Guest Post: Stop Selling AI Agents
The glaring issue with agent economics
This week features a post from AC Summer Scout Sai Mattapalli, a Vanderbilt University Computer Science undergrad and serial entrepreneur.
Selling “agents” is like selling a hammer in a world where everyone can make hammers for $0.01. The value capture is in the house you build, not the hammer you use to build it.
Most agents are built on the same foundation models, making them cheap to produce, easy to replicate, and nearly impossible to differentiate in a way that drives meaningful customer LTV. As inference costs fall, so will the price customers are willing to pay, until it’s effectively nothing.
The winning move is to make the agent invisible, the unseen engine inside a larger, more valuable product. The best companies won’t sell agents, they will sell outcomes. Envision a claims process cut from weeks to hours, or a logistics network that always delivers on time. The agents are just the secret ingredients.
Mike, the GP here at Also, once told me that the best businesses have this almost magic to them. They deliver some product or process that just…works, in a way that feels impossibly too fast, too accurate, and too good to be true. Agents are a modern unlock for delivering that kind of magic, not a product to be sold.
I’m Sai, a 19 year old who was graciously given the opportunity to work alongside Mike the last two summers. Prior to my time at Also Capital I worked on and published AI research before founding my first startup called Vytal. I’m now working on something new called River where I can assure you, I will not be selling AI agents.