Is the AI investment boom creating real value?

Billions are flowing into artificial intelligence, but the size of the investment tells us remarkably little about the value being created.

For boards, investors and executive teams, this distinction is becoming difficult to ignore. Organisations are spending heavily on AI infrastructure, models, products and transformation programmes. Companies are being valued partly on expectations of what AI may enable in the future. At the same time, consumers and employees are being told that the technology will make their lives easier, businesses more productive and economies more prosperous.

The difficult question is whether those promises will translate into outcomes.

Episode 15 of The AI Values Podcast brings Edosa Odaro and Lindley Gooden together to examine the widening gap between AI investment and demonstrated value. Their discussion considers previous periods of technological investment, including the expansion of the railways and the dot-com boom, to ask what leaders can learn from earlier moments when genuine technological progress became entangled with speculation.

The issue is not whether AI has value. It is whether organisations can distinguish that value from the expectations surrounding it.

Investment is not evidence of value

Large investment figures have become part of the story organisations tell about AI.

New infrastructure projects, model development and corporate commitments are frequently presented as evidence of momentum. Yet money entering a technology does not prove that equivalent value will emerge from it.

Gooden expresses his frustration with that assumption directly:

"I'm fed up of hearing about how many tens of billions have been spent on AI."

His concern is what happens after the announcement. Investment can generate headlines and demonstrate ambition, but organisations eventually need to show what changed as a result.

"It's a branding exercise. It's a PR exercise. But we all need to have a positive result of all that investment."

This is particularly relevant for boards approving significant AI expenditure. The governance question cannot end with how much is being invested or whether competitors are spending similar amounts. Leaders need to understand the intended outcome and how they will recognise whether it has been achieved.

Efficiency may be one outcome. Growth may be another. AI might improve a service, reduce repetitive work or allow an organisation to create something previously impossible.

Whatever the objective, investment needs a credible connection to value.

Without it, expenditure can become its own measure of progress.

The railway problem is happening again

Periods of major technological change often attract more investment than the market can ultimately justify.

Odaro uses the development of the railway as an analogy. Rail fundamentally changed economies and societies. Its long-term value is not in doubt. But that did not mean every railway project deserved to succeed.

"In the end, what's happened is we have lots of rail, but not every rail idea was successful."

Capital followed the promise of the underlying technology. Numerous projects appeared, investment aligned itself with those projects and expectations grew. Some railways produced lasting value. Others did not.

The distinction is useful because a speculative boom and a valuable technology can exist at the same time.

The dot-com era offers a more recent example. Many companies failed and considerable investment disappeared, but the internet itself did not prove worthless. Digital services went on to become fundamental to commerce and everyday life.

AI could follow a similar pattern. Some of today's investments may become essential infrastructure. Others may disappear. The existence of weak investments does not invalidate the technology, just as the importance of the technology does not validate every investment made in its name.

For leaders, this means avoiding a binary debate about whether AI is either a bubble or the future. The more useful task is determining which investments are connected to sustainable value.

AI has an expectation economy

One of the most useful distinctions in Episode 15 is between two economies developing around AI.

The first is based on value. It concerns what AI actually enables organisations and individuals to accomplish.

Odaro describes this in practical terms:

"The value based bit of it is very much the one that promises productivity gains."

Those gains can include completing work faster, creating more products, serving more customers or improving satisfaction. They are outcomes that can eventually be observed and assessed.

Alongside this sits an expectation-based economy.

"Another one is kind of an expectation based economy."

This economy is driven by what people believe AI will eventually deliver. Expectations can influence company valuations, investment decisions and the perceived importance of particular technologies long before their benefits have been demonstrated.

That distinction matters for AI governance because expectations can create their own incentives. Organisations may invest because competitors are investing. Executives may fear appearing slow. Investors may reward an AI narrative before operational evidence exists.

None of those pressures necessarily mean an investment is wrong. They do, however, make disciplined decision-making more difficult.

Boards should therefore understand which part of an AI business case depends on demonstrated value and which part depends on expectations about the future. Treating the two as equivalent can make speculation look like performance.

AI washing makes value harder to see

When a technology attracts significant attention, organisations have an incentive to associate themselves with it.

AI is no exception.

Products and services are increasingly described as AI-powered, AI-driven or AI-enabled. Sometimes the description reflects a material change in capability. In other cases, the label contributes more to positioning than to the product itself.

