For most of the history of modern software, there was a fairly obvious tradeoff between being small and being powerful.
A freelancer could move fast.
A small company could stay focused.
A large enterprise could do things neither of them could: operate at enormous scale, coordinate thousands of people, build deep infrastructure, maintain global distribution, carry regulatory risk, and deploy capital that an individual could never access.
AI is changing that tradeoff.
Not because enterprises are disappearing.
And not because everyone is suddenly going to become a freelancer.
The more interesting change is that the minimum efficient size of an organization may be collapsing.
At the same time, the maximum potential power of a well-run enterprise may be increasing.
That creates a strange new world.
A single person can increasingly behave like a company.
And a company can increasingly behave like a small group of extraordinarily capable people.
The question is which side adapts faster.
The Historical Advantage of the Enterprise
Large organizations exist for good reasons.
Many things are simply difficult to accomplish alone.
Someone needs to build the product.
Someone needs to sell it.
Someone needs to run infrastructure.
Someone needs to answer customers.
Someone needs to negotiate contracts.
Someone needs to deal with security, compliance, finance, taxes, procurement, recruiting, marketing, legal issues, and hundreds of operational details that accumulate around any sufficiently successful business.
The enterprise solved this problem through specialization.
Instead of one person doing twenty things badly, twenty specialists did one thing well.
And as organizations grew, another layer appeared to coordinate all those specialists.
Managers.
Program managers.
Product managers.
Architects.
Operations teams.
Planning processes.
Reporting systems.
Meetings.
Documentation.
The enterprise became a machine for coordinating human specialization.
That machine created enormous leverage.
But it also created enormous overhead.
And for decades, there was no obvious alternative.
AI Attacks the Coordination Problem From Both Sides
AI changes this structure in two important ways.
First, it increases the range of things a single person can competently perform.
A software engineer can suddenly do more design.
A designer can produce prototypes.
A founder can analyze contracts.
A marketer can build software.
A product manager can query data.
A salesperson can create personalized materials.
A researcher can build working tools.
The boundaries between professions become less rigid.
You no longer need to become an expert in every adjacent discipline.
You need enough understanding to direct AI effectively, evaluate its output, and recognize when something is wrong.
That is a radically different skill requirement.
The second change is even more consequential.
AI can perform some of the coordination work itself.
It can maintain context.
Summarize decisions.
Track dependencies.
Generate documentation.
Review work.
Discover inconsistencies.
Prepare reports.
Translate intent between domains.
Eventually, much of the information-moving work inside an organization can happen automatically.
And in many organizations, information-moving work may be a surprisingly large share of what people do.
The Rise of the One-Person Enterprise
This is why the idea of the solo founder is becoming much more interesting.
In some cases, a single person with strong judgment and a collection of AI systems can already operate something that would previously have required a small team.
They can research a market.
Develop a product.
Build a website.
Create marketing material.
Run experiments.
Provide support.
Analyze usage.
Manage infrastructure.
Handle basic finance.
Prepare contracts for review.
Produce documentation.
And continuously improve the product.
The important point is not that AI does all of these things perfectly.
It does not.
The important point is that it changes the economics.
Previously, the question was:
Can one person do all of this?
Usually, the answer was no.
Now the question becomes:
Can one person supervise systems that do most of this?
For more and more kinds of work, the answer may be yes.
That is a fundamentally different business model.
Freelance Becomes Something Bigger
The word freelancer may eventually feel inadequate.
Historically, a freelancer sold personal labor.
You had a limited number of hours.
You charged for those hours.
Your income was tightly coupled to your own time.
AI breaks that relationship.
A highly capable independent professional may increasingly resemble a miniature agency.
Or a studio.
Or a consulting firm.
Or even a software company.
The distinction between freelancer and company begins to blur.
Imagine an independent engineer who has:
- AI research agents
- coding agents
- testing agents
- documentation agents
- a design agent
- a sales assistant
- a financial assistant
- and a persistent knowledge system coordinating all of them
There may still be only one human.
But economically, this person is no longer operating alone.
They are managing a small digital organization.
That changes pricing.
It changes competition.
And it changes what a single person can credibly promise to a customer.
The Death of the Hourly Model
This creates an uncomfortable problem for traditional freelancing.
If AI allows a task that once required forty hours to be completed in four, charging by the hour becomes increasingly irrational.
The customer should not care how long something took.
They care whether the outcome is valuable.
