Uber gives you a simple way to express what you want: a ride from one place to another. The network finds a driver willing to provide it at a price you’ll accept. More drivers improve availability for riders, and more riders give drivers a reason to join.
I believe agents will make it possible to build networks like this around a much wider range of goods and services. They can help people express more complicated preferences and find suppliers willing to meet them. Some of these exchanges already happen, but at a price or a level of complexity that keeps most people out. Others have been too much work to arrange at all. Agents will also make possible exchanges that we haven’t yet thought to ask for. All of these are opportunities for founders.
Consider a simple, familiar example: Organizing dinner with a group of friends. Everyone has a different schedule, budget, and opinion about where to eat. Today, one friend ends up running the group chat: collecting those preferences, searching for restaurants, and going back to the group when the available options don’t work. Restaurants, meanwhile, have tables to fill on a Tuesday night and limited ways to reach a group whose plans are still flexible.
Your personal agent coordinates with your friends, or with their agents, and agrees on a request: dinner for six in the Mission, under $60 a head, anytime from 5:30 to 8:00, anything but sushi. Restaurants’ agents respond with offers 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 the offers against everyone’s preferences and books within the limits you’ve agreed to.
A restaurant with an empty table at six can make an offer directly to groups willing to eat early, before they’ve chosen where to eat. The group has preferences and a budget. The restaurant has capacity and a price it would accept. What’s missing is the network that lets their agents find each other and agree on a purchase.
Uber shows how a network can bring buyers and suppliers together around a relatively simple request. Agents could extend that model to purchases with more complicated requirements, as with dinner for a group. Most networks have been built for humans. We need them for machines. Founders can build places where agents bring those requests and suppliers compete to fulfill them.
Building the network
Crypto already offers a model for organizing these requests with intents: A user specifies an outcome, and providers compete to deliver it. On UniswapX, for example, a user signs an order specifying the tokens they want to trade and the terms they’ll accept. Providers compete to fill that order, finding liquidity across different sources. Agents could help extend the intents model to everyday commerce. An LLM can turn a conversation into a request that suppliers can act on, and agents can search for offers and negotiate the details. That could make it cheap enough to arrange even a modest purchase, like dinner for six, this way.
The same network effects that make Uber useful could develop around these more complicated purchases. Each restaurant that joins adds another possible match for diners, especially when a group has specific requirements. More groups give restaurants more opportunities to fill tables, making it worth connecting their booking systems and responding with tailored offers. As participation grows in a city, the network becomes useful for more kinds of requests.
A founder could start with dining in one city and build from there. The initial product would need to save people enough time or money to bring their friends. Restaurants would need to see bookings they wouldn’t have received otherwise. Those results keep both sides coming back and inviting others.
I think a consumer-facing agent could be free to use and earn money through a model similar to payment for order flow. A user specifies what they want and what they’re willing to pay, and suppliers compete to fill the request. If the agent offers to arrange dinner for $300 and finds a restaurant willing to provide it for $270, the network could keep the difference. It could also receive a fee for routing the purchase to the restaurant. Revenue from these transactions could pay for the consumer’s inference. Users should understand how their agent gets paid and how it chooses among offers. Their willingness to delegate purchases will depend on trusting those choices.
Why these businesses can last
As AI makes software easier to reproduce, 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, became more useful as people joined.
An agent commerce network could develop the same kind of advantage. A competitor can build similar software, but it still has to bring together buyers and suppliers who can reliably transact. In the restaurant example, that means enough local restaurants willing to make offers and enough groups ready to book.
Making these transactions reliable is part of building the network. A restaurant needs to know that an agent is authorized to commit the group’s money. The diners need to know that accepting an offer secures the table at the agreed price. The network will need a way to verify the agent’s authority, record the booking terms, and arrange payment. Both sides should know what happens if the group cancels or the restaurant fails to honor the booking.
Founders can build these services alongside the market itself. As buyers and suppliers transact, the network can develop a record of who delivers what they promise and a process for resolving disputes. A network that helps participants find good offers and reliably complete purchases is the one they return to for the next transaction.
What else could founders build?
Dining is my toy example. There are many other networks agents could 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 on dates or destination creates room for offers that a fixed search would miss.
- Business procurement. A business’s agent stays online around the clock, negotiating future supply and price commitments with suppliers. It compares delivery schedules and payment terms alongside price, helping buyers lock in costs and giving sellers more predictable demand.
- Compute markets. Agents match workloads with available capacity based on hardware requirements, location, and timing. Businesses can also buy future capacity and hedge the cost of compute as their needs grow.
- Arbitration. Parties agree to use a shared forum to resolve disputes faster than the existing legal system. Each buyer and supplier that adopts the same rules makes the forum more attractive to the next.
- Physical-device networks. Agents buy services from drone networks or purchase data from connected devices. Each additional provider extends the network’s coverage or adds a service that other participants need.
- Collaborative health research. People pool health data and participate in crowdsourced studies. More contributors support larger studies and research into questions that smaller groups couldn’t answer.
There are human-centered network opportunities too. As intelligence gets cheaper, people will still need ways to establish identity, vouch for someone’s qualifications, and build trusted relationships. Credential networks, members’ clubs, and other organizations could become more valuable as more work happens outside traditional institutions.
***
The opportunity is to find exchanges that people want to make but that cost too much time or money to coordinate today. Some of the largest will almost certainly be exchanges we haven’t yet imagined. Start with a specific market, make those exchanges work, and reward each side for bringing in more participants. If you’re building an agent network, I’d love to hear from you.
***
The views expressed here are those of the individual AH Capital Management, L.L.C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the current or enduring accuracy of the information or its appropriateness for a given situation. In addition, this content may include third-party advertisements; a16z has not reviewed such advertisements and does not endorse any advertising content contained therein.
This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments for which the issuer has not provided permission for a16z to disclose publicly as well as unannounced investments in publicly traded digital assets) is available at https://a16z.com/investments/.
The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information.