Odaro refers to this as AI washing:

"AI washing is a classic where you kind of stick an AI label onto everything."

The governance concern is not simply about marketing language. When AI becomes a catch-all label, it becomes harder for decision-makers to distinguish meaningful capability from superficial association.

This can distort investment decisions. Leaders may find themselves evaluating whether something contains AI rather than whether it solves an important problem.

A stronger approach begins with the problem, the intended beneficiary and the outcome. Only then should leaders determine whether AI is the appropriate mechanism for achieving it.

That discipline becomes increasingly valuable as the AI label itself acquires commercial value.

The customer eventually decides what matters

The debate about AI investment can easily become detached from the people who are supposed to benefit from it.

Gooden repeatedly returns the discussion to the consumer.

"What I care about is features that will help me in my life."

This provides a useful counterweight to investment announcements and technical benchmarks. From the customer's perspective, the amount spent developing a capability is largely irrelevant. What matters is whether the resulting product or service becomes more useful.

The same principle applies inside organisations.

Employees experience AI through changes to their work. Customers experience it through changes to products and services. Investors eventually experience it through business performance. Public institutions experience it through whether outcomes improve for citizens.

AI value therefore depends on perspective. A deployment that reduces organisational costs but makes a service significantly worse for customers may create one form of value while destroying another.

Boards need visibility across these competing outcomes.

Responsible AI governance should ask who receives the benefit, what form that benefit takes and whether costs or risks are being transferred to someone else.

Every benefit has another side

AI can produce considerable benefits. It can increase productivity, remove repetitive work and create new products and services.

But assessing only what AI adds produces an incomplete business case.

Odaro argues that organisations also need to consider what may disappear as AI becomes embedded:

"AI is going to give us tons and tons of things."

The corresponding responsibility is to understand what is being exchanged for those gains.

"We need to avoid being blindsided by the things that it takes out on the other end as well."

Those costs will differ between organisations. They may involve jobs, skills, human interaction, resilience, privacy, flexibility or dependence on external technology providers. Some may be acceptable trade-offs. Others may undermine the value the organisation hoped to create.

This is why AI investment decisions need more than a return calculation based on productivity.

Leaders need to examine both sides of the value equation. What becomes better because of AI, and what might become weaker? Who receives the gain, and who absorbs the downside? Are short-term efficiencies creating dependencies or consequences that will become expensive later?

An investment can generate measurable financial returns and still create organisational costs that have not been properly recognised.

Is AI a bubble?

The question inevitably emerges when investment, expectations and technological enthusiasm rise together.

Is AI a bubble?

The answer is unlikely to be captured by either extreme.

The railway analogy shows that enormous speculation can accompany infrastructure that subsequently changes society. The dot-com period shows that inflated expectations and failed companies can coexist with a technology that goes on to become indispensable.

Gooden captures the uncertainty in a more provocative question:

"So is it a bubble only if you blow one?"

Whether today's investment cycle eventually experiences a major correction cannot be established from the transcript. What leaders can control is whether their own organisations contribute to an expectation-driven cycle without maintaining discipline around value.

That means resisting the assumption that investment is justified simply because the technology is important.

AI may prove enormously valuable while individual AI investments fail.

Both things can be true.

Listen to Episode 15

Going deeper: Boards need to govern expectations as well as investment

AI governance is usually discussed in terms of model risk, data, privacy, accountability and responsible use. Episode 15 highlights another responsibility: governing expectations.

Expectations influence how capital is allocated. They affect the targets executives set, the promises organisations make and the speed at which technologies are adopted. When expectations become detached from evidence, leaders can find themselves pursuing AI because the organisation feels it cannot afford not to.

That creates a governance problem.

Boards should be able to distinguish between investments intended to generate measurable value now, investments designed to build capabilities for the future and speculative investments made because future opportunities remain uncertain. Each can be legitimate, but they should not be presented as though they carry the same evidence or risk.

The same discipline applies to consequences. A credible AI business case should account for more than what the technology adds. It should consider what may be displaced, weakened or made more dependent as a result.

This does not require leaders to predict the future perfectly. The railway and dot-com examples demonstrate why that is unrealistic.

It requires something more achievable: clarity about what the organisation believes, what evidence supports that belief and what outcome would demonstrate that the investment has actually created value.

The size of the AI investment race makes that discipline more important, not less.