How to Audit Delegated Work

What Is the Problem?

Unaudited delegation has three familiar accidents.

Silent decay

The delegated task runs flawlessly for its first weeks and watching slackens; then the input pattern shifts — a new product, a new client type, a changed format — and the output starts spoiling quietly. Nobody reads, so nobody sees; spoiled output ships for weeks. When it surfaces, the work is recalled piece by piece and every hour the delegation saved is repaid in one remediation week.

Faults pass for one-offs

The occasional caught error gets a shrug — “these things happen” — and single-case treatment. Without a log, nobody sees that three errors share one root; a pattern is visible only in records. The unlogged line pays for the same lesson as if it were new, every time.

The stopping hand is undefined

When the work goes wrong, “should we stop it” floats: whose call, by what measure, and how does the work run while stopped? Forbes’ analysis of stalling deployments places the same gap in the typical picture: no human defined with the authority to stop the system. A line without that authority argues first and stops second in a crisis — and the hours in between are already invoiced.

Why Does It Happen?

The audit gap has three well-meaning roots.

Delegation is mistaken for release

“We’re rid of that job” closes the attention along with the handover. But what was delegated is the doing of the work, not its accountability — the output still carries the business’s signature, and a signature demands maintenance. The feeling of release is the silent killer of the audit budget.

The honeymoon gets mistaken for the baseline

Delegation typically starts with the most suitable task and the cleanest inputs; the opening period shines. An audit not built during the shine cannot be built later — the healthy baseline it needs was never measured. Oversight is installed on the line’s best day and searched for on its worst.

Audit cost never gets netted against the gain

When sampling and log hours are absent from the delegation arithmetic, the audit reads as “overhead” and gets pruned in the first busy stretch. The honest formula is set upfront: delegation’s net gain is what remains after the audit share — a share that is small but never zero.

How Is It Done?

The routine has four parts; all fit one page per task.

Part 1: the sampling schedule

Not all output is read — a regular slice is: a high rate at the start (say one in ten), thinning on a written schedule as the line earns trust. The sample is drawn at random — never just the first outputs or the easy ones — and is never skipped on busy days, because decay begins on busy days. Reading follows the one-question discipline of the approval flow: what to look at is written down.

Part 2: the three-class fault log

Every caught fault enters the log as one line and takes one of three classes: cosmetic (style, format — does not stop dispatch), material (wrong information, missing item — ships only after correction), critical (touches a client, the ledger or the law — halts the line). Classes are defined in advance; the log is scanned weekly and a recurring root gets written into the task definition.

Parts 3-4: thresholds, alarms and the stopping authority

Each class gets a threshold: cosmetic share above this band triggers a template review, material faults above that frequency tighten the sampling, a single critical fault stops the line. Thresholds must produce alarms — not dashboard décor — and the alarm’s address is fixed: the task’s owner. A threshold without an alarm is a thermometer nobody reads. The routine’s last stone is authority: the power to stop the line rests in one name and in a crisis is used, not searched for. Two more lines sit beside it — how the work runs on the human side while stopped, and on what condition the line reopens. The business that can stop, delegates bravely; the right to accelerate belongs to whoever has working brakes.

How Long, at What Cost?

The audit’s economics read directly off the delegation’s.

Half a day to build, one hour a week to run

Writing the four parts is half a day per task; running them — reading samples, scanning the log — settles at about an hour weekly on a mature line. That hour is a small percentage of what the delegation recovers, and it works like an insurance premium: skip a month and nothing happens, skip a year and something does.

The cost measure doubles as the decision measure

If audit hours approach the delegation’s gain — if verifying the output takes as long as the work itself — the task sits in the wrong tier: it returns to the test or steps down to the approval flow. Audit cost is the delegation decision’s permanent referee.

The log accumulates a second fortune

Over time the fault log becomes a business-specific treasury: which input type produces which fault, which class rises under which setting. That treasury is the most reliable data behind the next delegation decision — an industry report describes the average; your own log describes you.

The Common Mistake

The audit gets built; it rots in three places.

Thinning the sample on success

“No faults for months, let’s stop looking” is cancelling the insurance because the year was claim-free. The thinning schedule is written in advance and never undercut; a zero-fault period proves the audit is working, not that it is unnecessary.

Attaching the log to a culprit

The day fault records feed personal performance reviews, the records turn pretty and the truth disappears. The log measures the line, not the person — until that principle is on the wall, no honest data gets collected.

Counting a stop as a defeat

Halting the line on a critical fault is not the delegation’s bankruptcy but the audit’s victory. A line stopped, root-caused and reopened comes back stronger; a line afraid to stop grows a small accident into a public one.

Frequently Asked Questions