Agentic swarms are an org-chart delusion
The "agentic swarm" vision of productivity is comfortingly familiar.
Which should be an immediate red flag...
You take the existing corporate hierarchy, you replace the bottom layers with a swarm of AI agents, and you keep humans around as supervisors. It's an org chart with robots instead of interns. The VP of Engineering becomes the VP of Engineering Agents.
Congratulations. You've reinvented middle management.
This is what Clayton Christensen would have called a sustaining innovation in the guise of disruption: you're using new technology to do the same thing slightly more efficiently, in a way that looks and feels like the old thing, and the incumbents love it because the power structure stays intact.
The person at the top still delegates. They still think in terms of roles and departments and functional areas. They've just swapped out the people underneath them for software that doesn't need health insurance.
But when an actually disruptive technology arrives, it makes the existing structure irrelevant.
And AI is that tech.
Roles are an artifact, not a law
The entire "swarms of agents" model is based on the idea that work naturally decomposes into roles. You ~need a marketing agent, a sales agent, a support agent, a development agent - because marketing, sales, support, and development are existing job titles and, for humans at least, fundamentally different activities that belong in different boxes.
This feels like an obvious truism, but it's a depreciable artifact of organizational scaling, not some deep // universal truth about work itself.
Adam Smith's pin factory example is famous because he showed that dividing labor into specialized roles made pin production dramatically more efficient. But the pin factory was a specific solution to a specific constraint: individual humans are slow, they get tired, and they can only hold so much context in their heads at once.
Specialization was a workaround for bio-cognition. If you could have one person who simultaneously understood metallurgy, wire-drawing, straightening, cutting, pointing, grinding, and packaging - and could do all of it concurrently without fatigue - Smith would never have divided the labor in the first place.
That hypothetical person is what a solo practitioner with a capable AI already looks like.
Outcomes over org charts
When I sit down with an AI assistant and say "write me a marketing brief, then generate the landing page copy, then draft the ad variants, then build the page, then set up the analytics tracking," I'm not managing a team of five agents with five different specializations. I'm issuing a sequence of commands to the same system from the same interface.
The boundaries between "marketer" and "developer" and "analyst" dissolve, because those boundaries were never real boundaries in the work itself. They were boundaries in human capacity.
The people who will thrive aren't "agent managers." They're people who can say what they want and evaluate whether they got it - and whether what they got was either good or shit.
The workflow looks less like a CEO directing department heads and more like a musician working in a DAW: one person playing every instrument, mixing, mastering, and producing, toggling between tasks at the speed of thought rather than delegating through layers of abstraction.
Brian Eno talked about the recording studio as a compositional tool — something that collapsed composer, performer, and engineer into one creative role. AI is doing the same thing to knowledge work, collapsing strategist, executor, and analyst into one operational role.
This is ~already happening
Plenty of one-person businesses are shipping products, running marketing, handling support, and managing finances through a single AI-augmented workflow. They're not thinking about it in terms of agents with job titles. They're thinking about it as "what do I need to get done today" and then doing all of it, fluidly, without ever switching between conceptual departments.
The "swarm of agents" idea appeals to people who either come from or aspire to the world of management. If you've spent your career hiring people and organizing them into teams - or dreaming // LARPing about becoming a CEO - then naturally you look at AI and see a new kind of team to organize. It's Maslow's hammer. When all you have is an org chart, everything looks like a headcount decision.
Why this matters
If you believe the future is agent management, you'll build tools for orchestrating fleets of specialized bots. You'll create dashboards for monitoring your marketing agent separately from your sales agent separately from your dev agent. You'll recreate Salesforce, but for robots.
If you believe the future is unified execution, you'll build tools that let one person express intent and get outcomes across every domain from a single surface. The interface collapses. The abstraction layers disappear. You don't manage agents any more than you manage the individual transistors in your laptop.
The first path leads to a world that looks a lot like the one we already have, just with fewer humans in the lower tiers.
The second path leads to something actually new: a world where the unit of economic production isn't the company or the team but the individual, with a general-purpose cognitive tool that makes specialization itself an anachronism.
I know which version the people who currently sit atop org charts would prefer. And I know which version the technology is actually pushing toward. Those two things, for the moment, are very different.
But the technology tends to win these arguments eventually. It won in music production. It won in publishing. It won in video. Every time a tool collapses specialized roles into generalist capability, the generalists inherit the earth — no matter how loudly the specialists insist their particular expertise can't be automated or absorbed.
The future of work isn't managing a swarm.
It's being the swarm.