Sam Altman's WHAT vs HOW: The 5% of CEO Time That Actually Matters

No Master Plan

Asked at Sequoia Capital's AI Ascent 2025 how OpenAI plans its enormous ambitions, Sam Altman gave an answer most strategy decks would treat as malpractice:

"We're going to go make great models and ship good products, and there's no master plan beyond that."

This is not a scrappy startup talking. It is the company that triggered the AI era, describing itself at massive scale. And it isn't new: Altman described early OpenAI in the same conversation as almost impossible to overstate how much it was

"…a research lab with a very strong belief and direction and conviction, but no real kind of like action plan."

Now hold that against your own operating rhythm. Most leadership teams run the opposite pattern: an elaborate plan, weakly held convictions, and a quarterly fight about both. Altman's model inverts it — be stubborn about what matters, be flexible about how you get there — and the rest of this post is about what that actually takes, including the part of it his own company's history complicates.

The 5% / 95% Rule

The framework is older than OpenAI's fame. In his interview for Elad Gil's High Growth Handbook — published in 2018, while Altman was still president of Y Combinator and OpenAI had roughly 80 employees — he defined the job in one sentence:

"The role of the CEO is basically to figure out and decide what the company should do and then make sure it does that."

Then he split the time between the two halves:

"The hard part is that most people want to just do the first part, which is figure out what the company should do. In practice, time-wise, I think the job is 5% that and 95% making sure that it happens."

Read the qualifier carefully. Altman did not say strategy is worth 5% and execution is worth 95%. He said time-wise. And he was specific about where the 95% goes — not into heroic problem-solving, but into repetition:

"The way you make it happen is incredibly repetitive. It's a lot of the same conversation again and again with employees or press or customers. You just have to relentlessly say, 'This is what we're doing, this is why, and this is how we're going to do it.' And that part — the communication and the evangelizing of the company vision and goals — is time-wise by far the biggest part of the job."

That gives the rule its honest reading: the highest-leverage act of leadership may be choosing what matters — but almost all leadership time is spent making that choice real in other people's heads.

The WHAT Is Still the Leveraged Part

It is easy to hear 5%-versus-95% and conclude that execution matters and strategy is overrated. Altman's own writing explicitly rejects that reading. From his essay Productivity:

"It doesn't matter how fast you move if it's in a worthless direction. Picking the right thing to work on is the most important element of productivity and usually almost ignored. So think about it more!"

His later essay How to Be Successful makes the same two-step argument without the numbers: spend more time than feels natural deciding what to focus on, because working on the right thing beats working more hours — and once the priority is chosen, become extremely difficult to stop. The 5% isn't the unimportant part. It's the part every subsequent hour compounds on: months of flawless execution against the wrong priority just reaches the wrong destination faster. That is the core discipline of strategic decision-making — the quality of the choice sets the ceiling on the value of everything downstream.

The 95% Is an Alignment Problem, Not a Control Problem

The second common misreading treats making sure it happens as the CEO personally gripping every decision. Altman argued against that in the same 2018 interview:

"At some point your job becomes more about hiring people and working with them to get what you want done than doing it yourself."

By 2025, the same principle shows up at OpenAI scale as an argument about organizational shape. Companies, he told the Sequoia audience, tend to grow headcount faster than output — and the cure is structural:

"You want everyone to be busy. You want teams to be small, you want to do a lot of things relative to the number of people you have. Otherwise, you just have, like, 40 people in every meeting."

So the 95% is not CEO decides → CEO controls everything → team executes instructions. The healthier loop is: leadership sets direction → teams get real ownership → leadership maintains alignment through repetition → teams determine most of the how. On the question of when to override that and coordinate top-down, Altman drew the boundary himself:

"There are some projects that require so much coordination that there has to be a little bit of, like, top down quarterbacking. But I think most people try to do way too much of that."

Intervene heavily when coordination genuinely requires it; don't turn every decision into a coordination problem. It is the same discipline behind Steve Jobs staying out of Pixar's early Braintrust meetings — the leader's presence, applied indiscriminately, degrades the very process it's meant to steer.

How ChatGPT Actually Happened

If the WHAT/HOW split were clean, OpenAI's history would read as plan-then-execute. It doesn't. Nobody sat down in 2016 with a six-year roadmap ending in a chatbot. The company experimented with video games and a robotic hand. Researchers drifted toward unsupervised learning and language models. GPT-3 shipped as an API partly because OpenAI wasn't sure what product to build with it.

