Sequoia Is Right. And Still Incomplete.

The Error Is Not in Execution. It Is in the Decision.

On March 5, 2026, Sequoia Capital published an essay that quickly became a reference point in the debate over the future of software: "Services: The New Software." Written by Julien Bek, the piece makes a claim that, at first glance, sounds merely provocative. It is not. It is structural.

"The next $1T company will be a software company masquerading as a services firm."

The idea is simple enough to be ignored by those who do not pay attention, and deep enough to redefine capital allocation.

For decades, software sold tools. The customer bought capacity. Potential. Access. The work, however, was still done by people. Excel does not deliver the report. The CRM does not close the deal. Accounting software does not close the books. It merely organizes human effort.

Sequoia is saying this model is over.

The new software does not sell the tool. It sells the work done.

And by doing so, it shifts the center of value capture. Because the relevant budget was never software. It was always labor. Sequoia illustrates this with an asymmetry that, once seen, cannot be unseen: companies spend a fraction on software and multiples of that on the services required to operate that software. The expensive part is not the system. It is the person using it.

The consequence is unavoidable. Once software begins to execute, it stops competing with other software and starts competing with people. And people are the largest cost inside an organization.

So far, the thesis is correct. More than correct. It is inevitable.

But this is where the argument begins to fail.

Sequoia describes a transformation at the level of execution. What it does not describe, and perhaps has not fully perceived yet, is that execution was never the main problem inside companies.

Companies do not fail because they cannot execute tasks. They fail because they execute the wrong tasks. Because they prioritize badly. Because they interpret signals inconsistently. Because they make decisions under noise, not under system.

Automating execution does not solve this. It amplifies it.

If the decision is right, automatic execution is a multiplier of efficiency. If the decision is wrong, automatic execution is a multiplier of error.

This distinction is uncomfortable because it moves the problem to a less visible place. Execution is tangible. It can be measured, optimized, automated. Decision cannot. Decision is diffuse, contextual, dependent on memory, feedback, and interdependence between variables. It does not fit neatly into a simple interface.

But precisely for that reason, it concentrates value.

There is a fundamental difference between markets and companies that is rarely examined with the seriousness it deserves.

Markets are systems.

Companies are organized improvisations.

In financial markets, principles are tested, refined, transformed into rules. Rules are encoded into systems. Systems execute. The human interacts with the system, but the human is not the primary decision point. The system is.

Inside companies, the process remains incomplete. Principles exist: strategy, culture, directives. But they do not become consistent operational rules. They do not become systems. The result is predictable: decisions fall back onto individuals. And individuals do not scale.

This creates a pattern that repeats with almost statistical regularity. The same types of companies, with the same leadership profiles, under similar conditions, produce similar results. And still, they continue to operate as if every decision were unique, as if every problem were new, as if isolated human judgment were sufficient.

It is not.

Sequoia's thesis misses this point because it starts from an implicit assumption: that the work to be executed has already been correctly defined.

In practice, it has not.

Most of the invisible cost inside a company is not in execution. It is in defining what should be executed. It is in decisions made too late, too early, or simply wrong. It is in opportunities not prioritized, risks not detected, actions not initiated.

This does not appear in the budget. It is not classified as a cost line. But it determines the result.

And, more importantly, it is not solved by task automation.

There is a structural limit to the "services-as-software" model that few are willing to admit.

Execution can be compressed. It can be standardized. It can be replicated. As models advance, every operational workflow tends toward automation. This is not a hypothesis. It is a direction.

But what happens when everyone automates execution?

Execution loses differentiation.

If everyone can close the books automatically, the value is no longer in closing the books. If everyone can generate outreach, screen candidates, answer tickets, the value is no longer in those tasks. It is in deciding which books matter, which candidates to hire, which customers to prioritize, which actions to take.

Execution converges.

Decision does not.

Decision depends on accumulated context, operational memory, relationships between events, and the interpretation of weak signals. It is not merely calculation. It is structure.

And structure is not easily copied.

The error in the thesis is not what it says. It is what it omits.

It describes the industrialization of execution.

But ignores the systematization of decision.

And that is the level where the game is actually decided.

Today, most corporate systems operate through a broken flow. Data is collected. Interfaces organize the data. Humans interpret. Humans decide. Systems execute. Feedback is fragmented and rarely incorporated in a consistent way.

Sequoia's implicit proposal removes part of this flow. It removes the human from execution. But it keeps the human in the most critical point: the decision.

That is partial optimization.

The complete system is something else.

A complete system does not separate data, decision, and execution. It integrates them. It connects context, preserves memory, structures decisions, and closes the loop through execution and continuous feedback. It does not depend on episodic intervention. It does not depend on isolated interpretation. It operates as a cycle.

Detect. Interpret. Decide. Execute. Measure. Adjust.

Without rupture.

The consequence of this shift is less technological than strategic.

Whoever controls execution gains efficiency.

Whoever controls decision gains direction.

And direction precedes efficiency.

It is possible to execute something irrelevant perfectly. It is impossible to decide correctly without context.

The real control of an organization is not in who does the work. It is in who defines the work.

The trajectory of software makes this clear.

Traditional SaaS increased productivity. The new generation, driven by AI, begins to replace labor.

The next stage will not merely replace labor.

It will replace judgment.

Not in the philosophical sense. In the operational sense.

Systems capable of deciding, consistently, what should be done next, based on memory and context, not merely on instruction.

This does not eliminate the human. It repositions him.

The human stops being the executor. He stops being even the primary decision-maker. He becomes the supervisor of the system.

There is a risk attached to this transition, and it is rarely discussed with the seriousness it deserves.

Automating execution without systematizing decision does not merely fail to solve the problem. It amplifies it.

If a company already makes inconsistent decisions, giving it a system that executes faster does not improve the result. It accelerates the error. It reduces the time between bad decision and consequence.

Efficiency, in this context, is dangerous.

Without decision coherence, it is only speed.

Sequoia Capital is right to say that software will stop selling tools.

But that is not what defines the next generation of dominant companies.

What defines it is not the ability to execute work.

It is the ability to decide work.

Services-as-software solves the problem of execution. But it does not solve the problem of direction.

And without direction, execution is irrelevant.

The final distinction is simple, but not trivial.

Software that executes wins workflows.

Software that decides controls companies.

Leo Bentier