When expertise moves faster than authority

There is a paradox I have become increasingly interested in.

I have spent much of my professional life around technology, from the early Internet to mobile communications, payments, digital transformation and now artificial intelligence. Over those decades, almost everything about the production and circulation of knowledge has accelerated.

What once required days of research can now take hours. What required access to specialised libraries can be retrieved from a laptop. What once depended on knowing the right expert can increasingly be explored through digital networks and, today, AI.

Knowledge has become extraordinarily fast.

Organisations have not.

And after working across different environments – technology companies, financial institutions, international initiatives, universities and public organisations – I have come to believe that this difference in speed is becoming one of the most important organisational problems of the AI era.

It is not simply a technology problem.

It is a problem of institutional velocity.

When knowledge was slow

For much of the twentieth century, organisations were designed around the scarcity of information.

Information travelled through hierarchies. Expertise was concentrated in particular departments and professions. Seniority often corresponded, imperfectly but reasonably, with accumulated knowledge and experience.

The organisational pyramid therefore had an informational function.

The person at the top was not necessarily the most knowledgeable person in every domain, but information was progressively aggregated as it moved upwards. Decisions then travelled downwards.

It was slow, but the external world was also relatively slow.

That assumption has disappeared.

A professional today can access research, regulation, market intelligence, technical documentation, international experiences and specialised communities almost instantaneously.

AI accelerates this further.

It does not merely give us more information. It dramatically reduces the time between question and knowledge.

And this is where the problem begins.

Because reducing the latency of knowledge does not automatically reduce the latency of an organisation.

I have seen technology accelerate. I have seen organisations resist acceleration.

This contrast is particularly striking to me because I have watched several technological transitions from relatively close range.

The early Internet dramatically reduced the cost of moving information.

Mobile technologies reduced the constraints of place.

Digital platforms reduced the friction of transactions.

Cloud computing reduced the friction of accessing computing resources.

And AI is now reducing something even more fundamental: the friction involved in accessing and manipulating knowledge itself.

Yet many organisational mechanisms remain surprisingly familiar.

Committees.

Approval chains.

Organisational boundaries.

Hierarchical permissions.

Informal networks.

Internal politics.

And, inevitably, personal relationships.

None of these is inherently illegitimate. Organisations need governance. Institutions need accountability. Relationships create trust, and trust is an essential component of collective action. Public institutions in particular cannot simply optimise for speed; legitimacy, transparency, due process and responsibility matter.

But there is a point at which necessary governance becomes organisational latency.

And there is an even more uncomfortable point at which professional competence and accumulated experience compete with proximity, friendship, internal alliances or political convenience for the right to influence a decision.

That is where the speed-of-knowledge problem becomes a problem of organisational intelligence.

The knowledge may already be inside the organisation

We often assume that organisations struggle because they lack skills.

Sometimes they do.

But my experience increasingly suggests another possibility: organisations may already possess much more intelligence than they are capable of using.

Someone somewhere may understand the problem.

Another person may have solved something similar twenty years earlier.

Someone may know the technology.

Another may understand the institutional constraint.

Someone may have international experience that could prevent the organisation from repeating an error already made elsewhere.

The knowledge exists.

But organisational structures determine whether that knowledge ever reaches the place where a decision is made.

This distinction matters enormously.

A knowledge deficit can be addressed through recruitment, consulting, education or technology.

An intelligence-routing deficit is much harder.

It means that the organisation has knowledge but cannot bring the right knowledge, from the right people, into the right conversation at the right moment.

AI alone cannot solve that.

In fact, AI may make the problem more visible.

When professionality is not enough

There is another dimension that organisations rarely discuss openly.

Formal organisations have organisational charts, roles and processes.

Real organisations also have invisible architectures.

• Who knows whom.
• Who trusts whom.
• Who belongs to which internal network.
• Who has access to decision-makers.
• Who is considered politically convenient.
• Who is perceived as “one of us”.
• And who, regardless of professional experience, remains outside those circles.

This is not exclusive to public administration. I have seen variations of it in corporations, universities, associations and professional communities.

The problem appears when this invisible architecture becomes stronger than the architecture of competence.

At that point, an organisation can paradoxically employ extremely experienced people while systematically underusing their experience.

It can recruit talent without listening to it.

It can invest in knowledge while rewarding proximity.