Then came the observation that changed everything. Watching how developers actually used the API, the team noticed something the roadmap never predicted:

"People love to talk to it in the Playground… that was the kind of only killer use, other than copywriting, of the API product that led us to eventually build ChatGPT."

Sit with what that means. The most consequential product launch of the decade came not from the 5% — a leadership team choosing WHAT — but from the 95% generating evidence nobody planned for. The direction persisted; the implementation emerged; and then the implementation quietly rewrote the direction.

Four Practices That Make the Split Work

Taken together — the 2018 rule, the 2025 operating description, and the ChatGPT counterexample — the framework turns into four concrete practices:

Where Management Theory Pushes Back

There is a serious case against dividing leadership into WHAT and HOW at all. Roger Martin has spent years attacking the distinction — in The Execution Trap (HBR, 2010), Stop Distinguishing Between Execution and Strategy (HBR, 2015), and Strategy and Execution Are the Same Thing (HBR, 2016). His argument: so-called execution is a cascade of choices, and choices are strategy. A brilliant strategy poorly executed is, on inspection, usually just an incomplete strategy.

Notice that Altman's own elaboration half-concedes the point. His 95% isn't mechanical implementation — it's evangelizing the vision, deciding again and again what to emphasize, to whom, in what words. By Martin's lens, that is strategy work. The 5/95 split survives as a description of where the hours go, not as a wall between two kinds of thinking.

Henry Mintzberg and James Waters go deeper. Their classic 1985 framework distinguishes deliberate strategy — intended and executed as planned — from emergent strategy: the pattern an organization discovers in its own actions. Real strategy, they showed, always lives on the continuum between the two.

Which is precisely OpenAI's story. Conviction, experiments, information, adjustment — until the experiments produced a product that redefined the company. The HOW helped discover the WHAT. Any version of Altman's framework that can't absorb that fact is a version his own company refutes.

The Scale Question: 80 Employees vs 3,500

There is a harder question the 5%/95% framing doesn't answer: does the ratio survive scale? When Altman gave the original quote in 2018, OpenAI had roughly 80 employees. By 2025, when he spoke at Sequoia, it had grown to over 3,500 — a 40× increase. The nature of the 95% changes dramatically across that gap.

At 80 people, a CEO can repeat the vision in every all-hands, every hallway conversation, every one-on-one. The 95% is literally the CEO's voice, the CEO's time, the CEO's repetition. At 3,500 people — let alone the 7,000+ OpenAI is projected to reach — that becomes physically impossible. The CEO speaks to the executive team; the executives speak to their directors; the directors speak to their managers; the managers speak to their teams. The repetition necessarily cascades through layers.

This changes the CEO's job in one of two ways. Either the 95% becomes alignment of the aligners — the CEO spends most of the time ensuring that the people doing the repetition are saying the same thing — or the ratio shifts. At scale, more of the CEO's time may move toward the 5%: more strategic decisions, more bets, more external representation, while a layer of executives and chiefs of staff handle the communication cascade that once consumed the founder's calendar.

Altman's framework may still hold as a description of what the organization does — 5% direction-setting, 95% alignment work — but at scale, it is no longer a description of what the CEO personally does. The honest reading: the 5%/95% rule is most literally true for founder-CEOs of companies under a few hundred people. Beyond that, the rule describes the work, not necessarily the CEO's calendar.

A Better Frame: Direction → Execution → Learning

So the most useful version of the framework isn't a static 5/95 split. It's a loop:

Then repeat. This is also where a documented decision process pays off twice: the record of why you chose the current direction is exactly what tells you, later, whether the assumptions behind it still hold — the connection between this loop and the decision quality chain.

Stubborn About What Matters

The 5%/95% line has survived seven years of retelling because it names something most new leaders discover painfully: deciding what should happen can take an afternoon; making an organization actually do it consumes nearly everything else.

But Altman's more recent account adds the half the aphorism leaves out. The 95% isn't the faithful execution of a fixed plan — it's building an organization that moves fast toward a clear objective while continuously adapting its tactics, and occasionally its assumptions, as reality reports back. That's a harder problem than either strategy or execution alone. It's also the one great companies actually solve.

Be stubborn about what matters. Be flexible about how you get there. And build the feedback loops that tell you when the what itself needs to change.