Era of Human Agency

We are entering a new era of work. Most people are framing it around what AI can do — the agents, the copilots, the automation layers. I am framing it around what humans still must do.

This is the Era of Human Agency.

AI is about to absorb a massive share of coordination, execution, and operational labor. The work that kept organizations running — routing decisions, managing handoffs, maintaining workflows, synthesizing reports — is becoming software. Fast. What remains scarce is not output. It is judgment, context, taste, empathy, and the ability to decide what actually matters.

That is a structural shift, not a feature upgrade. And it changes who matters inside an organization, how roles are defined, and what work is actually worth paying a human to do.

The New Division of Labor

The logic is simple: humanness-rich humans guide, decide, and empathize. Context-rich AI agents coordinate, execute, and scale.

As that happens, roles collapse back to their highest-agency form. Not into new titles or redrawn org charts — into four fundamental modes of work that have always existed but were buried under layers of operational overhead.

Builders create. Products, systems, content, infrastructure — anything that did not exist before and now does. AI accelerates what builders produce. It does not replace the decision of what to build or the taste required to build it well.

Observers see what others miss. They read markets, interpret data, identify patterns, and name the problems worth solving. This is the analyst, the strategist, the researcher — not as a title, but as a function. AI can surface information faster than any human. It cannot yet determine which information changes a decision.

Sellers persuade. They create demand, build relationships, close deals, and move people to act. AI can generate outreach, qualify leads, and personalize at scale. It cannot replace the trust, timing, and conviction that make someone say yes.

Activators make things happen across people and systems. They are the operators, the project leads, the integrators — the ones who turn a decision into coordinated action. This is the role most transformed by AI, because the majority of an activator's work was coordination. In this era, activators shift from managing workflows to owning outcomes — directing AI agents rather than managing human task chains.

Every role that survives will map to one or more of these modes. Roles that were mostly coordination reorganize into something leaner or disappear entirely. Roles that were mostly judgment gain leverage they never had before.

What This Means in Practice

A marketing director does not get "replaced by AI." But the 60% of her week spent on campaign scheduling, reporting, vendor coordination, and status updates does. What remains is the 40% that was always the most valuable part of the role — positioning decisions, creative direction, and market judgment. AI does not remove her job. It compresses it to its highest-value form.

Consider the difference between a RevOps manager routing leads across teams, reconciling CRM data, and building pipeline reports — and a growth lead deciding which market to enter, what the messaging should be, and when to kill a bet that is not working. The first set of tasks is a coordination problem. The second is a judgment problem. AI absorbs the first. The second becomes more valuable.

Any role fundamentally defined by "ops" is a bundle of coordination tasks. Coordination is software now. The high-agency work remains: build, observe, sell, activate.

This is not a prediction about the distant future. The SBA Office of Advocacy found that small business AI adoption nearly closed the gap with large enterprises in just 18 months — faster than broadband internet equalized. MIT's NANDA research found that mid-market companies are moving from AI pilot to production in 90 days. The shift is not coming. It is here, and the pace is accelerating.

The Organization

Today's organizations were designed around a constraint that no longer exists: human input required to generate business output.

For decades, coordination was expensive. Scale meant more people and more layers. More people meant more managers, more meetings, more process — not because leaders wanted bureaucracy, but because there was no other way to move information and decisions through a complex operation. Headcount was the price of capability.

AI changes the cost structure of coordination itself. The same economic output can now be achieved by a fraction of the workforce — not because people are unnecessary, but because the work that justified most of the headcount was never judgment in the first place. It was coordination.

Start With Tasks, Not Roles

The way to see this clearly is to stop thinking about roles and start thinking about tasks.

Every role in an organization is a bundle of tasks. Some of those tasks require judgment, relationship, or creative decision-making. Others are coordination — moving information between people, systems, and processes so the judgment tasks can happen. AI does not eliminate roles wholesale. It absorbs tasks selectively, stripping the coordination layer and leaving the judgment core exposed.

When you evaluate any role through the lens of its component tasks, a clear pattern emerges. Coordination tasks migrate to AI. Judgment tasks stay with humans. The ratio between them determines how much a role transforms. Roles that are 80% coordination and 20% judgment do not survive in their current form. Roles that are 80% judgment and 20% coordination become dramatically more productive.

This is the diagnostic tool. Do not start a restructuring conversation by asking which roles to cut. Start by mapping the task composition of every function. The answer will be obvious — and more defensible — than any top-down headcount target.

