The Greatest Risk to Asset Managers May Not Be AI…

…It May Be Who Controls It.

For the past two years, the asset management industry has been obsessing over one question:

How can we use AI?

The better question may be:

What happens when AI starts using us?

Every conference agenda now contains a panel on artificial intelligence. Every strategy document references machine learning, automation or generative AI. Every CEO is asking their management team how the business can become more efficient, more productive and more competitive through technology.

The assumption underpinning most of these discussions is that AI represents a tool.

A powerful one, certainly.

But still a tool.

What if that assumption is wrong?

What if AI is rapidly becoming infrastructure?

And what if the greatest threat to asset managers is not that they fail to adopt AI quickly enough, but that they become dependent on systems they neither control nor fully understand?

Recent discussions among CFOs and COOs highlighted a fascinating and troubling phenomenon: being "de-agented."

Several organisations reported concerns following incidents where enterprise users of large language models found their access suspended, restricted or altered with little warning, minimal explanation and virtually no customer support.

The immediate reaction was frustration.

The more thoughtful reaction was far more concerning.

What happens when a critical business capability suddenly disappears?

The Bloomberg Terminal Moment

For decades, asset managers have relied on third-party infrastructure.

Bloomberg terminals.

Market data providers.

Custodians.

Index providers.

Ratings agencies.

The difference is that everyone understood those relationships.

Contracts existed.

Service-level agreements existed.

Escalation routes existed.

Accountability existed.

The emerging AI ecosystem often feels very different.

Many firms are embedding AI into:

• research workflows,

• client reporting,

• compliance processes,

• operational controls,

• marketing content,

• investment analysis.

Yet relatively few have fully considered the implications of dependency.

Imagine if an asset manager's investment team increasingly relies on one AI platform to summarise earnings calls, analyse company filings and prepare investment briefs.

Then one day access is restricted.

Or pricing changes dramatically.

Or data permissions change.

Or the provider simply decides that your use case no longer aligns with its commercial strategy.

Suddenly, what appeared to be a productivity tool reveals itself to be a critical dependency.

Asset managers have spent decades diversifying portfolios.

Few are currently diversifying their AI risk.

The New Key-Person Risk

Traditionally, firms worried about key-person risk.

What happens if the star portfolio manager leaves?

What happens if the CIO retires?

What happens if a crucial relationship disappears?

AI may be creating a new form of concentration risk.

What happens if your firm's most important employee isn't actually an employee?

What happens when that employee is a language model controlled by someone else?

Across the industry, thousands of professionals are already using AI to:

• draft investment commentary,

• analyse documents,

• conduct research,

• prepare presentations,

• review contracts,

• summarise meetings.

In many cases, these systems are becoming embedded before governance frameworks have fully caught up.

The irony is remarkable.

The same industry that spends enormous resources measuring operational risk may be introducing entirely new forms of operational dependency without fully understanding the consequences.

Data Is the New Custody Risk

Asset managers have always been custodians of trust.

Clients provide capital.

Managers provide expertise.

Underlying that relationship is data.

Sensitive data.

Proprietary research.

Investment decisions.

Client information.

Commercial intelligence.

This is where AI becomes particularly uncomfortable.

Because unlike traditional software, many AI systems improve by learning.

The obvious question therefore becomes:

Learning from what?

Every asset manager must eventually confront difficult questions.

Where is data stored?

Who has access?

How is it used?

Can it be retrieved?

Can it be deleted?

Can it be separated from broader training environments?

The industry has become increasingly sophisticated in managing cyber risk.

Yet many firms appear far less sophisticated when assessing AI-related data risk.

The danger is not necessarily malicious behaviour.

The greater risk may be misunderstanding.

Many users simply do not know enough about how these systems operate.

And ignorance has rarely been an effective risk management strategy.

The Death of Competitive Advantage

There is another issue that receives surprisingly little attention.

If everyone uses the same AI systems, what happens to differentiation?

For decades, asset managers have sold:

• unique insight,

• proprietary research,

• differentiated thinking.

But what happens when everyone has access to similar models trained on similar information?

The danger is not that AI produces bad analysis.

The danger is that it produces remarkably similar analysis.

Asset management has always rewarded original thought.

Consensus rarely generates alpha.

Yet there is a growing risk that AI creates an industry-wide homogenisation of research, commentary and decision-making.

The result could be an uncomfortable paradox.

The more firms adopt identical tools, the harder it becomes to create differentiated outcomes.

The Human Premium

Perhaps the most interesting consequence of AI is that it may increase the value of being human.

That sounds counterintuitive.

Most discussions assume the opposite.

Yet throughout history, automation has tended to increase the value of skills that machines struggle to replicate.

Judgement.

Trust.

Relationships.

Context.

Experience.

Asset management is ultimately not a technology business.

It is a trust business.

Clients do not allocate capital solely because a model generated a recommendation.

They allocate capital because they trust the people making decisions.

No institutional investor appoints a manager because their AI writes particularly good meeting notes.

They appoint managers because they believe those managers understand risk, opportunity and uncertainty better than their competitors.

AI may improve efficiency.

It cannot eliminate the need for judgement.

At least not yet.

The Real Future of Asset Management

The industry's current AI conversation is largely focused on productivity.

How many hours can be saved?

How many reports can be automated?

How many processes can be streamlined?

These are important questions.

They may also be the least interesting ones.

The more profound question is whether asset managers are becoming dependent on infrastructure they do not own, cannot control and may not fully understand.

The firms that thrive over the next decade may not be those that adopt AI fastest.

They may be those that govern it best.

Those that diversify providers.

Those that understand operational dependencies.

Those that protect proprietary information.

Those that recognise where human judgement remains irreplaceable.

And those that remember a simple truth:

Technology has always promised efficiency.

What clients ultimately pay for is confidence.

Why OFFLINE May Be the Most Important Event in Finance

Perhaps the most revealing observation from recent discussions among CFOs and COOs was not about AI itself.

It was about the value of stepping away from it.

At a time when technology dominates every conversation, there is something almost radical about creating space for people to disconnect.

To talk.

To challenge assumptions.

To compare experiences.

To learn from peers.

Because the biggest questions surrounding AI are not technological.

They are strategic.

They are operational.

They are human.

And unlike a language model, those questions rarely have simple answers.

The future of asset management will undoubtedly involve AI.

The real question is whether the industry is building a future where it controls the technology—or one where the technology increasingly controls it.

That distinction may determine the winners and losers of the next decade.