The Founder Bottleneck

When Box had about 35 employees, Aaron Levie still wanted to be consulted about almost everything. If a button moved, if some wording changed or if a colour was adjusted, he expected to be involved. Performance reviews, pay rises and other approvals also made their way to him.

This was not happening because Levie was an unusually controlling founder. It was happening because his attention to detail had helped Box survive its earliest years. He knew what the product should feel like and could spot decisions that might move it away from that vision. While the company was small, bringing those decisions to him was often the fastest way to resolve them.

But the volume of decisions kept increasing. Levie later described the point at which this approach broke down: “everything that could possibly happen in the organization had to flow through me.” By his own account, the founder had become a significant bottleneck at roughly 35 people.1

The problem for many founders is behaviour which creates a bottleneck does not feel like a problem.

The company learns to wait

In the earliest stage of a company, centralisation is often sensible. There are few people, little spare cash and no established way of making decisions. The founder may be the only person who understands the customer, the product, the finances and the history behind earlier choices.

When a team member asks for approval, the founder can usually answer quickly. When something goes wrong, the founder steps in and corrects it. This feels efficient because, at that size, it often is.

The difficulty is that the organisation starts adapting itself to this behaviour. Employees learn which decisions they can make and which ones might be questioned later. If those boundaries are unclear, taking initiative becomes risky. Waiting is safer than making the wrong call and having the work reversed.

Gradually, decisions that could have been made elsewhere begin moving upwards. The founder’s experience reinforces the pattern. People appear reluctant to take responsibility, so the founder becomes more involved. The more involved the founder becomes, the fewer opportunities the team has to develop its own judgement.

Research into decision authority in startups suggests that this is common. A 2024 study examining 241 startups found that strategic decision-making remained highly centralised in many of them, with the chief executive frequently retaining the final say. The researchers also identified the underlying trade-off. Delegation can make use of knowledge held throughout the company and allow decisions to happen in parallel, but it creates a risk of misalignment and loss of control.2

Founders are not imagining that risk. They may have seen employees make confident decisions without understanding the consequences. They may also have learned that an apparently minor choice about a customer, hire or feature can carry much larger implications.

The mistake is not caring about those implications. It is designing a company in which the only defence against a bad decision is the founder’s continued involvement.

Amazon turned Bezos’s judgement into a system

Jeff Bezos faced an early version of the same problem at Amazon. Former Amazon executives Colin Bryar and Bill Carr recount that, until early 1997, Bezos interviewed every salaried candidate and retained the final hiring decision. In a very small company, this gave him direct control over the quality of the people being hired. As Amazon grew, he could no longer participate in every interview.3

The obvious response would have been to withdraw from hiring and trust each manager to apply their own standard. Amazon took a more deliberate approach. It built a process intended to preserve the principle Bezos cared about, that every new hire should raise the quality of the company.

In 1999, Amazon introduced what became known as the Bar Raiser programme. Trained interviewers participated in hiring discussions as objective third parties. They were not there to reproduce Bezos’s personal opinion about every candidate. Their role was to protect a shared hiring standard and challenge decisions that appeared driven by urgency, convenience or local pressure.4

This is an important distinction. Bezos did not solve the problem by ceasing to care about hiring. Amazon separated the founder’s underlying judgement from the founder’s personal availability.

The principle remained. The mechanism changed.

For a small company, the mechanism does not need to be as formal as Amazon’s. It could be a written scorecard, a trained interviewer from outside the hiring department or a rule that every candidate must demonstrate how they raise the standard in a defined area. What matters is that the organisation no longer needs the founder in every interview to remember what good looks like.

Delegating the decision is not enough

Founders are often advised to “let go”, as though delegation were mostly an emotional hurdle. Sometimes it is. More often, the founder has knowledge that has never been made explicit.

Consider a salesperson asking whether they can offer a customer a discount. The founder may understand that the decision depends on the type of customer, the length of the contract, the support burden, the likelihood of expansion and whether the concession will become a precedent. When the founder gives a quick answer, all of that reasoning remains hidden.

If the salesperson is simply told to take responsibility next time, they have been given authority without the context required to use it. If they make a poor decision, the founder concludes that they were not ready. Control returns to the centre.

