Governing Risk · Part 1 of 8 · Richard Anderson · 17 November 2026

Who is judging?

Four ways of seeing the same decision

Listen to this articleRunning time 20 min

Download MP39.5 MB, to listen offlineRead by Richard Anderson

Or follow Talking Risk in your podcast app, and new audio editions arrive by themselves:
Apple Podcasts Spotify

St Paul's Cathedral seen through a blur of people crossing the Millennium Bridge, with the words Governing Risk, part 1 of 8: Who is judging?
St Paul’s Cathedral from the Millennium Bridge. Photograph by Richard Anderson

This is the first of eight essays in Governing Risk, published every other Tuesday until March, with a break at Christmas. Each takes an idea I have used with boards and on my courses over thirty years and asks what it means in a world knocked off its axis: how boards see risk, how they judge it and how they govern it. Some of the ideas are old, and some I have changed my mind about. All eight are free to read and to pass on, and I would welcome challenge as much as agreement. I start with the people in the room, because every other question about risk depends on who is judging it.


A board is asked to let an AI assistant act on behalf of customers. One director wants a small trial so the firm can find out what customers actually do with it. Another wants to know what the assistant will be forbidden to do. A third asks who will bear the loss when the assistant makes a mistake. A fourth wants to know whether the firm can switch it off if the supplier changes the underlying model without warning.

They have all read the same paper. Each thinks they are asking the question a prudent director ought to ask.

Later in this series, in “Dimensional Control, twenty years on”, I ask how we decide what kind of control a risk needs. Where the evidence is thin, there is a prior question: what are we each assuming about the world when we make that judgement? Those assumptions are rarely in the board paper, yet the decision often turns on them.

Two questions, four views

The anthropologist Mary Douglas, and later Michael Thompson and others, developed what became known as cultural theory: an account of how the way we live together shapes what we notice and what we fear. John Adams drew on it to show why people can look at the same risk and reach very different conclusions about what to do.

In the boardroom I reduce it to two plain questions, and I ask them before I explain anything.

  1. Is human nature individualist or collectivist? I or C?
  2. How do we live? Do “superiors” tell “inferiors” the rules, or do equals negotiate them? T or N?

Directors answer I or C, and T or N, for how they instinctively think rather than how they think they ought to. I take the answers without names and put them on the grid. Only then do I say what the four boxes mean. Describe the types first and everyone places themselves in the most flattering one. Ask cold and people are often surprised where they land, and more surprised where their colleagues land.

Try it yourself before you read on. Note down I or C, and T or N. Then look at the grid.

Four ways of seeing risk, after John Adams and his myths of nature

T · Tell “superiors” tell “inferiors” the rules

T and I

Fatalist

Nature capricious

A ball on a flat plain could roll anywhere. What will be will be.

T and C

Hierarchist

Nature perverse/tolerant

A ball in a valley is safe within limits. Push it too far and it is over the hill.

N · Negotiate equals negotiate the rules

N and I

Individualist

Nature benign

A ball in a bowl rolls back when pushed. The world is forgiving, so take the chance.

N and C

Egalitarian

Nature ephemeral

A ball on a dome falls at the slightest push. The world is fragile, so look after it together.

The top row hides an important difference. A hierarchist wants rules because they believe authority and expertise can keep a system within safe limits. A fatalist experiences rules as something other people impose, with little say over the outcome and little expectation that anyone will help them deal with it. From the boardroom both may look like compliance. On the ground they feel very different.

Four ways of seeing the AI decision

Adams gave each view a “myth of nature”: a ball in a landscape that shows how each expects the world to behave when pushed. I take them in the order I use in the room, starting bottom left and working anti-clockwise, with the fatalist last.

The individualist

The ball sits at the bottom of a bowl. Push it and it rolls back. The individualist sees a world with room to try things and recover from mistakes. In the AI discussion, this is the director who wants a bounded trial. Pick actions the firm can reverse. Give a small group of customers access. Find out where the assistant helps and where they would rather deal with a person. The individualist sees a cost in waiting while competitors learn and the firm’s own assumptions go untested.

That is a necessary voice on a board that is quick to see the risks of doing something and slow to see the risks of standing still. Its weakness is the assumption that mistakes are recoverable. A transfer may be reversible in the firm’s systems while the customer has already missed a payment. Before agreeing to the trial, I would ask what can be put right, by whom, and in time for whom.

The egalitarian

The ball is balanced on top of a dome. The slightest push and it falls, perhaps for good. The egalitarian sees a world in which apparently small changes can do lasting damage, and asks how the decision affects people who have little say in it. The average customer might save a few minutes. A customer who struggles to reach a human being may be worse off. Who gets the benefit, who takes the risk, and who was involved in deciding that the trade was acceptable?

This view brings the experience of affected people into a discussion that can otherwise be dominated by averages. It might lead the board to require a clear route to a person and to look separately at the outcomes for customers who use it. Its weakness is that, in being careful, it can lose sight of the cost of the service people have today. Some customers may already spend half an hour getting help with a simple task. Their experience belongs in the comparison too.

The hierarchist

The ball rests in a valley. It is safe within limits, but push it too far and it goes over the hill. The hierarchist expects the system to work within limits that competent people can establish. Who sets the assistant’s permissions? Who tests that they hold? Who is accountable when something changes? A hierarchist wants answers before the firm gives the assistant authority to act. Regulated finance is hierarchist by design, so this is a familiar and often productive instinct in our boardrooms.

