The thinking

Three bodies of work.

One condition, one territory and one thesis. Each begins where the account an organisation holds of itself stops being accurate.

01 / ConditionThe Filter

An account is edited on its way to the top.

Reporting a problem early and in full carries a professional risk that softening it does not. The person softening it is behaving rationally. So is the person above them.

Across five reporting layers the softening compounds in one direction. What arrives at the top is coherent, current and wrong in the places that decide.

The judgement is sound. The picture is not.

01What goes

The unwelcome number, while it is still small

A figure carried early is a problem. The same figure carried late is a crisis somebody else can be held to. The incentive runs one way.

02What goes

The objection that was raised and lost

It is settled before the meeting, so the minute records agreement. The record is accurate about the vote and silent about the argument.

03What goes

The contradiction between two stated priorities

Each is defensible on its own. Naming the conflict requires naming whose priority yields, which is a decision the level below does not hold.

Three channels the hierarchy does not control.

A leader cannot repair the filter by asking harder questions through the same line. The line is the mechanism. What works is a route the line does not carry.

AChannel

Attribution removed

What is said cannot be traced to who said it. Non-attribution is a structural condition rather than a promise, and the transcript is what the leader receives.

BChannel

The line bypassed

The account arrives without passing the people it describes. Board packs, exit interviews and minutes are read as evidence, rather than as reporting.

CChannel

Consequence detached

Raising a condition costs the person raising it nothing. Restorative just culture is what keeps the channel open once the first uncomfortable finding lands.

Ask an organisation what is wrong and it will tell you what it has agreed to say is wrong.

02 / TerritoryFirst Generation Leaders

The filter is thickest where there is nothing to check it against.

A leader with an inherited network hears the unedited version at dinner. Somebody who has held the seat tells them what the report left out, at no professional cost to either.

A first-generation leader has one source: the organisation. Every account of it arrives through the structure the account is about.

The second opinion is a network effect. It is not evenly distributed.

Inherited

Two accounts

First generation

One account

The board pack, and a private reading of it
The board pack
A predecessor who has met this failure before
A first encounter, treated as a first encounter
A peer group outside the organisation
A peer group inside it, reporting upward
A calibrated sense of what a number should be
The number, and the person presenting it
Advice that carries no consequence for the adviser
Advice from people the decision affects

What the second channel replaces.

The work is not encouragement. It is building the check that an inherited network supplies for nothing, and doing it deliberately, from outside the reporting line.

Southern and East African infrastructure is led disproportionately by people in this position. That is a structural observation about who holds these seats, rather than a claim about capability.

Competence is rarely the constraint. Information is.

03 / ThesisPrecisely Wrong

An organisation that filters its own information is automating the filter.

Machine systems do not challenge a premise. They do not ask whether the record is complete, or notice that a category has been quietly redefined. They commit, at speed.

A model trained on a curated account of an organisation learns the curation. The output is confident, internally consistent and precisely wrong.

Speed is not the risk. Speed applied to an edited account is the risk.

01Exposure

The training record

Where bad news is carried late, the record of what happened is a record of what was reported. A system built on it inherits the delay as a fact.

02Exposure

Contradiction at scale

Where an executive team holds three readings of one strategy, automated decisions express all three. Human teams reconcile these in a corridor. Systems do not.

03Exposure

Reconstruction after the event

When an automated decision goes wrong, the board asks what the organisation believed. Human-factors investigation is built to answer that question. Most governance frameworks are not.

The governance question, put once.

Before an organisation scales a decision with machine intelligence, it has to establish that the account the system will act on describes the organisation as it operates.

That is an investigation, and it belongs before the approval rather than after the incident.

04 / Enquire

If one of these describes your week.

Thirty minutes, direct to me. Its purpose is to establish whether the problem you are describing is the problem you have.