The Second Workforce Is Already Here. Most Companies Don’t Know How to Run It.
There’s a shift happening right now that almost nobody in business is naming clearly.
There’s a shift happening right now that almost nobody is naming clearly.
Not because the information isn’t out there. Because the people who should be paying attention are too busy managing the old world to notice the new one forming underneath it.
Here it is:
A second workforce has arrived. And the collapse of average is accelerating alongside it.
These two things are connected in ways that matter enormously, for founders, for investors, for anyone trying to build something that compounds rather than just survives.
The Second Workforce
For the last two years, the AI conversation in business has been about tools. Faster writing. Smarter search. Automated admin.
That conversation is already out of date.
What’s actually happening now is that organisations are beginning to run two workforces simultaneously. One human. One not.
Software agents that don’t just assist tasks, they execute workflows end to end. They handle decisions, manage processes, talk to customers, write code, and run operations. They work while you sleep. They don’t burn out. They don’t have bad Mondays.
And they’re already inside most serious businesses, whether those businesses have designed for them or not.
The problem is almost nobody has designed for them.
Every org chart, leadership framework, HR system, and management philosophy that exists was built around human workers. None of it maps cleanly onto a non-human workforce. And so you get what I’m seeing across the businesses I invest in and advise: AI capability deployed on top of human-shaped infrastructure, creating complexity where there should be leverage.
Tools bought. Architecture missing.
I made a version of this mistake at ContentCal. We had the tools before we had the systems to make them compound. Speed without structure just creates expensive noise. The lesson cost us months and money we didn’t have.
The same principle is playing out now at a much larger scale.
The companies getting this right aren’t necessarily the ones with the most AI. They’re the ones who’ve asked the harder question: how do we structure human and agent work so they actually reinforce each other? Where does the agent run autonomously? Where does human judgment hold the line? Who’s accountable when the agent gets it wrong?
That’s not a technology question. It’s an operating model question. And most organisations haven’t started answering it.
The Collapse of Average
At the same time, the middle is disappearing.
Not just in the job market. In business outcomes, in competitive position, in what it means to be a capable professional or a functioning company.
For a long time, being reasonably good at something was enough. You could build a solid business in a category and hold ground. You could be a competent operator and have a long, stable career. The average was a viable place to be.
That’s ending.
When leverage concentrates, when one person with the right tools and the right clarity can do what used to take a team, the average stops being a market position. It becomes a waiting room.
I’ve watched this play out across the 60+ companies I’ve invested in. The gap isn’t between the great and the good anymore. It’s between the people and companies that have designed for leverage, and everyone else running harder on a treadmill that keeps speeding up.
The ones falling behind aren’t lazy. They’re not stupid. They’re optimising for a world that’s quietly being replaced.
What I find striking is how invisible this feels until it suddenly isn’t. The company that seemed fine last year is somehow not fine this year. The operator who was reliable is somehow behind. Nothing dramatic happened. The ground just moved.
That’s how structural shifts work. They don’t announce themselves. They accumulate.
Why These Two Things Are Connected
The second workforce and the collapse of average aren’t separate trends. They’re the same trend.
Agents amplify whoever is holding them well. That means the gap between companies and individuals who’ve designed for leverage and those who haven’t doesn’t grow linearly. It compounds.
One team with clear architecture, knowing which decisions agents handle, which humans hold, how quality is governed, can operate at the scale of a much larger organisation. One operator who thinks carefully about how to deploy intelligence alongside their own judgment can outperform a team that’s just adding tools and hoping.
The compounding works in reverse too. Add agents to a chaotic system and you don’t get a more productive chaotic system. You get faster chaos. More expensive chaos.
This is why the firms and operators I see pulling ahead quietly are not the ones spending the most on AI. They’re the ones who got deliberately small before they got leveraged. Narrow focus. Clean systems. High judgment. Then agents on top of that.
It looks effortless from the outside. It isn’t. It’s designed.
What To Actually Do
If you’re building something or backing something, a few things I’d be thinking hard about:
The question isn’t “which agents should we deploy?” It’s “what’s our architecture for human and agent work?” Without the architecture, the agents create overhead, not leverage. Design the operating model first. The tools follow.
Hire for judgment above all else right now. As agents handle more of the execution layer, what appreciates in value is the quality of thinking sitting above them. The ability to frame a problem correctly, make the call agents can’t make, catch the error before it propagates. That’s what compounds. Raw output is getting cheaper every quarter. Judgment isn’t.
Governance is the moat nobody is building. Clear permissions. Defined accountability. Escalation paths when agents fail or hallucinate. This sounds bureaucratic. It’s actually competitive. The organisations building governance infrastructure for their agent workforce now are creating something their competitors will spend years trying to replicate.
Get specific faster than the market does. The collapse of average rewards precision. One customer segment, owned completely. One metric that actually matters. One channel that works before you add the second. In a world where leverage concentrates, being broad and average is the most dangerous place to operate. The barbell has no middle.
Build distribution like it’s a second business. Because in a leverage economy, it effectively is. Content that builds belief, narrative that opens doors before you need them, brand that creates inbound rather than requiring outbound, these compound in a way headcount never could. The founders I see compounding fastest are treating this seriously, not as a side project.
The Real Divide
The question isn’t whether AI changes work. That’s settled.
The real question is which side of the divide you end up on.
On one side: companies and individuals who’ve designed for the hybrid human-agent world, who’ve built clean systems, hired for judgment, governed well, and gotten specific.
On the other: everyone who layered AI onto existing assumptions and called it transformation.
The frustrating thing about structural shifts is that the window for getting ahead of them is always shorter than it looks. And it always looks longer than it is.
The second workforce is already here. The collapse of average is already underway.
The only question worth asking right now is whether you’re designing for it, or waiting for the moment it becomes undeniable.
By then, the gap will already be very hard to close.