Freelancing therefore moves closer to outcome-based economics.
| Instead of | The model becomes | |
|---|---|---|
| The offer | “I charge $200 per hour.” | “I can solve this problem for $20,000.” |
| Priced on | How long something took | Whether the outcome is valuable |
If AI allows the work to be completed in three days instead of three weeks, that efficiency belongs partly to the person who built the capability to produce the result.
This is similar to what happened with software itself.
Nobody asks how many CPU cycles were required to generate a search result.
The value comes from the outcome.
AI pushes human knowledge work in the same direction.
But Enterprises Get AI Too
It would be a mistake to conclude that AI therefore favors only small organizations.
Large companies receive exactly the same leverage.
Potentially more.
An enterprise has something an independent professional usually does not have:
context.
Massive amounts of it.
Customer relationships.
Historical decisions.
Operational data.
Proprietary software.
Internal documentation.
Distribution.
Brand.
Capital.
Regulatory knowledge.
Existing workflows.
Millions of interactions describing how the company actually operates.
If AI systems can effectively understand and operate across that context, the enterprise becomes extraordinarily powerful.
A company with 50,000 employees may not need AI simply to make those 50,000 people slightly more productive.
The deeper opportunity is to redesign how the organization works.
The Enterprise Becomes a Software System
This is where things get interesting.
Traditional companies are mostly human networks.
Information moves through meetings, email, documents, Slack messages, dashboards, and management chains.
The organization is effectively executing a distributed human program.
AI allows parts of that program to become explicit.
| Instead of | The explicit version |
|---|---|
| Relying on humans to continuously move information between systems | AI agents can observe events and react |
| Managers manually assembling status updates | Organizational state can be continuously synthesized |
| Someone asking five teams why a metric changed | An agent can investigate the dependency graph |
| Knowledge disappearing when an employee leaves | Organizational memory can become persistent |
Over time, the enterprise starts to look less like a hierarchy of people and more like a software system composed of humans and AI agents.
Humans become nodes where judgment, creativity, accountability, leadership, and ambiguity matter most.
AI handles more of the connective tissue.
That could dramatically reduce the traditional penalty of organizational scale.
The Real Enemy Is Coordination Cost
This matters because company size itself is not necessarily the problem.
Coordination cost is.
A ten-person company where everyone needs constant meetings can be slower than a thousand-person company with excellent systems.
Conversely, a huge organization may contain enormous talent but spend most of its energy synchronizing itself.
AI has the potential to reduce that synchronization cost.
If it succeeds, enterprises gain a surprising advantage.
They retain scale.
They retain capital.
They retain distribution.
They retain specialized expertise.
But some of the bureaucracy required to coordinate all of those resources disappears.
That is an extremely powerful combination.
Maximum Velocity Changes
This connects directly to a broader question I have been thinking about: what determines the maximum velocity of a team or an organization?
For a long time, execution capacity was one of the main constraints.
How many engineers do we have?
How many designers?
How many analysts?
How quickly can they physically produce the work?
AI changes that constraint.
Execution becomes cheaper.
Generation becomes faster.
The cost of trying ideas falls dramatically.
As that happens, the bottleneck moves.
The maximum velocity of an organization becomes less constrained by how fast people can type, code, analyze, or assemble documents.
It becomes constrained by things like:
- how quickly the organization can make decisions
- how clearly intent can be expressed
- how much context is available
- how quickly feedback reaches the people or agents doing the work
- how much coordination is required between groups
- how good the organization’s judgment is
- how much ambiguity leadership can resolve
- and ultimately, how ambitious the organization is willing to be
AI does not remove the idea of a maximum organizational velocity.
It changes what determines it.
In a traditional organization, production capacity may have been the limiting factor.
In an AI-native organization, decision bandwidth may become the limiting factor.
That is an important distinction.
| You can | And still have |
|---|---|
| Give every engineer an AI coding agent | A slow company |
| Generate ten times more proposals | A leadership team capable of evaluating only the same number as before |
| Create hundreds of experiments | A product organization that takes weeks to decide which results matter |
The bottleneck simply moves upstream.
AI increases the theoretical speed of the machine.
The organization still needs to learn how to steer it.
Two Forces Moving Toward Each Other
We may therefore see two transformations happening simultaneously.
| At the bottom of the market | At the top | |
|---|---|---|
| The transformation | Individuals become organizations. | Organizations become more like individuals. |
| What it gains | The freelancer gains leverage. | The enterprise gains agility. |
| At company scale | The small company gains capabilities previously available only to large organizations. | The large company gains the ability to operate with the coherence of a much smaller one. |
Both sides move toward the same destination:
small numbers of humans directing enormous amounts of computational labor.
The difference is what resources surround those humans.