This helps explain why trust is such an important part of delegation. A 2026 study of 186 entrepreneurs, supported by responses from 47 co-founders, found that founders were more inclined to delegate to employees they perceived as capable and trustworthy. It also found that greater perceived venture risk discouraged delegation. Highly ambitious founders were less inclined to delegate, although the researchers noted limitations in how ambition was measured.5

The finding is intuitive. When the stakes feel high, keeping decisions close feels protective. Yet this is precisely when a company can become dangerously dependent on one person.

The practical alternative is to delegate judgement in stages. The employee first brings a recommendation and explains the reasoning. Later, the employee makes the decision and informs the founder. Eventually, the employee acts independently within an agreed boundary and only escalates exceptions.

That progression allows the founder to see how the employee thinks. It also allows the employee to absorb the considerations that previously existed only in the founder’s head.

Google discovered that removing managers was not the answer

There is another trap here. Once founders recognise that too many decisions are moving upwards, they may conclude that the answer is to remove hierarchy and make the company completely flat.

Google tried something close to this while it was still a comparatively small company. Larry Page had helped grow it to more than 200 employees by 2001.6 According to an account later given by Google’s former head of people operations, Laszlo Bock, Page removed the company’s engineering managers in an attempt to reduce bureaucracy and allow engineers to work more directly.

The experiment was reversed. One executive ended up with an impractical number of direct reports, while everyday questions and disagreements still needed somewhere to go.7

The episode matters because it exposes a false choice. A company does not have to choose between founder control and the absence of management. Removing managers does not remove the need for decisions, coordination, feedback or conflict resolution. It simply makes it less clear who is responsible for providing them.

A founder bottleneck is therefore not solved by telling everybody to act like an owner. Employees still need to know what they own, which constraints matter and where one person’s authority ends and another’s begins.

Good delegation requires more structure, not less. The structure can be light, but it must make responsibility visible.

Stripe taught people how the founders thought

Stripe offers another version of this transition. In its early years, founders Patrick and John Collison were closely involved in product and design decisions. According to former Stripe chief operating officer Claire Hughes Johnson, the founders and early employees developed design review checkpoints that allowed them to teach the questions and criteria behind their decisions.

Over time, design leaders could take over the reviews. The founders’ judgement had not disappeared, but the company no longer depended on them examining every design personally.

Stripe applied a similar approach to commercial decisions. In the beginning, senior leaders were involved in setting the terms of deals. As similar deals recurred, the company standardised the common decisions and reserved leadership attention for unusual cases. Johnson argued that leadership would have been failing if it still needed to review every deal as the company grew.8

This is what transferring context looks like in practice. The founder identifies the recurring questions behind a decision and turns them into principles, examples and boundaries. Responsibility can then move without requiring the company to forget what the founder has learned.

It also gives the founder a better role. Instead of correcting every individual output, the founder can improve the system that produces those outputs.

What should still reach the founder?

The answer is not “nothing”. Some decisions genuinely belong with the founder, particularly while the company is small.

These tend to include decisions that could change the identity or survival of the business: raising capital, changing the company’s strategic direction, hiring a senior leader, entering a large or unusual partnership, making a commitment that is difficult to reverse or taking a risk that could threaten the company.

The rest can usually be divided into three groups:

1. Decisions someone else can make independently within an agreed boundary.

2. Decisions someone else can make, provided they inform the founder afterwards.

3. Decisions where someone else prepares the recommendation but the founder retains the final call.

This is more useful than a vague instruction to be more empowering. It forces the founder to distinguish between decisions that require their judgement and decisions they have simply become accustomed to making.

A short audit can make the problem visible. For two weeks, the founder records every decision, approval and request brought to them. At the end, they examine which items genuinely needed founder involvement, which were escalated because ownership was unclear and which could have been resolved using an agreed rule.

The founder should then choose one recurring category, not twenty, and transfer it properly. That means naming the owner, describing the desired result, explaining the important constraints and agreeing which exceptions should return to the founder. The early decisions should be reviewed so the rule can be improved.

The objective is not to reduce the founder’s workload for its own sake. It is to redirect that workload towards decisions that alter the future of the company.

A preliminary 2026 study using a representative survey of German entrepreneurs found that delegation was associated with greater incremental innovation. Interestingly, the researchers did not find that entrepreneurs simply worked less. Instead, delegation appeared to shift their attention towards strategy, market understanding, growth and investment decisions.9

This remains associational evidence, not proof that delegation automatically produces innovation. It does, however, challenge the assumption that delegation is mainly a way for founders to reclaim their evenings. Its more important purpose may be to change the work the founder is able to do.

