Picture the resignation landing on a Tuesday. Not a bad leaver: plenty of notice, a proper handover offered, no hard feelings. The problem is what they actually do. For the last four years they have owned the month-end reconciliation, the new-client onboarding sequence and the settlement checklist. None of it is written down in a way anyone else can follow. It lives in their head, a spreadsheet on their desktop, and a set of habits nobody else has watched closely.
Most firms have at least one person like this. The work runs smoothly right up until it doesn't. This article looks at how that dependency forms, what it costs while the person is still there, what it costs when they leave, and five ways to reduce it that do not require a documentation project the team abandons by week three.
Key takeaways
- A process that depends on one person is a continuity risk, not a personality quirk. When they take leave, throughput drops. When they resign, it can stop.
- The dependency builds up without anyone deciding it should. In a 2018 study of about 1,000 workers, 42% of workplace knowledge sat with a single person and was not shared with colleagues (Panopto study, via HR Dive, 2018).
- The cost lands before anyone leaves: no cover for annual leave, quiet bottlenecks, an owner who cannot step back. In research reported in January 2026, half of Australian small-business owners were reluctant to take time off because of the impact on the business (AMP Bank GO, 2026).
- You can reduce it without a big documentation effort: capture the process while it runs, give the task a second pair of hands, automate the mechanical steps, and test the result with a planned absence.
What "one person is the process" actually looks like
It is rarely a formal role. It is a pattern you notice in passing: one name comes up every time a particular piece of work needs to move.
The common signs, across accounting firms, law firms, advisory practices and property businesses:
- A recurring process (month-end close, trust-account settlement, client onboarding, payroll, board-pack preparation) that only one person can complete end to end.
- A personal spreadsheet, email folder or notebook that is the real system of record, sitting alongside whatever software the firm pays for.
- Steps that are never questioned because only one person knows why they exist.
- A visible drop in throughput whenever that person is on leave, and a quiet reluctance in the team to book around them.
None of this shows up on an org chart. It shows up in how nervous people get when that person mentions a holiday.
The reframe worth making early: this is not a documentation failure by one lazy manager. It is the default outcome of how work accumulates. Knowledge that is used every day feels obvious to the person using it, so it never gets written down. The absence of a record is the natural state. A shared, current record is the thing that has to be deliberately built.
Why capable firms drift into it
A firm ends up here by doing sensible things. Someone competent takes on a fiddly process. They get good at it. Because they are reliable, more of that kind of work flows to them. Each individual decision is reasonable. The cumulative result is a single point of failure that nobody chose.
Tacit knowledge is sticky. In Panopto's 2018 study of around 1,000 workers, 42% of institutional knowledge was unique to the individual who held it and was not shared with any colleague. The same study found employees losing about 5.3 hours a week waiting on information from co-workers or rebuilding something that already existed elsewhere (Panopto study, via HR Dive, retrieved 31 August 2026). Those numbers are from a US sample and are a few years old, but the pattern is familiar to anyone who has run a services team: the knowledge that matters most is the knowledge that never leaves someone's head.
For an Adelaide firm running five to ten tools that were bought separately over the years, there is an extra push toward single-person processes. When the software does not join up, someone has to be the join. That person becomes the integration layer: they know which report to pull from which system, in what order, and how to reconcile the two when they disagree. The more disconnected the stack, the more valuable and less replaceable that person becomes.
What it costs before anyone resigns
The dependency is expensive while the person is still sitting at their desk. The costs are just easy to miss because nothing has broken yet.
Leave becomes a problem to manage. Annual leave, sick days and long service leave all turn into throughput dips. Teams learn to schedule around one person's calendar, which slows everything tied to that process.
The owner cannot step back. In small firms the single-person process is often the owner's. Research reported in January 2026, based on a survey of 2,000 Australian sole traders and small businesses with 20 or fewer staff, found that one in four owners felt they had no time to take a holiday, half were reluctant to take leave at all because of the impact on the business, and 90% were preoccupied with work outside normal hours (AMP Bank GO, retrieved 31 August 2026). A business the owner cannot leave for two weeks is carrying a cost every one of those weeks.
Improvement stalls. A process only one person understands is a process only one person can improve. Everyone else works around it. The steps that no longer make sense stay in place because the cost of questioning them is a long conversation with the one person who knows.
Quality depends on a mood. When there is no written standard, "done properly" is whatever that person does on a good day. On a bad week, or a busy one, the output drifts and nobody has a reference to check it against.
In practice: the single-person process usually is not the slowest or most visible part of the operation. It is a quiet, reliable bottleneck that the firm has organised itself around so thoroughly that the cost has become invisible.
What it costs when they do resign
The bill that has been accruing quietly now arrives in full.
Replacing the person is not the same as replacing the knowledge. Widely cited Gallup analysis puts the cost of replacing an employee at one-half to two times their annual salary, and estimates that voluntary turnover costs US businesses around a trillion dollars a year (Gallup, 2019). Those figures are US and directional, but the shape holds in Australia: recruitment fees, a vacancy, a ramp-up period, and senior people pulled off billable work to cover and to train.