And it can purchase sophisticated AI systems while leaving its fundamental decision architecture untouched.

That is why I increasingly think that the discussion about AI transformation is starting in the wrong place.

The real AI readiness test

We ask organisations:

Do you have an AI strategy?

Do you have the right data?

Which models are you using?

Have you deployed copilots or agents?

Do you have an AI governance framework?

All legitimate questions.

But I would add another one:

How quickly can relevant knowledge influence a decision in your organisation?

That question reveals something that technology assessments often miss.

Imagine that a person discovers something important today.

How many organisational layers separate that knowledge from the person capable of acting on it?

And then ask an even more uncomfortable question:

Does the probability of that knowledge being heard depend primarily on its quality — or on the organisational position and relationships of the person presenting it?

That is an institutional intelligence test.

And AI makes it increasingly urgent.

From information latency to institutional latency

In computing, we obsess over latency.

Milliseconds matter.

Networks are optimised.

Databases are tuned.

Caches are introduced.

Computational bottlenecks are identified and removed.

Yet organisations can tolerate extraordinary latency in the movement of knowledge.

An insight can spend weeks moving through meetings.

Expertise can remain trapped inside a department.

A proposal can wait months for the appropriate organisational sponsorship.

Experience accumulated over decades can be ignored because it comes from the wrong part of the hierarchy.

We would never design a computer system this way.

Imagine a distributed architecture in which the best available node possesses the answer, but the system refuses to query it because that node is not sufficiently close to the central server.

Technically, we would call that poor architecture.

Organisationally, we often call it normal.

AI will widen the gap

This is why I believe the next phase of AI adoption will produce an unexpected divide.

The difference will not simply be between organisations that use AI and organisations that do not.

Almost everybody will eventually use it.

The more important difference may be between organisations that accelerate knowledge production and organisations that also accelerate knowledge absorption.

The first will generate more reports, more analyses, more presentations and more recommendations.

The second will actually become more intelligent.

That distinction is fundamental.

If an organisation introduces AI into a slow decision architecture, it may simply produce knowledge faster than the organisation can consume it.

The result is not necessarily transformation.

It may simply be a larger queue.

Experience matters differently now

There is also a personal dimension to this for me.

After several decades working through successive waves of technological change, I do not think experience should give anyone an automatic right to be heard.

Experience can become obsolete. Seniority can produce certainty where curiosity would be more valuable. Younger professionals frequently see possibilities that experienced people overlook.

But the opposite mistake is equally dangerous.

Experience is not simply the number of years someone has worked.

At its best, it is compressed pattern recognition.

You recognise situations because you have seen versions of them before. You remember which apparently revolutionary ideas were attempted twenty years ago. You know which constraints turned out to matter and which fashionable assumptions disappeared.

AI makes access to explicit knowledge dramatically easier.

That may make genuine experience more valuable, not less, because judgement becomes increasingly important when information itself becomes abundant.

The intelligent organisation therefore needs both.

New knowledge and accumulated experience.

Technology and institutional memory.

Young perspectives and senior judgement.

External signals and internal understanding.

The problem begins when the organisation selects among these not according to their relevance, but according to hierarchy, relationships or political convenience.

The organisation as a knowledge-routing system

Perhaps we therefore need to change the metaphor.

An organisation is not simply a hierarchy of responsibilities.

It is a knowledge-routing system.

Its intelligence depends on its ability to discover where relevant knowledge exists, move it across boundaries, combine different forms of expertise and bring them into decisions.

Seen this way, organisational design becomes very different.

The important question is no longer merely:

Who reports to whom?

It becomes:

Who needs to know what – and whose knowledge needs to reach whom?

That is a much more interesting question for the AI era.

Because AI is making knowledge faster every month.

Technology will continue to accelerate.

Markets will continue to accelerate.

The circulation of ideas will continue to accelerate.

The scarce resource may therefore no longer be knowledge itself.

It may be the institutional capacity to recognise, absorb and act upon knowledge.

And this leaves organisations with an uncomfortable choice.

They can use twenty-first-century intelligence inside twentieth-century decision architectures.

Or they can redesign themselves so that expertise, experience and evidence can travel at something closer to the speed of the world around them.

The most dangerous latency in the AI era may not be computational latency.

It may be the distance between knowing and being heard.