Why Most Teams Call for Help Too Late.

Written and narrated by Serguei Poppeleer.

By the time a programme calls me, it’s rarely a question anymore. It’s a claim. A deadline. A number with a lawyer attached.

The same work would have been a tenth of the size three months earlier. It almost always would have.

Teams don’t call late because they’re careless. They call late because the system told them everything was fine — right up until the moment it very obviously wasn’t. They were watching the dashboard. The dashboard was green.

The cheapest moment to fix a programme is always earlier than it feels.

The signals were there the whole time. A transformer delivery that kept “confirming” but never quite landed. A fire-suppression package that kept getting deferred. A cable-routing decision that stayed open so long everyone forgot it was still open. None of those trip an alarm. That is precisely the point.

Contracks LawThe alarm is calibrated to the crisis, not the cause.

I have watched this same shape repeat across transport hubs, power plants, pipelines and government programmes. Different sectors, different countries, identical sequence: quiet, quiet, quiet — then a number with a lawyer attached.


Why the alarm is always late

Programmes are wired to detect crises, not drift. The escalation thresholds are set where the damage is visible: a missed milestone, a blown budget line, a formal notice. By the time a signal is loud enough to clear that bar, the cheap window has closed. You are no longer preventing the problem. You are managing its consequences.

Drift lives below the threshold by design. Each individual slip is survivable. A week here. An unpriced change there. A decision parked “until the next meeting.” Stacked over a quarter, they are a claim. But no single one was ever loud enough to make someone pick up the phone.

Equatorial Guinea — a remote camp build

What happened: built at a fraction of the expected cost, in 28 months against a 36-month standard, with zero claims.

Why it could have gone the other way: remote terrain, no established supply chain, every standard assumption pointing to 36 months and a fight at the end.

Why it didn’t: the small signals got acted on while they were still small, by people close enough to the work to see them. Decisions closed in days, not weeks. The state client released payment on a single note: “We have started the works.” That outcome was not luck. It was attention applied early, before anything could harden.

Early did not mean dramatic. It meant visible, owned, and closed before it became formal.

The distance between reporting and control is where claims are born.


What “early” actually looks like

1
A date confirms but never quite lands.

That is the signal. Not the missed date — the pattern of soft confirmations before it. Act on the pattern and you never reach the miss.

2
A decision keeps getting deferred.

Deferral is not neutral. Every week a decision sits open, the options narrow and the cost of the eventual call rises. The deferral itself is the warning.

3
A scope conversation keeps getting postponed.

Unpriced scope does not disappear because nobody discussed it. It compounds quietly until it surfaces as a claim. The postponement is the signal.


The objections this raises

“We can’t escalate every small thing — we’d cry wolf.”

Acting early is not escalating loudly. It is closing a decision, verifying a date, pricing a change — quiet, cheap moves while they are still cheap. The wolf only needs crying about once you’ve waited for the crisis.

“Bringing someone in early looks like we can’t handle it.”

Bringing someone in at the claim stage is what looks like you couldn’t handle it. Early help is cheap, quiet and invisible to the board. Late help has a number attached and a name in the minutes.

AI Reality Check

This is exactly the gap AI is built to close. Humans cannot watch every soft confirmation, every parked decision, every deferred scope item, every week. AI can — surfacing the small signal while it is still small, so a person can make the call early.

But it only works if the things worth watching are defined first. AI accelerates a discipline; it cannot invent one. Decide what “early” looks like, then let the machine watch for it.

AI can tell you it’s early. It cannot make you act while acting is still cheap.


The line worth keeping

Frontier markets don’t change the rule. They just make “too late” arrive faster.

You don’t need help when it’s a crisis.

You need it while it still looks fine.

The call you don’t want to make is the one that’s still cheap to make.

Catch it while it still looks fine.

The Delivery Score shows where your programme is drifting below the alarm line — while there’s still time to act cheaply. Fifteen Yes/No questions. About four minutes. Instant result.

Get my Delivery Score — free → Read the core governance report →