When preparing a decision well is not enough

AI is making it easier to arrive prepared for a decision.

It can summarise documents, compare information, retrieve precedents, build scenarios, anticipate questions, organise alternatives.

It is compressing a portion of the work that, until recently, required time, people, and access to different skills.

Yet preparing a decision better does not necessarily mean making it better.

The distinction is subtle.

And for this exact reason, it risks becoming increasingly important.

More preparation does not automatically mean more judgment

In the 2026 Annual Corporate Directors Survey, PwC highlights an interesting finding: 81% of the interviewed directors believe that their board could challenge management more effectively on strategy.

Among the elements indicated to improve this capacity are clearer metrics, strategic alternatives, explicit trade-offs, scenario analysis, and benchmarks.

In the same study, 71% identify AI as the primary skill to strengthen on the board. But 35% also indicate, among the risks tied to AI, an excessive reliance on its outputs with the consequent weakening of human judgment.

These two pieces of evidence, read together, open an interesting question.

It is not about the contrast between human beings and technology.

It is about what happens when our capacity to prepare a decision increases enormously, while the question remains open regarding the quality with which we test what is leading us to decide.

Decision preparation, decision judgment, decision ownership

They are three different functions.

Decision preparation is about what we bring to the decision.

Information. Evidence. Precedents. Scenarios. Alternatives. Summaries. Simulations.

It is a territory in which AI is becoming extraordinarily powerful.

Decision judgment is about what we do with what we have in front of us.

Which element truly carries weight?

Which alternative is missing?

Which assumption have we stopped treating as a hypothesis?

Which trade-off are we accepting without having named it?

What in the context is shaping our reading?

And above all: what question are we not asking?

Then there is a third function.

Decision ownership.

Someone, ultimately, takes the decision.

And with it, the consequences.

These three functions can collaborate.

They are not interchangeable.

The risk of the very well-documented decision

A decision can have dozens of pages of analysis behind it.

It can be supported by correct data.

It can have alternative scenarios.

It can have been discussed by competent people.

It can even appear perfectly rational.

And still have a blind spot.

Because the blind spot is not necessarily missing information.

It can be an alternative that no one has an interest in bringing into the conversation.

A trade-off that the group prefers not to make explicit.

A reputation that attributes more weight to one voice over another.

A hypothesis repeated so many times that it has become a fact.

A real time pressure — or a perceived one.

A relationship.

A role.

A distribution of authority that makes some questions much harder to formulate than others.

The problem, at that point, is not having more data.

It is managing to see what is already acting upon the data we have.

AI is entering decision preparation

The phenomenon is already visible in boards.

OnBoard's 2026 Board Effectiveness Survey reports that 92% of interviewed directors used AI for board-related activities in the previous six months.

The applications clearly tell the story of what is happening: summarising board books and documents, preparing minutes and agendas, researching and analysing data, presentations, anticipating questions.

It is difficult to imagine this direction stopping.

And it is equally difficult to imagine it remaining confined to boards.

Preparing a decision will become progressively faster.

Retrieving information will cost less.

Generating alternatives will be simpler.

Building an initial scenario analysis will require less time.

This does not necessarily diminish the value of human work.

It can make the point where it begins more evident.

Before the decision

Perhaps, then, the question will no longer be simply:

Do we have enough information to decide?

Other questions might become the important ones.

What is the real decision?

Which alternatives are not entering the conversation?

Which trade-offs have not yet been made explicit?

Which assumptions are we treating as facts?

What in the field is influencing the decision without being named?

Who contributes to the decision and who truly possesses the responsibility for it?

What changes if we decide — and what changes if we do not decide?

These are not questions that replace analysis, data, or technical competence.

They come after.

Or, perhaps, they should arrive exactly when everything else already seems sufficient.

Judgment is not the residue left by AI

There is an understandable temptation in the current debate: trying to establish what will be left to human beings when AI is capable of performing a growing share of cognitive work.

I do not believe that judgment should be thought of as what remains.

It is a different function.

A machine can enormously expand what enters the field of the decision.

It can find elements we had not considered.

It can connect quantities of information that no person could process in the same time.

It can help us prepare the field better.

But the availability of more elements does not eliminate the responsibility of assigning them a weight.

And above all, it does not eliminate the responsibility of the choice.

PwC notes simultaneously a growing need for AI competence in boards and the concern that an excessive reliance on its outputs could weaken human judgment.

Perhaps the issue is not choosing between the two.

It is avoiding confusing their functions.

When the preparation is finished

In high-responsibility work, a moment arrives when continuing to add information produces diminishing returns.

The dossier is ready.

The scenarios are there.

The numbers have been checked.

The competent people have been heard.

The decision, however, must still be taken.

That is exactly where it can become useful to change the question.

Not only:

Have we prepared this decision well?

But:

What have we not yet tested before transforming it into a commitment?

Because a well-prepared decision can still have a blind spot.

And no amount of preparation makes it less important to know who, ultimately, possesses the judgment and the responsibility to decide.