I have been sitting with a question lately: why does the modern corporation look the way it does?
Not in a philosophical sense. In a practical one. Why layers of management? Why full-time employees instead of contractors for everything? Why did "the firm" become the dominant unit of economic organization in the first place?
I put ChatGPT Deep Research to the test, and it compiled a report tracing corporate evolution across 700 years, from Florentine merchant partnerships to the English East India Company to the modern enterprise. What struck me is that the answer turns out to be remarkably consistent: firms exist to solve coordination problems, and their structure at any moment reflects the costs of production and coordination at that moment. There is no master plan, no grand vision. It's the invisible hand of the market, pushing organizations into certain configurations that optimize for productivity and success. And like evolution, it's the ones that adapt that survived.
When information was slow and enforcement weak, you coordinated through kinship and reputation. When railroads and telegraphs made large-scale production feasible, you built management hierarchies to substitute for market contracting. When computers and the internet reduced coordination costs again, you outsourced, modularized, and spun up platforms.
Same underlying logic, different shape. Every time.
What the Numbers Actually Show
U.S. labor productivity has grown nearly 5x since 1950. This one struck me — I keep hearing economists say that productivity has largely been flat. Perhaps it's how one group measures, perhaps it's the time frame that matters. But this number squares more with my own experience.
IT investment as a share of GDP went from about 1.6% in the early 1970s to 4.7% today. R&D intensity has more than doubled in that span. To be honest — that surprised me too. I would have thought IT spending would be a larger percentage. No doubt it will be in the future.
This is not just "technology getting better." This is organizations learning slowly, messily how to restructure themselves around new coordination tools. The technology came first. The productivity showed up later, after firms finally redesigned their processes, decision rights, and incentive systems to match.
This is the part most leaders miss. It wasn't the computers that created value. It was the organizational change the computers eventually forced.
So Here We Are
AI, and especially agentic AI that can execute tasks, drops coordination costs again. Dramatically. For cognitive work this time, not just information relay.
Three plausible near-term scenarios:
The most likely path: Hierarchies mostly survive, but decision cycles compress. Middle management stops being the information layer and becomes the judgment layer: goal-setting, exception-handling, cross-functional integration. Hybrid teams of humans and AI agents become the normal unit of work.
A real possibility in some sectors: Small core teams orchestrate fleets of specialized agents, shifting the firm-size distribution in software, marketing, and professional services toward leaner "modular orchestrators" that own the strategy but rent everything else. I am seeing this in action, and it's exciting and empowering, especially for those that are resource-constrained or have otherwise been locked out of the economic engine.
The regulated reality in high-stakes domains: Health, finance, infrastructure. AI autonomy expands slowly, constrained by liability and auditing requirements. Governance gets more formal, not less.
What This Means If You're Leading
Don't roll out AI tools. Build an AI operating model — with a clear owner, a tiered governance structure, and explicit decision rights for what AI can draft, recommend, and act on autonomously. "Shadow autonomy" (where systems make consequential decisions nobody's accountable for) is a real organizational risk, not just a policy one.
Instrument before you automate. Agents amplify whatever process reality exists underneath them. If your workflows are messy and your baselines unmeasured, automation makes that worse faster. Get clean data on cycle times and quality before you scale anything.
And maybe most importantly: build a workforce transition plan before you need one. AI changes task composition faster than the external labor market adjusts. The organizations that handle this well will be the ones that created internal pathways — reskilling, mobility, new role definitions — rather than discovering the problem on the back end. If you look back in history, the bedrock companies, the ones spotlighted in books like Jim Collins' Good to Great and Built to Last, were thoughtful and strategic about bringing their people along.
The Pattern Holds
Every major coordination technology in history reshaped corporate structure, not by making the old structure obsolete overnight, but by making a different structure newly possible. The leaders who moved early, who reorganized with the technology instead of waiting for the settled picture, captured the gains.
We're in that window now.
The firms that treat AI as a tool to bolt onto existing org charts will get incremental improvements. The firms that ask "what coordination problems does this actually solve, and what does our structure look like when those problems are cheaper?" — those are the ones building something that lasts.