When Your Knowledge Can Be Cloned: The Coming Agent Economy
What happens when a CPA's AI agent can serve unlimited clients?
A colleague at LinkedIn recently said something that stuck with me: “The agent economy might be winner-take-all — not at the platform level, but at the individual level.”
Here’s what he meant. Take a CPA. Today, she serves maybe 50 clients. She’s good, but she’s human — there are only so many hours. That constraint is what keeps the market competitive. There’s room for thousands of CPAs because no single one can serve everyone.
Now give her an AI agent. One that’s absorbed her expertise, her judgment, her way of working. Suddenly, she’s not limited to 50 clients. She’s not limited at all. The first CPA to build a great agent captures the market. The second one? There may not be room.
This isn’t hypothetical. It’s the logical conclusion of where things are heading. And it changes everything about how we think about careers.
The End of the Capacity Moat
For most of modern history, knowledge workers have been protected by a simple fact: there’s only one of you. Your expertise has value partly because you can only be in one place at a time. Scarcity creates opportunity.
Agents remove that scarcity. When your knowledge can be cloned and deployed infinitely, the economics flip. First movers don’t just get an advantage — they get the whole market. Late entrants find there’s nothing left.
This isn’t just a problem for CPAs. It’s lawyers, consultants, designers, developers, financial advisors — anyone whose value comes from knowledge and judgment rather than physical presence.
Entrepreneurship Becomes a Survival Skill
If traditional career paths get compressed by winner-take-all dynamics, what’s left? I keep coming back to the same answer: entrepreneurship. Not “quit your job and raise a Series A” entrepreneurship. Something more fundamental — the ability to identify opportunities, test ideas, and create value independently.
The problem is, most people aren’t wired for that. And our current systems don’t prepare them for it either. We train people to be excellent employees — to build expertise, join organizations, climb ladders. That playbook is about to get a lot less reliable.
The Gap Nobody’s Filling
Here’s what I find interesting: the tools we have today assume you’re already on one side or the other. You’re either a consumer (”Hey Siri, set a timer”) or a founder (”Here’s my AWS architecture for my SaaS startup”). There’s almost nothing for the person in the middle — someone with a skill, maybe an idea, working a 9-to-5, who wants to dip their toes in but doesn’t have the time or infrastructure to try.
What if your AI agent could be your first employee? Not in a cute metaphorical way, but literally — testing a service offering, finding potential clients, handling the grunt work, while you keep your day job. Low risk, real signal. Entrepreneurship with training wheels.
I’ve been setting up my own AI agent recently (through a project called OpenClaw), and the experience changed how I think about this. Giving an agent a persistent runtime — letting it actually do things rather than just answer questions — dramatically expands what’s possible. But the gap between “I have a capable agent” and “my agent can participate in the economy on my behalf” is still enormous.
What Needs to Exist
For any of this to work, we need infrastructure that barely exists today:
Discovery — How does someone find your agent? How does your agent find services it needs? We need an agent marketplace, not just an app store.
Trust and auth — If I hire your agent to do my taxes, how do I know it’s competent? How does it securely access my financial data? The trust layer is completely unsolved.
Transaction rails — Agents need to transact. Not just call APIs, but negotiate, quote, invoice, get paid. Someone needs to build the Stripe of the agent economy.
Accessibility — Setting up a capable agent today requires real technical skill. That needs to be as easy as opening a Shopify store.
The Dilemma
Here’s the tension that keeps me up at night — and it’s personal, because I work at LinkedIn.
LinkedIn’s mission is to create economic opportunity for every member of the global workforce. That mission assumes opportunity can be distributed. But if the agent economy is winner-take-all at the individual level, enabling it could concentrate opportunity instead.
Do you build the agent marketplace knowing it might contradict your mission? Or do you not build it, knowing someone else will — and they won’t care about distributing opportunity at all?
I don’t have the answer. But I think it’s one of the most important questions in tech right now. And pretending it’s not coming isn’t an option.
Next in this series: I'll go deeper on the winner-take-all problem — why agents break the natural equilibrium that keeps markets competitive, and what (if anything) can be done about it.
I'm Yusi Zh, an Engineering Manager at LinkedIn. I write about AI, the future of work, and building things. The opinions expressed here are my own and do not reflect the views or positions of LinkedIn. Follow along at yusizhang.com.