Will the Agent Explosion Really Come?
Henry Kim ·
Will the agent explosion really come?
I have said it in past posts. I use agentic AI well, and I am building agentic AI for small businesses. My experience and trust are baked into my agent. That is why people use my agent instead of the leading companies' apps. That was the last post: scarce things hold value.
But a question remains. How much will all knowledge workers use agents in their work? How deep will it go? Like coders today, will they use agents for fundamental problem-solving? Will usage really explode?
My answer splits. Personal use will grow; everyone gets a chief of staff. Commerce will grow too. My predicted picture: me and my agent versus the seller and the seller's agent, a tug of war of outsmarting each other by turns. Negotiation becomes agent versus agent, not human versus human. Haggling, terms, review verification: all agents' work. Shopping becomes negotiation, not search. My agent brings budget and taste; the seller's agent brings inventory and margin.
But the explosion where every knowledge worker leans on agents deep in their work: that, I am not sure about. Coders use agents as partners in debugging and design. Will marketers, accountants, lawyers go that deep? Using a tool and leaning on a tool are different, and the explosion comes from the latter. Most use today is the former: drafts, summaries, translation. Easier life, unchanged work. The explosion comes when the work itself changes.
We are still in the early stage of the superintelligence era. History shows the early stage always follows the same pattern: overinvestment, bubble, consolidation. Railroads did. Fiber optics did. Will AI skip the process and keep productivity rising exponentially in abundance? When I ask myself, my answer is no. There are no shortcuts. The sequence will not be skipped this time either.
Right now the AI wheel is spun by overinvestment and market expectations feeding each other. Buy chips, build data centers, grow models, raise funds again. The aftershock will come; its size I do not know. It is adjustment, not failure. What remains after the foam clears is real. Remember: YouTube came after the fiber bubble burst.
I am waiting for the consumer market to open. In the wireless internet era came the iPhone, and after it YouTube exploded. Infrastructure was laid, devices spread, and on top of them the real consumer market opened. AI's iPhone and YouTube have not arrived yet. When they do, agents become everyday life, not infrastructure. Right now is the era of laying infrastructure. What will be needed then is not a better model but a trustworthy agent.
Humanity holds A, A, A. The robot deals the last card. People are betting on four of a kind. All the money in the market right now is that bet.