The founder’s next product

When Aaron Levie brought in Dan Levin as Box’s chief operating officer in 2010, Levin began identifying the decisions and processes that were dependent on the founder. Levie later said that the appointment increased his leverage dramatically and helped the company resume its growth.1

The lesson is not that every 35-person company needs a chief operating officer. A senior hire cannot compensate for a founder who continues to reserve every meaningful decision. Nor should founders hand over decisions merely to demonstrate that they can.

The deeper lesson is that a growing company needs a way to preserve the founder’s useful judgement without requiring the founder’s presence everywhere.

Amazon created trained guardians of its hiring standard. Stripe turned recurring judgement into principles, review processes and exception rules. Google learned that removing managers did not remove the work of management. Box confronted the point at which one person could no longer remain at the centre of every important conversation.

In each case, the challenge was not to make the founder less important. It was to stop the company from being limited by the founder’s available hours.

A founder’s first product may be the service, technology or customer experience that gets the company started. The next product is the company itself: a system in which other people can make good decisions, learn from imperfect ones and move without waiting for the founder to enter the room.

You May Be Creating Office Politics Without Realising It

Everyone says they hate office politics and a small company is supposed to be different. People sit together, roles remain flexible and anyone can speak directly to the founder. There is no time for politics because everyone is busy building the business.

That informality can work beautifully when the whole company fits around one table. Everybody hears the same conversations. The founder knows what each person is doing, and disagreements can be resolved while the relevant context is still fresh. There is little need to describe how decisions happen because everybody can watch them happen.

Growth weakens that shared understanding before most founders notice. New employees did not hear the original conversations. Work begins to cross teams. A decision that seems obvious to one person looks like an intrusion to another. The founder remains accessible, but can no longer know what has already been discussed before somebody approaches them.

Politics enters through this gap. It rarely arrives as a deliberate attempt to manipulate the company. More often, it is the practical response of people trying to get work done in an organisation whose visible rules no longer explain where authority sits.

People learn the real organisation quickly

Imagine that two people both believe they own an important customer relationship. One has the relevant job title and manages the account day to day. The other joined the company earlier, helped win the customer and speaks regularly with the founder. When they disagree about a proposal, neither can point to an agreed decision owner.

At first, they try to resolve it directly. When that fails, each begins looking for support. They share different parts of the customer history with colleagues, test their argument in private conversations and try to understand which view the founder is likely to favour. Whoever reaches the founder at the right moment may acquire an advantage that has little to do with the quality of the proposal.

It is easy to describe this as personality conflict. Doing so misses what the organisation has asked both people to do. The customer matters, the decision affects their credibility and there is no dependable route to a conclusion. Building influence is a rational response.

The pattern becomes stronger when employees see that private influence works. A manager makes a decision on Monday. Someone who dislikes it mentions the issue to the founder on Tuesday, without the manager present. The founder, hearing only part of the context and believing the matter is still open, gives a different answer. By Wednesday, the original decision has been reversed.

Nobody needs to behave dishonestly. The founder thinks they have been helpful. The employee thinks they have protected the business. The manager learns that their authority lasts only until somebody finds a more powerful audience.

The next time that manager faces a difficult decision, they are likely to prepare politically. They may sound out the founder first, recruit support from influential colleagues or avoid making the call until consensus has formed. Other employees observe the same process and adapt. The official organisation may still say that the manager owns the decision, but everybody now understands the appeals system.

Ambiguity changes behaviour

Role descriptions are sometimes dismissed as corporate paperwork, especially by founders who value initiative. A detailed description of every task would be particularly unhelpful in a young company because the work will change faster than the document. But role clarity is not the same as task rigidity.

People need to know the result they are responsible for, the decisions they can make, the constraints they must respect and how their contribution will be judged. Without those boundaries, flexibility becomes difficult to distinguish from intrusion. Helping a colleague can be interpreted as taking over their area. Protecting a necessary standard can be interpreted as building an empire. Declining work that belongs elsewhere can be interpreted as lacking commitment.

