AI is fundamentally changing the way we interact with technology. It has started with language and voice as an interface, and it will continue with the rise of agents. I think of agents as pieces of software that can do your bidding on the internet. They can navigate web, mobile, and desktop applications to accomplish tasks on your behalf. The recent Rabbit launch demonstrated to us what this can look like.
Agents are an inevitability. If mobile applications are remote controls for real life, agents are the universal remote. This begs the question of whether web2 companies that have built business models on top of pre-AI interaction models will adapt or even participate in the agent world. If services like Uber, Kayak, and Instacart generate a meaningful amount of revenue through advertisements and cross-selling products, will they want their application interfaces to be abstracted away by an agent? Are they okay with their customers never interacting directly with their applications and losing their eyeballs? Their incentive structure is at odds with agent proliferation. This is a potential innovator’s dilemma in the making.
New conversational interfaces made possible by AI will create a new category of “agent native” applications. Since agents will interact with them instead of end-users, these applications will be headless by nature. They may be entirely unknown to users and simply run in the background, happily playing their part in a series of chain reactions. These will be protocols that our agents call on in the world of bits and they will also bridge to atoms (eg ride-sharing, grocery delivery, etc.).
This is likely a scenario where AI and crypto will be two sides of the same coin. AI will create the need for agent-native headless applications, and web3 will step in to fill the void. Web3 applications and protocols can accommodate the new ways in which we will use technology (i.e. conversational and voice-centric interfaces with agents working for us) because they do not need to “own” end-to-end customer experiences unlike most web2 companies. They are happy to run in the background as composable job-doers. They don't need their brand names front and center because their tokenized business models, which are primarily straightforward and network usage-based, are positioned to thrive in this world, especially relative to their web2 counterparts.
I suspect we will see more and more of these “two sides of the same coin” scenarios emerge as AI and crypto continue to weave their way into our daily lives.