AI will not just change what we do. It may change how we think.

Most discussions about artificial intelligence focus on capability. What can AI do? Which jobs will it disrupt? Which industries will benefit? Which companies will win?

Founded in 1597 by Sir Thomas Gresham, Gresham College occupies a unique place in British intellectual life. For more than four centuries, it has brought leading scholars, scientists and public intellectuals into public conversation on the defining questions of their age. Its professors are charged not simply with advancing knowledge within their disciplines, but with explaining its broader significance to society.

The printing press, the scientific revolution, industrialisation and the internet all altered how societies functioned. AI increasingly appears to belong in that category.

In a recent six-part lecture series, Professor Matt Jones, Gresham Professor of Information Technology, approached the subject from a notably different angle.

Rather than focusing primarily on what machines might become capable of doing, he repeatedly returned to a more interesting question: what happens to human beings when increasingly capable systems become woven into everyday life?

For investors and entrepreneurs, that shift in perspective matters. The most important consequences of technology are often not technical. They are behavioural. History is full of examples where the second-order effects proved more significant than the innovation itself.

Across six lectures, five themes stood out:

1. The real story may be human adaptation

Much of the public debate assumes that the central question is whether machines become more intelligent.

Professor Jones repeatedly turns the question around. The more consequential issue may be how human beings adapt to increasingly capable systems.

This is a familiar pattern in the history of technology. Cars changed settlement patterns. Television changed politics. Smartphones changed attention. Social media changed communication. In each case, the most important effects were not technical but behavioural.

AI may reach further upstream. Rather than changing how we travel or communicate, it may increasingly influence how we learn, analyse, decide and judge. That observation matters because technologies rarely transform society through their capabilities alone. They transform society by altering habits, incentives and expectations. 

The internet did not simply change access to information - it changed how people consumed it. Social media did not simply change communication - it changed behaviour.

The most important consequences of AI may therefore not be what machines become capable of doing. They may be how increasingly capable systems change the way human beings think, make decisions and relate to one another.

For investors and entrepreneurs, that shift in perspective matters. The most significant opportunities may not lie solely in the companies building AI, but in understanding how AI changes customers, employees, organisations and decision-making itself.

2. AI Is becoming part of the cognitive infrastructure

We often think of AI as a tool, but tools are things we use, whilst infrastructure is something we build upon:

•  Bloomberg became infrastructure for financial markets.

• Google became infrastructure for information retrieval.

• GPS became infrastructure for navigation.

Once adopted at scale, these systems stopped being optional aids and became part of the underlying architecture through which decisions were made. AI may follow a similar path.

The important point is not that humans stop thinking. Rather, thinking increasingly occurs in partnership with systems that can retrieve information, generate options, challenge assumptions and produce recommendations at extraordinary speed. That may sound like a subtle distinction, but it is not.

A tool changes what we can do, but infrastructure changes how an entire ecosystem can operate.

Few investors today could operate effectively without market data platforms. Few businesses could function without search engines. Increasingly, many organisations may find it difficult to operate without AI-assisted analysis, planning and decision support.

If that happens, AI ceases to be a technology story and becomes an infrastructure story.

For investors, the implications extend well beyond software companies and semiconductor manufacturers. The more interesting question may be which industries are transformed when AI becomes embedded in the underlying architecture of decision-making itself. Historically, infrastructure has often proved more valuable than applications built on top of it – and that possibility deserves serious attention.

3. Some friction is productive

Businesses instinctively seek efficiency. In most circumstances that is sensible - reducing waste, eliminating duplication and automating routine tasks generally creates value.

Yet one of the most thought-provoking themes in Jones’s lectures is that some forms of friction create capability. Investors develop judgment by analysing companies, making mistakes and living through market cycles. Entrepreneurs develop judgment by managing uncertainty, allocating capital and learning from failure. Leaders develop judgment through difficult decisions whose outcomes are not immediately clear. These experiences are not efficient, rather they are formative.

AI promises to remove friction from many activities. Research can be conducted more quickly. Analysis can be produced more efficiently. Administrative burdens can be reduced dramatically.

The gains are obvious.

The challenge is that some forms of friction are also the mechanism through which expertise develops. A young investor who never builds a model may still obtain an answer. The question is whether they develop the same intuition for cash flow, leverage and capital allocation. A founder who uses AI to generate business plans, marketing strategies and market analysis may move faster. The question is whether they accumulate the same lessons along the way.

This does not mean organisations should resist automation. Quite the opposite. It does suggest, however, that the goal should not be to eliminate every source of effort. The more important challenge is distinguishing between friction that consumes time and friction that creates ability and capability.

The organisations that benefit most from AI may not be those that automate everything possible. They may be those that understand which forms of effort develop judgment and which merely slow progress.

4. Performance is not understanding

One of the more revealing episodes in the recent AI debate came from an unexpected source. Richard Dawkins, the evolutionary biologist and bestselling author of The Selfish Gene and The God Delusion, recently described extended conversations with AI systems that left him feeling as though he was interacting with something genuinely human. Many scientists disagreed, arguing that he was confusing sophisticated performance with consciousness.