Research on role ambiguity has found a consistent relationship with poorer performance. A meta-analysis drawing on 74 correlations and 11,698 people found that greater role ambiguity was associated with lower job performance, although the strength of the relationship varied by occupation and method. The research does not say that every startup needs narrow jobs. It suggests that uncertainty about expectations and responsibility carries a measurable cost.

Research into employees' perceptions of organisational politics finds an even wider pattern. A meta-analysis linked perceived politics with lower job satisfaction and organisational commitment, and with greater strain and intentions to leave. These are correlations across many kinds of organisation, so they cannot prove that unclear roles alone caused the outcomes. They do show why dismissing politics as harmless gossip is unwise. When employees believe that informal manoeuvring determines what happens, their relationship with the company changes.

The practical issue is predictability. People can cope with a decision they dislike more easily than with a system they cannot understand. If the rules seem to change according to who asks, when they ask and how close they are to the founder, every important decision becomes a test of personal influence.

The founder's open door can become a side entrance

Founder accessibility is one of the genuine advantages of a small company. Problems can reach somebody with authority before they are filtered through several layers of management. Employees can hear the commercial reasoning behind decisions, and founders remain close to customers and the work.

The answer is not to close the door. It is to stop using that door as a private route around accountable owners.

When an employee brings a complaint about someone else's decision, the founder first needs to establish whether they are being asked to listen, advise or overturn it. Those are different acts. Listening may reveal useful information without changing the decision. Advice can be offered to the owner. An intervention may be necessary if the decision creates serious legal, ethical, financial or strategic risk.

What damages the organisation is the casual reversal. The founder gives an opinion in a five-minute conversation, and the opinion acquires the force of an instruction because it came from the most powerful person in the company. The manager may not know that the conversation occurred until their team begins acting differently.

A disciplined founder can remain accessible while refusing to become a private court of appeal. They can ask whether the concern has been raised with the decision owner and, if intervention is necessary, bring that owner into the conversation. If a decision is overturned, the reasoning should be visible enough for the organisation to learn from it.

This will sometimes take longer than giving an immediate answer. The comparison, however, is not between a five-minute answer and a 30-minute conversation. It is between the 30-minute conversation and the weeks of hesitation, lobbying and repeated argument created when nobody trusts a decision to remain settled.

Medium discovered the cost of coordination

Buffer was not alone in trying to distribute authority without reverting to conventional management. Medium adopted Holacracy while it was still a young company. The system divided work into explicit roles and circles, with the intention of giving individuals genuine authority rather than making every decision travel upwards.

In 2016, Medium moved away from Holacracy. Andy Doyle, then head of operations, did not describe the experiment as a failure. He credited it with encouraging ownership, initiative and nimbleness. The difficulty appeared when important projects crossed several functions. Achieving alignment could become time-consuming and divisive, while maintaining the detailed system of roles created a substantial administrative burden.

Medium's next step is instructive. It did not conclude that the founder should decide everything. The company retained principles of distributed authority and individual accountability, but began developing clearer ways to assign ownership to strategic initiatives, map cross-functional work and decide without requiring universal agreement.

The case exposes a misconception about collaborative cultures. Giving everybody a voice does not require giving everybody a veto. People affected by a decision should have a route to contribute. Relevant specialists should be consulted. Disagreement should be safe. Eventually, one person or clearly defined group must have the authority to decide and the responsibility to explain the choice.

Without that final point, consensus can become another form of politics. People with greater status, confidence or spare time can keep a discussion open until others give way. Decisions emerge from endurance and coalition building while still being described as collective.

Google's Project Aristotle research identified structure and clarity as one of the conditions associated with effective teams. Google's own guidance describes this in terms of understanding expectations, the process for meeting them and the consequences of performance. One of its diagnostic questions asks whether a team has an effective decision-making process.

The finding is often discussed separately from psychological safety, another condition in Google's work. In practice, the two support one another. People are more able to challenge a proposal when they know who will decide, how their objection will be considered and what happens after the decision is made. A meeting without that clarity may contain a great deal of talking while very little honest disagreement occurs.

Map the organisation employees actually use

Founders usually know the organisation they intended to build. To understand its politics, they need to examine the one employees have learned to use.

Take five consequential decisions from the previous month. Avoid ceremonial board matters and choose ordinary decisions involving customers, priorities, spending, hiring or delivery. Reconstruct each one from the first question to the final action.