The Middle May Be the Most Vulnerable
This raises an uncomfortable possibility.
The companies under the greatest pressure may not be the largest enterprises or the smallest startups.
They may be organizations stuck in the middle.
Large enough to accumulate bureaucracy.
Small enough not to have overwhelming distribution or capital advantages.
Too dependent on traditional coordination structures.
Too slow to redesign themselves around AI.
These companies historically justified their size because they could coordinate more people than a small competitor.
But if small competitors suddenly gain AI-driven scale, that advantage shrinks.
At the same time, they may lack the data, capital, and infrastructure advantages of the largest enterprises.
That could create a difficult strategic position.
Headcount Becomes a Strange Metric
For decades, growing companies celebrated headcount.
“We’ve grown to 500 employees.”
“We’re now a 10,000-person organization.”
Headcount was often interpreted as evidence of success.
In an AI-native economy, that relationship becomes less obvious.
If two companies generate the same revenue and one requires 10,000 people while the other requires 1,000, which organization is healthier?
Historically, the answer depended heavily on industry.
Increasingly, investors, employees, and executives may ask a more uncomfortable question:
Why does this organization require so many humans to operate?
This does not mean companies should blindly reduce headcount.
That would be an extraordinarily shallow interpretation of AI.
The better question is:
What would this organization look like if we designed it today around the capabilities now available?
The answer may involve fewer people.
But it may also involve the same number of people accomplishing dramatically more ambitious things.
Human Ambition Is the Limiting Factor
This is the part I find most interesting.
The AI productivity conversation often assumes a fixed amount of work.
If AI makes programmers twice as productive, perhaps we need half as many programmers.
But that assumes we already have enough software.
I do not think we do.
There are many problems that remain unsolved because solving them has historically been too expensive.
Internal tools that never get built.
Tiny markets nobody serves.
Scientific experiments nobody can afford.
Products personalized for extremely small groups.
Software for obscure workflows.
Education adapted to individual students.
Businesses that would never justify hiring a ten-person engineering team.
When the cost of creation falls, we may not simply produce the same things with fewer people.
We may produce far more things.
The scarce resource becomes less about labor.
It becomes imagination.
Judgment.
Taste.
Courage.
And the ability to decide what deserves to exist.
The Freelancer of the Future
The best independent professionals may therefore look very different from today’s freelancers.
Their primary skill will not simply be execution.
It will be orchestration.
They will know how to decompose a problem.
How to assign work between humans and AI.
How to evaluate outputs.
How to build reusable systems.
How to maintain context.
How to automate repetitive work.
How to know when AI is wrong.
And perhaps most importantly, how to maintain responsibility for the final result.
Their leverage may become extraordinary.
One excellent independent professional may outperform what previously required an entire consulting team.
The Enterprise of the Future
The best enterprises will likely learn the same lesson.
They will not merely distribute AI assistants to employees and declare an AI transformation.
They will redesign workflows.
They will ask why information moves through five humans before reaching the person who needs it.
They will ask why documents are manually assembled.
Why decisions repeatedly lose their context.
Why teams independently rediscover the same knowledge.
Why organizational memory depends on who happens to attend a meeting.
Why reporting consumes entire layers of management.
Why humans serve as APIs between software systems.
The AI-native enterprise will systematically remove these inefficiencies.
And as it does, the organization may begin to move with a speed we currently associate with startups.
Company Size Stops Telling Us Very Much
We are entering a world where asking how many people work at a company may reveal surprisingly little.
A ten-person company may operate like a hundred-person company.
A hundred-person company may serve millions of customers.
A single consultant may compete with an agency.
And a giant enterprise may dramatically reduce its internal friction without shrinking its ambition.
What matters becomes something else.
How much intelligence can the organization coordinate?
How quickly can it turn information into decisions?
How quickly can decisions become action?
How much useful output can each human generate?
And how much organizational complexity exists between an idea and its execution?
Those may become far better measures of organizational power than headcount.
The Convergence
The freelance economy and the enterprise world used to sit at opposite ends of the organizational spectrum.
One optimized for independence.
The other optimized for scale.
AI weakens that distinction.
Independence can now have scale.
Scale can now have agility.
The most interesting organizations may therefore combine characteristics we previously considered contradictory:
large capabilities with small teams,
deep specialization with broad individual autonomy,
global reach with founder-like speed,
and enterprise-grade reliability with startup-like experimentation.
The future may not belong specifically to freelancers.
And it may not belong specifically to enterprises.
It may belong to whoever learns fastest how to organize intelligence.
Human and artificial.
At any scale.