The Coordination Tax

Every additional person in an organization increases coordination cost faster than decision quality. This is not a new observation — it is the fundamental logic behind Brooks's Law, Dunbar's Number, and every founder's instinct that the team was faster at 15 than at 150.

The scale of the problem is now measurable. Microsoft's 2025 Work Trend Index found that the average knowledge worker spends 57% of their time communicating through meetings, email, and chat. Forty-three percent is left for actual creative work. Employees are interrupted every two minutes during core work hours — 275 times per day. Harvard Business Review research puts it even more starkly: managers and professionals spend 35–50% of their time coordinating, with some studies estimating 60–80% for knowledge workers in complex organizations.

That is the coordination tax, made literal. And until now, there was no alternative. Organizations added people because coordination required people. The meetings, the status updates, the alignment sessions, the layers of approval — all of it was the price of scale. AI makes that price optional.

This breaks the operating model that most large organizations are built on. Headcount was the input metric. Revenue per employee was an output metric, but rarely the one that drove decisions. Promotions, budgets, and organizational importance were all calibrated to team size. A VP with 200 people had more influence than a VP with 20, regardless of relative output. That logic is now a liability.

What the New Organization Looks Like

The post-AI organization is not a smaller version of the current one. It is a different shape.

Fewer layers between decision and execution. Smaller teams with broader scope. Senior people doing work that used to be distributed across three levels of reports. AI handling the synthesis, scheduling, reporting, and workflow management that previously required dedicated roles.

The winning organizations will not be the biggest. They will be the ones with the highest ratio of judgment to coordination in every role, on every team, at every level.

If headcount is no longer the signal of organizational capability, what replaces it? Outcomes per human. Not revenue per employee — that is a trailing indicator and too blunt. Outcomes per human captures the quality and velocity of decisions, the speed of execution, and the ability to create value without adding complexity. It is the metric that rewards the VP with a team of 12 who ships faster than the VP with a team of 120.

The Transition

The challenge of getting there is not technological. The tools exist. The challenge is that enterprise incentives, leadership models, and cultural identity were all designed around headcount as the primary input to capability. Rewiring those systems while the business continues to run is the hardest work most leaders will do in the next five years. There is no clean path. But there is a clear sequence.

Change how you evaluate leaders. Stop using headcount as a signal of progress. The default leadership competency model in most enterprises rewards scale: bigger teams, bigger budgets, bigger scope measured in people. A VP with 200 reports is treated as more senior than a VP with 20, regardless of relative output or decision quality. That model was rational when coordination required humans. It is now a liability. Redefine leadership competency around impact — outcomes delivered, decisions made, speed of execution. Reward leaders who redeploy headcount while improving results. The goal is not to punish leaders who built large organizations. It is to stop incentivizing them to keep building when the constraint that justified those organizations no longer exists.

Collapse tasks before you collapse teams. Before restructuring any team, map the task composition of every role: what percentage is coordination, what percentage is judgment? The sequence matters — identify the recurring coordination tasks that slow execution, automate those workflows with AI and shared memory, then redefine the remaining roles around judgment and ownership. Collapse tasks first. Team changes follow from the task map — they are not the starting point. This is harder than a top-down headcount reduction, but it is more defensible, more precise, and more durable. A headcount target tells you how many people to remove. A task map tells you which work to remove — and that distinction is the difference between a reorg and a rewiring.

Make displacement explicit, not implicit. Most enterprises handle displacement through euphemism. Roles are "evolved." Teams are "right-sized." This is corrosive. It destroys trust, slows the transition, and makes the next round harder. The honest version of displacement does not have to mean elimination. But it requires naming what is changing — clearly, early, and without pretending that a coordination role is going to become a judgment role through sheer willpower. Acknowledge that fewer people are needed for the same outputs. Separate performance conversations from structural change — they are different processes with different stakes, and conflating them is unfair to the people involved.

Build explicit pathways for reskilling and outskilling. If you are going to collapse coordination tasks and redeploy or exit the people who performed them, you owe those people a real pathway — not a generic learning portal and a wish of good luck. JFF's national survey found that 77% of workers expect AI to affect their career within five years, but only 31% report receiving any AI-related training from their employers. Set aside dedicated budget for reskilling — and split that budget explicitly between human transition costs and automation investment. Do not bury transition costs inside the automation business case. They are real costs, and treating them as an afterthought guarantees they will be underfunded. Support both internal and external mobility. Not every person whose coordination role dissolves will find a judgment role inside your organization. That is an honest statement, not a failure. Outskilling — helping people build capabilities that make them valuable somewhere else — is a legitimate and necessary investment.

