Research
What we found
Findings from long conversations about sixteen cases, across eleven sectors, each about one real decision.
01
Where we started
This work began with a belief: that good judgment depends on being able to come back to yourself before you act.
We thought many people were not struggling because they lacked capability. They were struggling because the pace around them never let them do that.
We built a framework around that belief, and set out to test it.
Then we started listening, and followed where it led.
Over the past months we held long conversations with people working inside organisations, and with a few who advise them. Sixteen cases, across eleven sectors and many functions: from engineering and finance to public service and the not-for-profit world, from R&D and compliance to HR and operations, in owner-run businesses and large institutions alike.
We asked each of them about one real decision: what it was, where it stalled, and what finally moved it.
What we heard was rarely about intelligence, effort or goodwill. The people we spoke to were capable, and most of them were working hard at the right things.
What they described was the ground they were standing on while they decided.
That ground is what we study.
Most attempts to improve decisions start by fixing the person, or fixing the system.
We start with the people making one real decision, and work outward to whatever is shaping it.
02
Held together
The first thing we found surprised us.
Most of the organisations we studied were working.
They were being held together.
In eleven of the sixteen, something human was quietly absorbing the friction the decision system produced. Nobody had designed it. Most people had stopped noticing it was there.
Sometimes it was one person carrying far more than anyone had realised, holding in their head what good looked like because it was written down nowhere else.
Sometimes it was a discipline: a team working in small, careful steps at its own level because the level above it moved so much more slowly.
Sometimes it was a culture: people covering for one another so consistently that the gaps never showed.
Sometimes it was simply commitment: someone who cared enough to keep going after the organisation had stopped asking them to.
All of these work.
That is exactly why nobody sees them.
What differs is how long they last.
A person is one resignation away.
A discipline travels only as far as the team that practises it.
A culture outlasts both until the organisation grows past what goodwill can carry.
We have not yet found an organisation whose decisions work without someone absorbing the gap.
We have not looked for it deliberately. That is next.
So the question we would ask of yours is not whether your decisions are working.
It is who is making them work, and what happens the week they leave.
03
The cost nobody counts
Organisations can stay commercially successful long after their decision systems begin to fail.
The second thing we found shows how.
When we asked what these problems cost, people sometimes knew. They would name a figure without hesitating: new hires who did not work out, time lost to systems that failed, months spent waiting for approval.
They said it the way you might mention the weather.
These were not hidden costs. They were known costs that had stopped registering as costs.
And sometimes nobody could put a number on it at all because nobody had ever thought to measure it. That told us as much as any figure.
This is how friction survives.
Not by being invisible, but by becoming normal.
A cost that has become normal sits in no budget.
So nobody is accountable for bringing it down.
04
Where ownership breaks
The third thing we found appeared in nine of the eleven sectors we looked at.
Decisions stall when ownership fails to attach to them.
It fails in more than one way, which is why the usual remedy so often misses.
Sometimes nobody can say who owns a decision. An owner probably exists. No one can find them. The decision stops.
Sometimes several people hold a stake and none of them holds the call. The decision is made, and made again, and never settles.
From a distance those look the same. One is decision paralysis and the other is churn, and they need opposite responses.
Sometimes someone clearly owns the decision, and is overruled by someone with more power. Next time, they stop trying.
Sometimes the decision is made by people who never asked the ones who knew. Of everything we heard, this came up most often.
And sometimes the person who decides is not the person who lives with the result. Nothing carries the consequence back to the choice, so the same choice gets made again.
An organisation can have a perfectly clear org chart and fail on every one of these.
Knowing who decides is not the same as owning what happens next.
05
The same symptom, a different cause
Underneath all of this sits a question most organisations never separate out.
Some decisions stall because nobody carried them forward.
Others stall because someone decided not to prioritise them, and never said so out loud.
From the outside they are identical. The meeting ends, the action sits, the quarter turns.
But one is an absence and the other is a choice. Treating a choice as an absence wastes everyone’s time.
We heard the second more often than we expected.
A decision postponed rather than refused because postponing never has to be defended.
The familiar option kept because changing it felt riskier than living with it.
A rule left in place by the people who could change it because the rule offered them cover.
Escalation works the same way.
When capable people keep sending decisions upward, the useful question is not whether they can decide.
It is what they would need in order to decide, and whether anyone has ever told them.
06
What survives
We asked what survives a change of leadership.
The honest answer, more than once, was: whatever you are made to do from outside.
Work a regulator requires survives. Work a client would notice survives.
Everything else, however proven, however valuable, depends on whether the next person chooses to carry it.
Things that work can disappear with the people who championed them. Not because they failed but because they belonged to someone who is no longer there.
That is not governance.
It is a dependency on outside pressure, doing the job internal accountability should be doing.
07
What we mean by conditions
Put these findings side by side and a pattern appears.
The people in these organisations were not failing.
They were responding sensibly to the conditions they were in.
By conditions we mean what people are operating under when they have to judge and decide: created by the organisation, by the work itself, or by the world around it.
Across these sixteen cases, four conditions came up again and again, in sectors that have little else in common.
Invisible deciders
Not knowing who decides, or what it would take to get a decision.
No path to influence
Knowing exactly who decides, and having no way to influence them.
The authority gap
Responsibility without authority: deciding without carrying the outcome, or carrying an outcome you had no say in.
Hidden load-bearers
A small number of people carrying far more than the structure accounts for.
None of these is exotic. Almost none of them appears in the way organisations measure themselves.
For decades organisations have worked on the system.
Technology. Process. Structure. Incentives. Governance. And now AI.
And then, somewhere near the end: now we need the people to adopt it.
But the system does not carry the consequences of a decision.
Neither does the AI inside it.
- People do.
- People notice.
- People interpret.
- People weigh the risk.
- People make the call.
- People live with it.
And they do all of that under conditions.
That is what we mean by the Human Decision Layer: the layer between the system you have designed and the people who have to make it work.
There is a second half we have not yet studied properly: what these conditions do to the person inside them. To attention. To fear. To recovery.
We have not asked about it yet, so we will not claim it.
It is the next round of the research.
08
Why now, and where we were wrong
We began with a belief about people. A view about technology came later: that AI was being adopted faster than organisations could decide what to do with it.
Often that is exactly what we saw. Capability landing, and the decisions around it still catching up.
But where we tested it directly, it did not always hold. The decisions were ready. The technology was not.
So we changed the claim.
The gap between what an organisation can do and what it can decide runs in both directions. Which way it runs in yours is the first useful question, and the two answers call for opposite responses.
What does not change is this.
When a machine produces ten options in seconds, someone still has to decide which one matters.
When it recommends an action, someone still has to decide whether to trust it.
When it can act on its own, someone has to have decided what it is allowed to do, and who answers for it.
Most thinking about AI oversight assumes there is a reliable person at the point of oversight.
What if that person is overloaded, short of time, unsure who owns the call, or reluctant to challenge the machine or the person above them?
The framework can be perfect.
The human can still fail.
Judgment is still done by people, under conditions.