If We Built the Company From Zero Today, Would We Create the Same Departments?

I keep coming back to a simple thought experiment.

If we were starting the company today—with today’s technology, today’s customer expectations and no inherited structure—would we organize the work the same way?

Would marketing, sales, product, technology, operations and finance have the same boundaries?

Would the same approvals exist? The same handoffs? The same management layers? The same ownership of customer information?

Some of them would.

I am increasingly convinced some would not.

The organization chart tells you who reports to whom. The operating model tells you how decisions, information and value actually move.

Departments solved real problems.

I do not think functional organizations are obsolete.

Specialization creates depth. Finance should understand finance. Engineering should understand engineering. Marketing should know markets and customers. Functional leaders create standards, talent systems and professional discipline.

Departments also reduce complexity. A business cannot have every person coordinate directly with every other person.

But organizational forms are partly responses to the economics of information and coordination.

What organizational economics tells us

Research on information technology and firm design has long found that changes in information and communication costs can alter employee autonomy, managerial spans and centralization. More recent NBER work finds AI investment is associated with meaningful changes in workforce composition and flatter hierarchical structures inside adopting firms.

That does not mean AI automatically eliminates middle management or collapses functions.

It means the economic assumptions underneath some structures are moving.

Coordination is getting cheaper.

A surprising amount of organizational work exists because information is fragmented.

Someone extracts the number. Someone reconciles it. Someone turns it into a deck. Someone schedules the meeting. Someone explains the decision to another function. Someone updates the system after the meeting.

Those are not useless activities. They are the cost of coordinating specialized people and systems.

When technology reduces that cost, the organization gets a choice.

It can use the savings to make the same structure move faster.

Or it can reconsider the structure itself.

A current signal

NBER research using U.S. firm data finds that information technology can enable “scale without mass,” with revenue growing faster than employment as routines become easier to replicate. Another 2026 NBER study of AI diffusion finds adoption is still concentrated in relatively few business functions, suggesting most firms are early in the transition from task-level use to broader operating change.

The danger is optimizing the functions while losing the outcome.

I see this most clearly in growth.

Marketing can optimize leads. Sales can optimize activity. Product can optimize engagement. Finance can optimize spend. Operations can optimize utilization.

Each function can hit its own metric while the customer or commercial outcome gets worse.

The problem is not that the functions are badly managed.

The problem is that the business outcome crosses their boundaries.

From operating experience

During digital transformation work, the biggest gains did not come from making one department more “digital.” They came from connecting acquisition, CRM, analytics, product, technology and revenue so the business could learn and act as one system.

That operating-model redesign contributed to approximately $2.8 million in annualized cost reductions while the digital revenue mix moved from roughly 20% toward 70%.

The lesson for me was not that departments should disappear. It was that the outcome should determine the architecture of the work before the org chart determines the handoffs.

AI makes the old boundaries more visible.

Most companies are currently adding AI inside existing functions.

Marketing gets AI tools. Finance gets AI tools. Sales gets AI tools. HR gets AI tools.

That is a rational adoption path because budgets, systems and accountability already live inside those functions.

But if AI reduces the cost of analysis, communication and coordination, the larger opportunity may sit between functions.

Why adoption may require redesign

OECD research warns that many managers still approach AI as a plug-and-play technology even though meaningful adoption can require changes across departments and business operations. NBER work on AI adoption similarly argues that interdependent tasks can make system-level redesign more valuable than isolated task automation.

That is where I think leaders need to be careful.

If we automate each silo independently, we may create faster silos.

The customer still crosses them.

Operating flowStart with how value and decisions move—not who reports to whom.

Operating model schematic

01Market / customer signal

02Decision

03Work across capabilities

04Customer / business outcome

05Evidence

The evidence should change the next decision. Functions support the flow; they should not obscure it.

I would organize around outcomes first.

If I were designing more of the operating model from zero, I would start with the outcomes and the decision loops.

How does a market signal become a decision?

How does a customer need become a product change?

How does a revenue target become resource allocation?

How does execution produce evidence, and how does that evidence change the next decision?

Then I would ask which capabilities need deep functional homes and which should be organized around cross-functional systems.

My hypothesis

The next organizational advantage may not come from eliminating functions. It may come from making the boundaries between them less expensive to cross.

What I would not do.

I would not flatten an organization because “AI removes managers.”

I would not create cross-functional teams for everything. Teams without clear decision rights can create more coordination cost, not less.

I would not confuse collaboration with accountability.

And I would not reorganize around the capabilities of today’s AI models. The technology will move faster than an org redesign.

The durable question is not which jobs the current model can perform.

It is where information, judgment and accountability need to sit for the business to make better decisions.

Questions I think leaders should ask.

Which handoffs exist because information used to be difficult to move?

Where do multiple functions influence the same outcome without one shared operating measure?

Which management layers primarily collect, translate or route information?

Where would lower coordination cost change the span of control or decision rights?

If we were starting today, which capabilities would we centralize—and which would we embed closer to the outcome?

What I may be wrong about.

Organizational structure tends to change more slowly than technology for good reasons.

People need identity, career paths and clear authority. Regulated work needs control. Complex businesses need specialization. A function can be an effective way to preserve expertise even when workflows cross it.

So I do not expect the org chart to disappear.

I expect the difference between the org chart and the operating model to become more important.

The question is not whether departments survive AI. The question is whether leaders continue treating inherited functional boundaries as the natural architecture of the work.