Every Project Still Needs Buyers

Last week I wrote about risk between ice and gas, and about how AI shrinks the risks we choose to take while enlarging the ones nobody can measure. Since then I have been thinking about a narrower question: which kinds of risk support entrepreneurship, and how they are distributed among the components that make up a project.

Three Components and the Models Built on Them

Anything we start needs three things. There must be an idea. There must be a team able to execute on it. And there must be resources, capital, to support the team while it does. Each of the three may be fully present, partly present or missing, and the missing piece is an opportunity for a particular model to step in and supply it.

When capital is missing, founders look to investors. Hackathons exist so that ideas and teams can find each other and gel in a weekend. Venture builders supply a team to founders who do not want to carry the risk alone, or who only want to carry it from a certain stage of development onward. Each model is defined by the component it brings and by the risk it agrees to absorb in exchange, and the equity split of a project records how that trade was made: whoever supplied the scarcest component, and took the largest share of the risk, owns more of the result.

What AI Changes in the Triad

Ideas were never the scarce component, and AI has made them more abundant still. What it changes is the cost of execution. A small team, or a single person, working with AI agents can now prototype, test, build and support a product that would have required a department a few years ago. That lowers team risk, because fewer people with fewer specializations can carry a project further. It lowers capital risk too, because the same work costs a fraction of what it did and the money required before the first customer shrinks with it.

When two of the three components become cheap, the weight of the whole project moves onto what is left. In a world of abundant ideas that are easier to execute, the important ingredient is the ability to test an idea in a market: to find people who are willing to pay for the product or the service, and who confirm by paying that they see value in it. The question of whether anyone wants to buy has not become easier at all, and none of the models above was designed to answer it. Investors, hackathons and venture builders each absorb a different risk, and market risk stays with the founder in all three.

Buying Attention Is a Zero-Sum Game

For a long time there was a standard way to address it. You could buy attention, if you chose to spend your money that way, by bidding for placement on Google, in the app stores or on Meta’s platforms. A large share of the pitch decks I receive still dedicate a substantial portion of the use of funds to go-to-market, advertising and marketing.

The trouble is that attention is a scarce space, and advertising for it is a zero-sum game. When many projects, each built more cheaply thanks to AI, crowd into the same auctions, they raise the price for everyone. The savings AI delivers on the side of building a solution are then transferred, almost entirely, to the platforms that sell the attention. A founder who relies on advertising to reach the market ends up passing the benefit of AI on to Meta, Google and the app stores.

Three Ways Around the Paradox

The first alternative is to find an anchor buyer, someone who commits in advance to purchase a product that may not yet exist, and in doing so creates the market for it. In 2007 five governments and the Gates Foundation pledged $1.5 billion in an advance market commitment for pneumococcal vaccines, guaranteeing manufacturers a market at a price the poorest countries could afford. In 2022 Frontier, a group of technology companies, committed $925 million to buying permanent carbon removal by 2030, a commitment it has since doubled, and that promise catalysed an industry that had no customers before it. The UAE is doing the same for agentic AI. Having committed to delivering half of its government sectors, services and operations through AI agents within two years, the state becomes the anchor customer of agentic solutions, and it guarantees symmetrically that a private enterprise market will form around the agents that need to speak and transact with the government’s own.

The second is to force markets open. The commissions Apple charged on purchases made outside its apps have been challenged by regulators and courts in Europe and the United States, on the grounds that they unfairly restrict competition. Each ruling of that kind returns part of the toll to the people who build the products.

The third is community. A message that circulates inside an ecosystem where people trust one another reaches buyers without passing through an auction, and the trust is what makes them listen. Crowdfunding turned that trust into a mechanism. On Kickstarter a community commits its money to a product before it exists, acting as a collective anchor buyer, and a campaign that falls short of its goal has tested the market at the cost of a page and a video. Patreon and similar platforms do the same for a continuing body of work, letting the people who already trust a creator pay for what comes next, without the creator having to buy their attention again each time. Our small community at Ikigai Collective is developing exactly that kind of trust, week after week, and I think its role in helping projects find their first customers will grow.

When the Buyer Is an Agent

How will all of this change when AI agents make it as easy to transact as they have made it to build? At least today, AI agents are not swayed by advertising, and they do not take part in building the relationships of trust that make human-to-human transactions possible. Neither of the two forces that currently move a purchase acts on them.

The tools that will send agents out into the world are arriving quickly. This week AgentMail, the service I use to let my own agents send and receive email through traditional channels, announced AgentID, a way for an agent to sign in to a platform as itself, much as we sign in with Google or with Apple. More and more platforms will accept it. As go-to-market strategies, purchase decisions and the choice of which tools to adopt come to be influenced, supported or led by agents, the market component of every project will change shape again, and I do not yet know what it will look like. I am eager to find out, and I expect we will discover it together, by building the projects that need a buyer and watching who, or what, turns up to buy.