a16z: The next Uber could belong to AI Agents, and new internet platform opportunities are emerging.

a16z Crypto partner Guy Wuollet argued that AI agents could enable new two-sided marketplaces far beyond Uber-style ride-hailing, citing a dinner-organization example in which personal agents negotiate a six-person booking under $60 per person in San Francisco’s Mission district while restaurant agents bid with empty-table offers. The firm pointed to crypto’s “intents” model, such as UniswapX, where users sign orders specifying tokens and acceptable terms and market-maker solvers compete across liquidity sources to fill them, as a template for agent commerce. Wuollet proposed a cold-start strategy focused on a single city’s restaurant scene, with a free consumer agent monetized through a payment-for-order-flow model: if an agent quotes $300 for a dinner and matches a restaurant willing to accept $270, the network could keep the spread or charge a routing commission, subsidizing AI inference costs. He argued that as AI drives software replication costs toward zero, durable moats will come from networks that accumulate fulfillment reputation and two-way transaction flow, requiring built-in verification of agent permissions, booking terms and settlement, plus dispute handling for cancellations or no-shows. Beyond dining, he flagged travel and enterprise procurement as promising verticals where agents can coordinate group schedules and budgets or negotiate supply contracts around the clock.