Rewire at an uncomfortable pace, but do not panic. The temptation in every enterprise transition is to move either too slowly or too fast. Too slow, and the coordination tax compounds while competitors adapt. Too fast, and you break the trust and institutional knowledge that still matter. Set outcome goals, not activity goals. "Deploy AI in three functions by Q3" is an activity goal. "Reduce time-to-decision in pricing by 40%" is an outcome goal. The first creates pilots. The second creates change. Communicate frequently — not performatively, but about the actual state of the transition: what is changing, what is not, what the timeline looks like, and what is still uncertain. People can handle uncertainty. They cannot handle silence followed by surprise.

The five moves above are not a checklist. They are a sequence — and the sequence matters. You cannot reskill people if you have not been honest about displacement. You cannot be honest about displacement if you have not mapped which tasks are moving to AI. You cannot map tasks if your leadership model still rewards headcount over outcomes. Start with how you evaluate leaders. The rest follows.

The Small Business Advantage

Everything above describes a painful, necessary transition for large enterprises. For small and medium-sized businesses, the story is different. This is not about transition. It is about opportunity.

Small businesses have never lacked for good ideas. They lacked the capacity to execute them. A four-person services firm knows exactly which clients need proactive outreach. A local retailer knows which products should be bundled for the season. A specialty manufacturer knows which markets are underserved. The insight was always there. The operational bandwidth was not.

For decades, the only way to unlock capability was to add people. Every new hire in a small business tightens cash flow, increases coordination cost, raises the stakes of a bad decision, and forces the owner further from the work that generates the most value. The old model made growth synonymous with complexity — and complexity is what kills small businesses.

AI breaks that link. The question is no longer "Can we afford to do this?" It is "Is this worth solving?" That is a fundamentally different constraint — and it favors the operator with judgment, not the operator with headcount.

Speed as Structural Advantage

The most underappreciated advantage small businesses hold is not their local knowledge, their customer relationships, or their operational flexibility — though all of those matter. It is their decision velocity.

Large enterprises deploying AI face a coordination tax at every step. Pilots require steering committees. Workflow changes require compliance review. New tools require IT approval, security assessment, and change management. MIT's NANDA research found that 95% of enterprise AI pilots fail to deliver measurable return. Mid-market and smaller companies, by contrast, are moving from pilot to production in 90 days — one-sixth the time it takes large enterprises to do the same.

Small businesses have none of that friction. The same technology that stalls inside a Fortune 500 org chart can be deployed by a small business owner in a weekend.

This matters because the value of AI compounds with use. A business that deploys an imperfect workflow today and iterates on it over six months will be in a structurally different position than one that waits for the perfect solution. The 70% solution deployed now beats the 100% solution deployed never — and small businesses are uniquely positioned to operate that way. They do not need consensus. They need one owner who sees the opportunity and acts.

Local Knowledge Becomes a Weapon

Proximity is the second advantage — and it is structural, not circumstantial.

A small business owner knows their customers in a way no enterprise CRM captures. They see the edge cases. They understand the community dynamics, the seasonal patterns, the unspoken preferences that never show up in a dashboard. Historically, that knowledge was trapped — useful in conversation, invisible in operations. AI changes the economics of acting on what you already know.

A local retailer can now use AI to analyze purchasing patterns and adjust inventory without hiring a demand planner. A services business can automate follow-up sequences that reflect the actual relationship with each client — not a generic drip campaign. A restaurant can run dynamic pricing and targeted promotions without a marketing team.

The leverage comes from the combination: human judgment about what matters, AI-enabled execution at speed. This is not replacing the owner's instinct. It is giving that instinct operational reach.

The Opportunity Is Asymmetric — and Time-Bound

The headcount constraint is dissolving. Local knowledge is becoming leverage. Capability can scale before payroll does. And the competitive gap between small and large is narrowing in ways that have not happened since the early internet.

But this window does not stay open forever. The businesses that adopt AI early and build compounding workflows — not just one-off experiments — will create advantages that are difficult to reverse. The businesses that wait will not just miss the window. They will find themselves competing against operators who already closed it.

The question is no longer whether AI is relevant to you. The question is whether you are willing to use it to operate at the level your judgment has always deserved.