For founders, owners, and operators

My Business Depends on One Key Employee. How Do I Reduce That Risk?

By Brendan Levin. 20+ years scaling businesses, from teams of 8 to 600 and budgets from startup to $100m.

Map every critical item that lives in one person's head, hands, or inbox using a single-point-of-failure register: what the item is, who holds it, who the named backup is, whether it is written down, and when the backup last did it for real. The register makes the invisible visible and gives you a repair list you can work through without alarming the person at the centre. Owner dependency works the same way.

Key takeaways

  • Key-person risk is most dangerous when the knowledge, relationship, or access is invisible to everyone except the person who holds it.
  • A single-point-of-failure register turns a vague anxiety into a prioritised list you can act on.
  • Building cover does not require telling the key person you are worried about them leaving: frame it as business continuity and team development.
  • The same register almost always reveals items that still sit with the founder, not just the key employee.

Why this feels different from owner dependency

Most writing on key-person risk focuses on the founder. But there is a quieter version that keeps growth-stage founders up at night: the one person on the team who is not you, but who holds something the business cannot run without. The operations lead who has the client relationship. The finance person who built the model everyone uses but nobody else understands. The senior technician who is the only one who can diagnose the thing that breaks on Friday afternoons.

The risk is not just that they might leave. It is that their presence gives you a false sense of safety. Everything runs smoothly, so you assume the business is robust. It is not. The knowledge, the relationship, or the access lives in one person. That is a single point of failure, and the signs are easy to miss until the person is gone.

The single-point-of-failure register

The most useful thing you can build is a register that makes every single point of failure visible on one page. Not a vague list of worries. A structured record with five columns for every critical item.

Here is what the register tracks. The critical item itself: what exactly would stop or break if this person were not available tomorrow. The category of that item: is it knowledge (how something works), a relationship (a client or supplier who deals only with them), system access (credentials or permissions only they hold), or approval authority (decisions that only they make in practice). Who holds it today. Who is the named backup for that item. Whether the item is written down anywhere the backup could find and use. And the date the backup last did it for real, not just shadowed it.

Those last two columns are where most registers fall apart. Something being documented is not the same as a backup being able to execute it under pressure. The only honest test is a real rehearsal.

Critical item (illustrative example)CategoryWho holds itNamed backupWritten down?Backup last did it
Monthly client review call with Account ARelationshipSara (Ops Lead)TomNoNever
Payroll sign-off and bank transfersApproval + AccessMike (Finance)OwnerPartial6 months ago
Server access and deployment credentialsSystem accessDev leadNo one namedNoNever
Supplier negotiation and pricing authorityKnowledge + RelationshipSara (Ops Lead)OwnerNoNever

How to build cover without signalling distrust

The most common reason founders do not act on this is that they worry the key person will read it as a vote of no confidence, or as preparation to replace them. That concern is understandable but it is also fixable, because the framing matters more than the action.

The honest frame is business continuity, not succession planning. Tell the person directly: if you were hit by a bus, or took a well-earned four weeks off, I want the business to be able to keep operating. That is a compliment to how central they are, not a threat. Most capable people hear it that way. The ones who push back hard on any knowledge-sharing are usually signalling something worth knowing earlier rather than later.

The practical approach is to build the register together with them, not around them. Ask them to name the things only they know how to do. They will almost always produce a longer list than you expected. Then work through the list item by item: what would it take for a backup to be able to do this without you? Document it, run a real rehearsal, update the date. Repeat quarterly. Delegation works on the same logic: the transfer is only complete when the other person can do it without you in the room.

The register will show you what still sits with the founder

Do not be surprised when you work through this and find that several of the items in the register point back to you. The client relationship that runs through the key employee on day-to-day calls, but where the key commercial decisions still route back to you. The approval that sits with them in theory but that they quietly escalate to you for anything above a small threshold. The knowledge that is documented, but that you are the only one who knows where the document lives.

This is not embarrassing. It is useful. The register gives you two repair lists at once: the items concentrated in your key employee, and the items concentrated in you. Both are single points of failure. Both go on the same register.

If every real decision still routes back to you, the register will show it clearly. That is a structure problem, and it does not go away by fixing the key-person risk alone.

What to do when the register reveals a gap you cannot quickly fill

Some items on the register will have an easy fix: write the process down, run a rehearsal, name a backup. Others will expose a deeper gap. The client relationship that only works because of ten years of personal history. The technical knowledge that would take eighteen months to genuinely transfer. The approval authority that sits with your finance person because no one else in the business understands the numbers well enough to hold it.

For those items, the register is still doing its job. It is telling you where the structural investment needs to go: a documentation project, a deliberate knowledge-transfer plan, or a decision about whether that item needs a second person who genuinely understands it. None of those are fast fixes, which is exactly why the register is more valuable the earlier you build it. Taking a real holiday is often the first honest test of whether the gaps are real.

Structure makes the register stick

A register built once and filed away is not worth the hour it took. The backup column only means something if the backup last did it recently. That requires a cadence: a quarterly review of the register, run as part of your normal operating rhythm, not as a one-off project.

Most founder-led businesses between eight and fifty people have not built that kind of operating rhythm yet. Building one is not a large project, but it does require deciding that this is how the business runs, not just how you plan to run it one day. The Momentum Engine is the program that installs that structure into the business in twelve weeks, so the register, the cadence, and the decision-making all run without routing back to you.

If the register shows that most of the real risk still sits with you, the Momentum Engine is the twelve-week program that fixes the structure underneath it.

See the Momentum Engine

Common questions

What if my key employee is also a close friend or long-term team member?

The register and the conversation are still the right move. Frame it as protecting them too: if they are ever sick, on leave, or simply want to step back from certain responsibilities, the business should not grind to a halt in a way that forces them back. That is a genuine benefit to them, not just to you. The discomfort of the conversation is almost always smaller than the disruption of being unprepared.

How do I know which items on the register are actually high-risk?

Prioritise by two factors: how fast the business would break without this item, and how long it would take a backup to be genuinely ready to handle it. A client relationship that would erode over three months if the key person left is lower urgency than a system access credential that, if lost, stops the business tomorrow. Start with the items where the answer to both questions is bad.

Is this the same as a succession plan?

Not quite. A succession plan is about who replaces a person. The single-point-of-failure register is about which specific items are at risk and whether a backup can handle each one. A person can be backed up across multiple items without any single person being their successor. The register is more surgical and more immediately useful for a growing business that is not thinking about exit.

What if my key employee is the only one who understands our clients deeply enough to hold those relationships?

That is common and it is a genuine risk. The fix is not replacing the relationship but building a second thread into the same client: another person from your business who has a real relationship with a real contact at the account, not just a name they have heard of. That takes time and deliberate effort, but a client relationship that runs through exactly one person on your side is fragile for both parties.

Do I need to tell the key employee I am doing this?

Yes, and it goes better when you do. Doing it quietly tends to feel like surveillance when it is discovered, and it usually is discovered. Doing it openly, framed as a business continuity exercise that benefits everyone including them, lands very differently. Build the register with them. They will know more about what is at risk than you do.

How often should I review the register?

Quarterly is the right cadence for most businesses in the eight-to-fifty-person range. The critical column is the last date the backup actually did the thing for real. If that date is more than six months ago, treat it as if there is no backup at all. Skills and relationships erode when they are not used.