Whether the systems are conscious is almost beside the point. The more interesting observation is how readily intelligent people respond to convincing simulations of understanding.

This connects directly to one of Jones’s recurring themes. AI is becoming extraordinarily good at producing answers, explanations and conversations that feel human. Yet producing the appearance of understanding is not necessarily the same thing as understanding.

The distinction matters because humans have a long history of confusing fluency with wisdom, confidence with competence and performance with expertise. For more than three decades, Bloomberg terminals have provided investors with access to quantities of data and analytical capability that would previously have been impossible to obtain, process or interpret in real time.

Yet Bloomberg did not eliminate the need for investment judgment. If anything, it increased it. When everyone has access to the same information, the differentiator is no longer information itself. The differentiator becomes interpretation.

An AI system can generate recommendations. It cannot bear responsibility for acting on them. This does not diminish the technology’s usefulness. It simply highlights the distinction between capability and accountability.

The risk is not that machines become human. The risk is that humans increasingly mistake sophisticated performance for wisdom, judgment or understanding and begin delegating decisions accordingly. As AI systems become more capable, distinguishing between assistance and authority may become one of the most important skills of all.

5. The next challenge is governance

Much of the public debate surrounding AI focuses on intelligence, but Jones repeatedly reframes the discussion around power. Historically, societies have learned to govern powerful technologies rather than eliminate them. Fire, electricity, aviation, finance and nuclear energy all created extraordinary benefits. They also created new risks.

The challenge was never simply technical. The challenge was deciding who remained accountable:

• Who bears responsibility when something goes wrong?

• Who has authority to act?

• Who controls the system?

• How are incentives aligned?

AI raises similar questions. This is particularly relevant in financial services, healthcare, education and public policy, where decisions affect real people and real outcomes.  An AI system may generate a recommendation, but a human being still bears responsibility for acting on it.

That distinction is more important than it first appears. For decades, technology has tended to automate tasks. AI increasingly assists with decisions. Yet accountability cannot be automated in the same way. In fact, the more capable AI becomes, the more valuable accountability may become.

Investors may place greater value on trusted institutions or advisors. Boards may place greater emphasis on governance. Clients may care less about who generated an answer and more about who stands behind it.

The defining question may not be whether AI can make decisions. It may be who remains responsible for them.

Conclusion: What this means for investors and entrepreneurs

Jones’s lectures are not investment lectures, but they do point towards several practical conclusions:

• Be cautious about businesses whose value proposition rests primarily on information, analysis or routine expertise. AI is likely to make many forms of professional knowledge easier to access and harder to differentiate.

• Pay close attention to businesses built around trust, reputation, distribution, proprietary data and long-term relationships. If anyone can generate information, the question increasingly becomes who you trust to stand behind it.

• Expect governance itself to become a source of competitive advantage. In highly regulated sectors such as financial services, healthcare and education, the winners may not be those with the most powerful AI systems, but those best able to combine technological capability with human accountability.

• Distinguish between automating tasks and automating judgment. Most organisations should seek to eliminate repetitive work. They should be far more cautious about removing the experiences through which future leaders, investors and entrepreneurs develop expertise.

• Look for businesses that use AI to augment people rather than simply replace them. The internet did not eliminate the need for expertise. Bloomberg did not eliminate the need for investment judgment. AI is unlikely to eliminate the need for either.

There is a tendency to assume that if AI makes information and analysis more abundant, the value of expertise must inevitably decline. History suggests otherwise. Bloomberg terminals transformed access to market data, and the internet transformed access to information, yet neither eliminated the need for judgment. If anything, they increased its importance.

When everyone has access to similar information, the advantage rarely lies in obtaining it. The advantage lies in deciding what matters. And AI may prove similar.

The defining question of the coming decade may therefore not be what machines learn to do. It may be whether increasingly capable systems leave us better at exercising judgment - or merely less practised at it.

Nick Perryman is Vice Chairman and Partner at Clarus Global Capital, and Chairman of its Investment Committee. Previously, he spent nearly two decades at UBS where was a Managing Director. He is co-author of the book, Leadership in Wealth: Mastering the Opportunities of Wealth in your Family, Firm and Society. He holds master's degrees from Durham and London universities, including in finance and organizational psychology, and is a doctoral researcher at Durham in financial services leadership, risk and governance. He is a Chartered Fellow of the Chartered Institute for Securities and Investment.

Important notice

This article is provided for general information, discussion and educational purposes only. It reflects the views of the author at the date of publication and is not intended to constitute, and should not be relied upon as, investment, financial, legal, tax, accounting or other professional advice. It does not constitute an offer, solicitation, recommendation or invitation to buy, sell or hold any investment, financial instrument or service, nor should it be regarded as a personal recommendation or as taking account of the objectives, financial circumstances or needs of any particular person.