The hiring market may not cooperate. For Australian firms in finance and accounting, the replacement is not guaranteed to be waiting. A Chartered Accountants ANZ survey of 395 members who advertised roles across 2024 found national vacancy fill rates of 59% for taxation accountants, 45% for management accountants and 37% for finance managers, all below the 67% mark that signals an occupation shortage (CA ANZ survey, via Accountants Daily, 25 March 2025). If the one person who runs your close resigns, the plan that assumes a fast like-for-like hire is optimistic.
Smaller firms feel it hardest. The Australian Legal Practice Management Association's 2025 survey of 345 firms found staff turnover had eased to 23%, the lowest in three years, but noted that "retention remains a challenge, particularly among smaller firms" (ALPMA, May 2025). A 12-person practice that loses its one operations lead has lost a larger share of its operating knowledge than a 200-person firm losing the same role.
It shows up when you sell. Buyers and acquirers of professional-services businesses look hard for key-person dependency, and they price it in. A practice that runs on documented, repeatable processes is worth more than an identical one that runs on three people's memories, because the buyer is purchasing something that keeps working after the handshake.
How do you de-risk it without a documentation project that stalls?
The usual response is to announce a documentation initiative. Templates go out, a few processes get written up in the first fortnight, and then real work resumes and the effort dies. The fix is smaller and more specific than a firm-wide programme.
Capture the process while it runs, not from memory
Do not ask the person to sit in a room and write down what they do. They will miss the steps that feel automatic, which are exactly the ones that trip up a replacement. Instead, record the next real run: screen recording, a shared doc filled in live, or someone sitting alongside taking notes as it happens. Capture the decisions, not just the clicks. "If the two totals do not match, I check the bank feed first" is the part worth keeping.
Make the checklist the system, not a document about the system
A checklist in a Word file is a description. A checklist built into the tool the team already opens every day is the process. Move the steps into your practice-management system, your project tool or a shared task template, so completing the work and recording how it was done are the same action. The record stays current because it is not optional.
Give the task a second pair of hands before you need them
Full cross-training of every role is unrealistic for a firm of 15 people. Pick the two or three processes where a single-week outage would actually hurt, and make sure one other person has run each of them, start to finish, at least once, recently. Not read the doc. Run it. A backup who has done the task once under supervision is worth more than a 20-page procedure nobody has tested.
Automate the mechanical steps so there is less to hold
Part of what the single person carries is tedium: exporting the same four reports, reformatting them, copying figures between systems that do not talk. Those steps do not need a human and they do not need documenting if they are automated. Automating the mechanical parts of a process shrinks it to the judgement calls, which are easier to teach and less fragile to hand over.
The point of automation here is not speed. It is reducing how much undocumented knowledge one person has to hold in the first place. A shorter process is a more transferable one.
Test it with a planned absence
Book the person off for a week. Not a trap, and not a surprise: tell them it is a resilience test, and have them available for genuine emergencies only. See what stalls, what questions come up, and what the stand-in cannot find. The gaps you hit in that week are your actual documentation backlog, ranked by real impact. Everything else can wait.
Where OrionX fits
A firm rarely has one single-person process. It has four or five, tangled together, usually where the tools fail to connect. OrionX runs a stack and process audit that maps them: which processes depend on one person, which steps are mechanical and can be automated, and where the manual joins between systems are creating the dependency in the first place. You get a ranked list of what to fix and what it would take, with no obligation to have us do the work.
Book a free consultation and we will walk your five to ten tools and show you where the knowledge is trapped.
Frequently asked questions
Isn't this just a matter of writing better SOPs?
Written procedures help, but on their own they tend to go stale the week after they are written, because keeping them current is a separate task that competes with billable work. The more durable fixes move the process into a system the team already uses, cut the process down through automation, and make sure a second person has actually run it. Documentation is one part of that, not the whole answer.
We're too small to have a backup for every role. Where do we start?
You do not need a backup for every role. List the processes where a one-week absence would cause real damage: missed client deadlines, a blown close, a compliance breach. That list is usually short, three to five items. Make sure one other person has run each of those, recently and end to end. Leave the rest until the business is bigger.
Should we document the process or automate it?
Automate the mechanical, repetitive steps: exports, reformatting, moving data between systems. Document the judgement: why a step exists, what to check, what to do when something does not reconcile. Most single-person processes are a mix of both, and splitting them this way makes the whole thing smaller and easier to hand over.
The one test worth running this month
If reading this made a specific name come to mind, that is the answer to the hook. You already know who the process is.
The single most useful thing you can do in the next month is the planned-absence test: pick the one process that worries you most, book its owner off for a week, and watch what happens. It costs you a week of mild inconvenience now instead of a quarter spent scrambling later, and it turns a vague worry into a concrete, ranked list of what to fix.