The danger comes when a limit that is easy to test stands in for a risk that is hard to understand. A firm can establish that an assistant cannot make a transfer above a set amount. That tells the board much less about what might happen if thousands of assistants respond to the same signal at once. The hierarchist’s discipline is valuable here too, provided someone asks what the test establishes and where its conclusions stop.

The fatalist

The ball sits on a flat plain and could roll anywhere. The fatalist sees a world in which important events are outside their control. Few directors volunteer for this label. It may be easier to hear the view from an operations manager who is told to own the outcome of a supplier decision they cannot influence. They have seen assurances given, procedures written and exceptions passed to the front line. They know who will be left to cope.

In the AI discussion, this voice asks what happens when the firm cannot get an answer from the supplier, or when the people who can suspend the service are unavailable. A board should take that question seriously. A stop button is of little use if the person watching the problem has to wait until morning for permission to press it. Fatalism can also lead people to give up on changes they could make. The board’s task is to find out which constraints are real and to give people authority where it would make a difference.

When nobody knows

Put a different problem in front of the same board. A colleague spots that a customer’s explanation and paperwork do not add up. The directors may have different reasons for saying so, but I would expect broad agreement that the colleague should pause and get help. They can all see enough of the problem to agree on what to do next.

Adams split risks into three kinds, and I put them more simply. Some risks everyone knows: you can see them, and experience tells you what to do. Some risks some people know: you need science, data or specialist expertise to see them. And some risks nobody knows. Adams called these virtual risks. The suspicious paperwork belongs to the first kind. The AI assistant is less tidy. Its transaction limit can be tested, which puts that part in the second kind. The virtual risk is what happens when many assistants, customers and suppliers interact in ways the firm has little basis for predicting.

Jonathan Hall, speaking personally as an outgoing member of the Bank of England’s Financial Policy Committee, asked in November whether AI will make the financial system more concentrated and more correlated, especially under stress. His answer: “at this point we don’t know, but there is a risk that it will.” Part of that is already visible. In the Bank and FCA’s survey, three cloud providers accounted for 79% of firms’ cloud relationships, and mapping the links between AI labs, cloud providers and chipmakers shows how much rests on a few names. What nobody can map is how they would all behave together under stress. That is what a virtual risk looks like.

With a virtual risk there is little institutional or personal experience to draw on, and no reliable data on how people, systems and suppliers will behave together. So people fall back on what they already believe. In the room I put it more bluntly: we argue from our ignorance and our prejudices. That is where the four views come out. The discussion can look like a dispute about the figures. Often it is a dispute about whether the downside can be reversed, whose experience counts and how much control the firm really has.

In my experience boards give least attention to risks that might happen but never have. Seeing them takes imagination and storytelling, which sits uncomfortably with a board’s natural wish for data. When the evidence is thin, a board tends to fall back on its dominant view without noticing it has done so. In a regulated firm that usually means the hierarchist answer. Sometimes that is right. On a virtual risk it is one view among four, presented as the only responsible one.

The same director can move

I learned this as a father before I saw it in a boardroom. When my daughters were born, I thought they could do whatever they wanted as they grew up, and the world would be kind to them. When they were a little older and due their MMR jabs, the scare over the vaccine was just beginning. For a while, a routine jab felt like a leap into the unknown, and I worried that one small nudge could bring their world to a shattering end. As they grew into teenagers, I thought that within the boundaries I set, they would be safe. And as they set out on their adult lives, I knew that all I could do was be there when they needed me. Che sera sera.

The bowl, the dome, the valley and the plain. I was the same father throughout.

This is where I part company with Douglas. She would say you are one of the four. My experience, at home and in many rooms over many years, is that people move.

In a board exercise I ask directors where their instincts put them, then change the decision. A director who wants clear rules on financial crime may favour giving a product team room to experiment. Someone who was comfortable with a pilot may take a more precautionary view once the proposal involves customers who will struggle to undo its effects. People move with the issue and with the circumstances.

Place matters too. I have run the exercise in Saudi Arabia, in the UK and in France, and the rooms came out very differently. That tells you nothing about how any one person thinks. It does tell you that the mix in a room is not an accident.

So the exercise is not there to label anyone. It is there to challenge what we assume about ourselves and about colleagues we think we already understand. The value of placing yourself in a box is seeing what makes you move out of it.

What changes in the board paper?

The four views explain why the three tensions I use in risk discussions are worth making explicit. I set them out properly in “Enough of each”. Taking risk and avoiding pitfalls means asking what the firm can learn from a trial and what it must prevent from day one. Performance culture and corporate ethics come into conflict if the team is rewarded for completing more transactions while customers who need help find it harder to reach a person. The here and now and multiple tomorrows means considering what happens if the trial succeeds and the firm becomes more dependent on the supplier, as well as what happens if it fails.

I would ask management to show how the proposal deals with the strongest objection from each view. In this case, the result might be a trial with a narrow set of reversible actions, explicit limits on what the assistant can do, a route to a person for customers who need one, and a tested way for the duty manager to suspend the service. Those are different design choices, and each comes from taking a different question seriously. They are also different routes through the design of control that I will describe in “Dimensional Control, twenty years on”. I would ask the board to agree in advance what would stop the trial and what evidence would justify extending it.

Some disagreements will remain. One director may think the benefit is worth a trial. Another may think the firm still lacks the means to put things right. The chair can ask what would change each judgement, and which of those things the firm could actually find out. The board can then decide with a clearer account of what it is relying on. That may mean holding a question open for longer than the executives would like.

At the next decision where evidence is thin and agreement comes quickly, I would ask one question. Which way of seeing this risk is missing from the room, and who will speak for it before we decide?

Sources

Richard Anderson
Founder, AndersonRisk