The Founder Is the Last Gatekeeper

The bottleneck in a growing startup is often the person who built it.

THE IDEA IN ONE LINE

At ten people, the founder is the advantage. At forty, the founder is the queue.

In the first year, everything goes through the founder, and that is why the company is fast. The founder hears every customer call, approves every price and answers every question before lunch. One person holds the whole picture, and decisions take minutes.

Three years and forty people later, the habit is the same and the result is the opposite. The team waits for the founder’s view on a price change, a hire, a customer request, a feature. Each decision still takes minutes. Getting to the front of the queue takes a week.

The founder has not changed. The company has. The path that once made everything faster is now the one every other path has to pass through.

How the founder became the gate

Henry Mintzberg called the early form the simple structure. Coordination happens by direct supervision: “the chief executive officer gives the orders.” Its strength, he wrote, is “rapid, flexible innovation, at least of the simple kind.” Its weakness is just as plain. Everything depends on one person (Mintzberg, 1981).

In a small company that is a fair trade. In a growing one, the founder stops being one path among many and becomes the hub. Every team has a direct line in. Many of the paths between teams quietly run through the founder too, because it is quicker to ask the one person who knows everything than to find the colleague who owns the answer.

The efficiency equation has a term for this.

THE TERM A FOUNDER CAN BECOME
η=(Pₛ × I) + (R × A)eᵁ + N

Network complexity (N) counts the gatekeepers and approval steps between the product and a paying customer. A founder who approves everything is a gatekeeper on every path.

The Entrepreneurial Efficiency Equation (η) · Dr. Hafiz Muhammad Ali

This is the third source of N in a growing company. Customers add it through exceptions. New hires add it through paths. The founder adds it by standing in the middle of both.

What the queue costs

Jay Galbraith described the mechanism in 1974. Organisations handle routine work through rules and plans, and refer the exceptions up the hierarchy. As uncertainty grows, the exceptions multiply until the top of the hierarchy is overloaded (Galbraith, 1974). In a startup, the top of the hierarchy is one calendar.

The cost is not only delay. When decisions wait, work waits with them. Teams prepare three options because they cannot tell which one the founder will pick. They hedge instead of committing. In time they stop proposing decisions they expect to be overruled, and the company slows to the speed of one person’s attention.

Larry Greiner mapped this turning point more than fifty years ago. Growing companies pass through phases, each ending in a crisis. The first, a crisis of leadership, is usually solved by strong direction from the top. The second is a crisis of autonomy: the people closest to the work know more than the centre does, and the centre does not let go (Greiner, 1972).

AI moves the queue

A newer version of this pattern is spreading. Instead of hiring, founders use AI for work a team once did: drafting, coding, analysis, customer replies. Fewer people means fewer paths, and that is a real gain.

But the decisions do not disappear. Work an employee would have owned now comes back to the founder to check, correct and approve. AI removes the paths between people and routes more of the judgement to one person. The queue moves from the founder’s calendar to the founder’s review.

Checking has a cost of its own, and it is easy to underestimate. In a 2025 trial by METR, sixteen experienced open-source developers took 19 per cent longer on tasks where AI tools were allowed, yet believed afterwards that AI had sped them up by 20 per cent (METR, 2025). METR calls this a snapshot of one setting, not a general law. The narrower point holds: review takes time, and the person doing it rarely counts it.

The routing rule still applies. Let AI and the team handle two-way doors end to end, with checks built into the process rather than into your inbox. Keep your review for the one-way doors, where judgement is the point.

Control has a price

Holding on is rarely vanity. Founders keep decisions because they care, because they are often right and because some have been burned by delegating too early. But control carries a measurable cost.

Noam Wasserman studied 6,130 American startups. Each additional level of founder control, whether holding the CEO role or controlling the board, reduced the company’s pre-money valuation by 17.1 per cent (Wasserman, 2017). His earlier work framed the choice as rich versus king: founders who keep control tend to build less valuable companies, and those who build the most value tend to give some control up (Wasserman, 2008).

The case for staying close

There is a strong counterargument. In 2024 Paul Graham described what he called founder mode. He recounted Brian Chesky’s experience at Airbnb, where following the standard advice to “hire good people and give them room to do their jobs” produced results that were, in Graham’s word, “disastrous.” Founders, he argued, should not treat parts of their company as black boxes, and skip-level meetings should become normal rather than unusual (Graham, 2024).

He is right that stepping back entirely is its own failure, and even he accepts that “there’s going to have to be some amount of delegation.” The question is not whether the founder stays involved. It is whether the founder is involved as a source of attention or as a checkpoint. Skip-level meetings add information. Approval gates add queues.

Route decisions, not people

The practical move is to sort decisions rather than hand them all away. Jeff Bezos drew the line clearly. Some decisions are “one-way doors”, consequential and hard to reverse, and deserve care. Most are “two-way doors”, which “can and should be made quickly by high judgment individuals or small groups” (Bezos, 2016).

Add a second question, how many teams the decision touches, and you have a routing rule. A reversible decision inside one team belongs to that team. A reversible decision that crosses teams belongs to the leads, with the founder informed. Only the one-way doors need to reach the founder at all.

Who decides what
Who decidesWhen the decisionFor example
The team
Can be undone, and stays inside one team
A pricing testA new onboarding email
The leads
Can be undone, but crosses teams
Moving a launch dateSharing an engineer
You
Hard to undo, or sets the company’s direction
A new marketA senior hire, a raise
Most decisions belong in the first two lanes. Only the last one needs the founder.

Routing works only if ownership is named. For each kind of decision, someone must be able to say “this is mine”, and the founder must be content to hear about it afterwards. Once the route is clear, the decision itself still has to be made fast, without waiting for consensus.

WHAT STILL RUNS THROUGH YOU

Which of your decisions need to reach you at all?

  1. What did you approve last week that someone else could have decided?

    Write the list down. Most founders find it longer than they expected.

  2. How long does a decision wait for your calendar?

    If the answer is days, the queue costs more than your judgement adds.

  3. Can the decision be reversed?

    If yes, it should not need you. Ask to be told afterwards.

  4. Who owns it when you are not in the room?

    If no one does, the decision has no route except through you.

  5. Do you need to be asked, or only told?

    Most of what founders approve, they only need to know about.

  6. Which AI outputs wait for your review?

    If most of them do, AI has given you more to approve, not less to do.

As the company grows, the founder’s job changes from answering every question to deciding which questions reach them. That is not stepping back. It is choosing where the founder’s attention is worth more than the team’s speed.

Stay close to the work. Stop being the door it has to pass through.

References

  • Bezos, J. P. (2016). 2015 letter to shareholders. Amazon.com, Inc.
  • Galbraith, J. R. (1974). Organization design: An information processing view. Interfaces, 4(3), 28–36.
  • Graham, P. (2024, September). Founder mode. https://paulgraham.com/foundermode.html
  • Greiner, L. E. (1972). Evolution and revolution as organizations grow. Harvard Business Review, 50(4), 37–46.
  • METR. (2025, July 10). Measuring the impact of early-2025 AI on experienced open-source developer productivity. https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
  • Mintzberg, H. (1981). Organization design: Fashion or fit? Harvard Business Review, 59(1), 103–116.
  • Wasserman, N. (2008). The founder’s dilemma. Harvard Business Review, 86(2).
  • Wasserman, N. (2017). The throne vs. the kingdom: Founder control and value creation in startups. Strategic Management Journal, 38(2), 255–277.
CONTINUE READING
THE FOUNDER LETTER

Think past
the obvious.

A monthly note on founder judgement, structural constraints, and building companies that remain coherent as they change.

Monthly. Considered. No noise.