Five Gates, Two Checks, and One Maria

On paper, the contract approval process was excellent.

Five approval gates. Two compliance checks. One final sign-off.

The book describes it in three words: elegant, clean, logical. I have seen a hundred process maps that deserved the same description, and I have drawn some of them myself.

Then, in Chapter Seven, David walks into the cramped office where that process actually lived.

A junior manager named Maria is hunched over a stack of documents, physical and digital. The book describes her as someone living in a constant state of quiet vigilance — eyes moving between monitors, shoulders raised as if braced for impact.

He makes a light remark about her being the nerve centre of the operation.

She answers: “I'm just… making sure nothing slips through.”

The Sentence That Should Stop a Meeting

I want to sit on that sentence, because it is the whole article.

The process did not ensure nothing slipped through. Maria did.

The five gates and the two checks were a diagram. She was the mechanism. And nowhere in the documentation of that elegant process does the word Maria appear.

There is a body of research about exactly this, and it comes from safety science rather than management, which is probably why most of us have never encountered it.

Erik Hollnagel draws a distinction between work-as-imagined and work-as-done. Work-as-imagined is what designers, managers and regulators believe happens, or believe should happen. Work-as-done is what actually happens.

His central claim is the uncomfortable one. Things do not go well because people follow the procedures. Things go well because people continually adjust to conditions the procedure did not anticipate.

Read that again if you have ever run an operation. Performance variability is not the enemy of the system. In complex work it is the reason the system functions at all.

Which means that when a process looks elegant on paper and works in practice, you should not conclude that the design was good. You should go and find out who is absorbing the difference.

Every process of any age has a Maria. The question is only whether you know her name.

The Fifty Thousand That Was Not the Finding

There is a second thing in that conversation, and it is easy to read as the main event.

Maria mentions, almost in a whisper, that there is a loophole in vendor renewals. It is costing the company more than it should — around fifty thousand a year.

David asks whether she has raised it.

“I did. Twice.” And then: her director told her it was not her place to question legal frameworks.

The book's line about this is exactly right — it was not malice. It was a system that rewarded silence over curiosity.

Now, most leaders reading that will fix on the fifty thousand. It is concrete, it is embarrassing, and it can be recovered.

I think the fifty thousand is the least interesting number in the story.

Because that figure came from one person, who tried twice, and only mentioned it on the third occasion because someone senior happened to sit down beside her and ask a question that sounded genuine.

It is not a finding. It is a sample.

And a sample implies a population. What is the rate of unreported fifty-thousands across everybody who tried once and stopped? Across everybody who watched Maria get told it wasn't her place and quietly drew the obvious conclusion?

You cannot know. That is rather the point. The only thing the number tells you reliably is that your detection system is a junior manager's willingness to whisper.

Amit, and Why Resignation Is Accurate

The manufacturing floor is worse, and quieter.

It looks efficient. Machinery humming, yellow lines, technicians who know their craft. But underneath it David senses something he names precisely: not anger, but resignation. A practiced disappointment.

A veteran technician, Amit, takes him aside. There are workarounds the floor has used for years. They could save thousands of hours if engineering approved them.

So why hasn't anyone implemented them?

“Because we've stopped suggesting things. Nobody listens until something breaks.”

And here is the detail that I think matters most in the entire chapter. The book notes that he did not say it with bitterness. He said it with the exhaustion of someone who had learned where the limits were drawn — and how small they were.

Bitterness would have been better news. Bitterness means somebody still expects something.

What Amit demonstrated was not disengagement. It was accurate learning. He had run the experiment repeatedly, observed the result, and updated. Any competent engineer would have done the same.

Which is why I have stopped treating resignation as a morale problem. Morale problems imply a mood. This is a conclusion — and it is usually a correct one, drawn from evidence the organisation supplied itself.

You do not fix a correct conclusion with a survey. You fix it by changing the evidence, repeatedly, for long enough that somebody risks a suggestion and watches what happens to it.

Why the Corner Office Genuinely Cannot See This

Hollnagel frames the gap explicitly as a hierarchy problem: work-as-imagined belongs to the blunt end of an organisation, work-as-done to the sharp end.

I would put it more plainly. Nobody in this story is lying.

The process document is accurate. There really are five gates and two compliance checks. The efficiency figures are real. Every report reaching the senior floor is true.

They are simply describing work-as-imagined, and there is no field on the form for Maria.

That is why this cannot be solved by asking for better reporting. Better reporting produces a more detailed description of the same imagined thing. The information you need does not exist in written form anywhere in the building — it exists as behaviour, in rooms you do not enter, performed by people who have concluded that mentioning it is not their place.

What a Guerrilla Audit Actually Is

David's method, which he called guerrilla audits half as a joke, is stripped of everything that would normally make it official.

No announcement. No calendar invitation. No leadership copied in. He walked floors he rarely visited, joined stand-ups uninvited, sat beside people, and asked questions that carried — in the book's phrase — more listening than agenda.

The informality is not style. It is the method.

You cannot schedule work-as-done. The moment a visit is announced, what you are shown is a performance of work-as-imagined, prepared conscientiously by people who assume that is what you came to inspect. They are not deceiving you. They are being helpful.

One caution, because this technique is easy to corrupt. Walking the floor unannounced can just as easily be surveillance, and people can tell the difference within about ninety seconds.

The distinguishing test is not how you behave during the visit. It is what happens to what you learn. If Maria's loophole gets fixed and her director is quietly told why, the next person talks. If it disappears into a report, you have taught the floor one more time that speaking up is a cost with no return — and you will have made things measurably worse than if you had never come.

What to Do on Monday

Pick one process you believe is working well. Not a troubled one — a good one, that reports clean numbers.

Go and sit where it actually happens, without telling anyone you are coming, and ask one question:

What do you do when this doesn't work?

Then stay quiet for considerably longer than is comfortable.

Whatever comes after that silence is your real process documentation.