Who initially believed they owned it? Who held information the owner needed? Whose approval was formally required? Who was consulted privately? Who could delay the work without being accountable for the delay? Did the person who announced the decision actually make it? Was the choice later reopened, and if so, by whom?

The gaps between the expected route and the actual route reveal the shadow organisation. Perhaps the head of marketing owns the budget on paper but the founder approves every campaign. Perhaps an early employee holds no senior title but can stop operational changes. Perhaps a project lead is accountable for delivery but six people possess an informal veto.

These observations are more useful than asking whether the culture feels political. They identify where political skill has become necessary. They also prevent the founder from responding with a generic request for better communication. Employees may already be communicating constantly. The problem is that information, influence and authority are travelling through different channels.

Once those channels are visible, the company can clarify four things without writing a policy manual:

1. The outcome each role or team owns.
2. The decisions included in that ownership.
3. The people who must be consulted, and the limits of consultation.
4. The conditions that justify escalation or reopening a decision.

Important decisions should also leave a short record: what was decided, who decided it, the principal reasoning and when it will be reviewed. The purpose is not surveillance or bureaucracy. It is to stop the same decision being reconstructed differently in several private conversations.

The founder needs a place in this map too. Before commenting on work, they can say whether they are asking a question, offering advice or giving an instruction. That distinction only matters if employees are genuinely allowed to reject the advice. If choosing differently is punished later, every founder suggestion will correctly be treated as an order.

Pay attention to the exceptions

Culture is often described through values and regular processes. Employees also learn from the moments when those processes are ignored.

They notice whose missed deadline is accepted, whose mistake becomes a performance issue, which manager can be bypassed and which decisions remain closed only until the founder changes their mind. There may be a sound reason for every exception. When the reason is not explained, people infer that status or access made the difference.

This does not mean leaders must reveal private personal information or submit every judgement to a vote. It means exceptions should not quietly rewrite the operating rules. Where the company cannot explain a choice in detail, it can still acknowledge that an exception was made and restate the principle that will govern the next case.

Office politics will never disappear. People have different ambitions, loyalties and ideas, and organisations depend on persuasion as well as formal authority. The goal is not to remove influence. It is to prevent proximity, private appeals and coalition building from becoming more dependable than the legitimate way of making a decision.

Buffer's experiment did not prove that autonomy was a mistake. It showed that removing visible hierarchy did not remove the human need for context, development and stable accountability. Medium did not return every decision to its founder. It tried to preserve distributed authority while making cross-functional coordination more intelligible.

Founders face the same choice as their companies grow. They can leave employees to discover how power really works, or they can make ownership, consultation and escalation clear enough that people do not need to manage an invisible organisation before they can do their jobs.

 

Sources

1. First Round Review, Aaron Levie on how to scale 10x as a CEO. Retrospective interview with the Box co-founder covering the company’s early decision bottlenecks and the appointment of Dan Levin.

2. Journal of Industrial and Business Economics, Venture capital and the delegation of decision authority in startups, 2024. Mixed-method study including a survey of 241 startups.

3. Colin Bryar and Bill Carr, Working Backwards, hiring chapter excerpt. Insider account of Amazon’s early hiring practices and the development of its hiring mechanisms.

4. Amazon, How Amazon hires. Company account of the purpose and history of the Bar Raiser programme.

5. International Journal of Entrepreneurial Behavior and Research, Empowering growth: delegating decision-making authority during the evolution of new businesses, 2026. Survey-based research on trust, workload, ambition, venture risk and delegation.

6. Google Research, Lawrence Page biography. Company biography confirming Google’s approximate size and stage in 2001.

7. World Economic Forum, The habits of the best managers, according to Google’s former HR leader. Summary drawing on Laszlo Bock’s account of Google’s early experiment with removing engineering managers.

8. First Round Review, Questions and operating practices from Stripe’s COO. Interview with Claire Hughes Johnson covering Stripe’s design reviews, commercial decisions and transfer of founder judgement.

9. Academy of Management Proceedings, Delegation, attention allocation and innovation in entrepreneurial ventures, 2026. Preliminary research based on a representative survey of German entrepreneurs.

10. National Bureau of Economic Research, CEOs and firm performance, 2023. Research programme examining the working practices and time allocation of 1,114 chief executives across six countries.