order flow. Users specify what they want and what they are willing to pay, and suppliers compete to fulfill the request. If the agent proposes arranging dinner for $300 and finds a restaurant willing to provide it for $270, the network can keep the difference. It can also charge a fee for routing that purchase to the restaurant. Revenue from these transactions can cover the reasoning costs for consumers. Users should understand how their agent gets paid and how it makes choices between offers. Whether they are willing to delegate purchases will depend on their trust in these choices.

### **Why These Businesses Can Sustain**

As AI makes software easier to replicate, the network of people and businesses willing to transact through it becomes a more important source of advantage. Some of the best businesses of the internet era, including Facebook, eBay, and Uber, become more useful as more people join.

An agent commercial network can develop the same kind of advantages. Competitors can build similar software, but it still needs to bring together buyers and suppliers who can reliably transact. In the restaurant example, this means having enough local restaurants willing to make offers and enough groups ready to book.

Making these transactions reliable is part of building that network. A restaurant needs to know that an agent is authorized to commit the group’s funds. Diners need to know that accepting an offer locks in that table at the agreed price. The network will need a way to verify the agent’s authority, record the terms of the reservation, and arrange payment. If the group cancels or the restaurant fails to fulfill the reservation, both sides should know what will happen.

Founders can build these services alongside the market itself. As buyers and suppliers transact, the network can form a record of who delivered what they promised and a set of processes for resolving disputes. A network that helps participants find good offers and reliably complete purchases is the network they return to for the next transaction.

### **What Else Can Founders Build?**

Dining is my toy example. There are many other networks that agents can help build. Some possibilities:

Travel. A personal agent coordinates a group’s schedules and budgets, then negotiates with hotels and other providers for a trip that works for everyone. Flexibility in dates or destinations creates space for offers that fixed searches would miss.

Business procurement. A company’s agent stays online around the clock, negotiating future supply and price commitments with suppliers. It compares delivery schedules and payment terms alongside prices, helping buyers lock in costs and giving sellers more predictable demand.

Computing power markets. Agents match workloads with available capacity based on hardware requirements, location, and timing. Businesses can also purchase future capacity and hedge computing costs as their demand grows.

Arbitration. Parties agree to use a shared forum to resolve disputes faster than existing legal systems. Each buyer and supplier adopting the same rules makes the forum more attractive to the next participant.

Physical device networks. Agents purchase services from drone networks or data from connected devices. Each new provider expands the network's coverage or adds a service that other participants need.

Collaborative health research. People pool health data and participate in crowdsourced research. More contributors support larger-scale studies and research into questions that smaller groups cannot answer.

There are also people-centered network opportunities. As intelligence becomes cheaper, people will still need ways to establish identity, vouch for someone’s qualifications, and build trust. Credential networks, membership clubs, and other organizations may become more valuable as more work happens outside traditional institutions.

The opportunity lies in finding exchanges that people want to engage in but currently spend too much time or money coordinating. Some of the biggest ones will almost certainly be exchanges we have yet to imagine. Start with a specific market, get those exchanges running, and reward each party for bringing in more participants. If you are building an agent network, I would love to hear from you.