AI investment will unleash innovation, but people need permission to fail: PwC boss

SINGAPORE – The vast sums flowing into artificial intelligence will produce a new generation of businesses and public listings, with the biggest breakthroughs likely to emerge in places where people are allowed to experiment, fail and try again, said Mohamed Kande, global chairman of professional services firm PricewaterhouseCoopers International (PwC).

“There is way too much money being channelled into the technology for it not to create a new age of innovation,” he told The Straits Times in Singapore, where he discussed the trends shaping corporate decisions he was seeing among PwC’s clients worldwide.

Besides navigating geopolitical tensions, rising energy costs and changing supply chains, he said, many global chief executives are thinking more about how to generate returns from AI as opposed to cutting costs, while others are using the technology to bring new ideas to fruition.

But the large volume of investment going to AI does not guarantee returns or success. Innovation often emerges only after repeated attempts, while a society’s tolerance for failure gives people the confidence to take the risks needed to succeed, he said.

“People always talk about what succeeded. They never talk about all the times that you failed. But it was many failures for one success.”

Asked where entrepreneurs had the best chance of recovering and succeeding after several failed attempts, Kande, who leads a team of more than 364,000 people in 137 countries and territories worldwide, pointed to the US, where there is generally greater acceptance of risk-taking.

That willingness to experiment will become increasingly important as AI moves beyond the frontier models built by a handful of technology giants such as OpenAI and Anthropic and into the inference phase, where trained models are applied to everyday business operations.

These applications range from using AI to operate a factory and support drug discovery to improving hospital operations and serving customers, said Kande, who previously led PwC’s global advisory business, overseeing investments in generative AI, managed services and new business models.

A new generation of IPOs

Kande said a new generation of businesses emerging from the AI boom will not be the builders of new frontier models such as Alphabet, Amazon or Microsoft, which require enormous computing power and vast amounts of data to train.

Instead, they will use existing models to solve problems in different industries.

Importantly, this wave of experimentation and investment should eventually lead to more companies capable of seeking initial public offerings (IPOs) globally.

Kande said recent hype over the anticipated blockbuster listings of firms such as Anthropic and OpenAI, as well as that of SpaceX, may have created the impression that a small group of large technology companies dominate IPO markets.

But he described this as the “tyranny of the few”, where the most prominent names obscure the many smaller businesses also raising capital.

In fact, many start-ups across the wider AI ecosystem could eventually go public, including those developing inference applications for factories, hospitals and other industries, as well as companies providing the infrastructure and equipment needed to support them.

“IPOs will remain an important way for these companies to raise capital,” said Kande, adding that whether a market develops a strong IPO pipeline will therefore depend partly on whether it has a healthy innovation ecosystem in which companies can start, experiment and grow.

An AI bubble?

Still, the sheer volume of capital going into AI has also raised questions about whether the sector is in a bubble and whether a correction could trigger another global financial crisis.

AI is no longer just a new technology or a collection of popular applications. It has become a vast ecosystem spanning the entire economy, Kande said.

While the initial conversation focused heavily on using AI to automate work, reduce costs and potentially cut jobs, more companies are now asking how the technology can help them make money by getting more done and creating new revenue sources.

While Kande agreed that there is “definitely hype” surrounding AI as the technology continues to attract billions of dollars, he noted that many businesses supplying the AI ecosystem have strong fundamentals and are struggling to keep up with genuine demand.

“Large technology companies may be spending heavily and putting pressure on their cash flow, but many are also delivering robust earnings,” he said.

Yet, their results are described as disappointing when measured against elevated investor expectations, rather than because the companies themselves are performing badly.

When asked, Kande noted that it is impossible to predict whether a bubble would form, when it might burst or what would cause the next crisis. But what is likely is that AI’s penetration into almost every part of the economy could mean that any future crisis involving the sector could look very different from the 2008 global financial crisis.

“The 2008 crisis was centred largely on financial services. The AI ecosystem spans eight to 10 industries, from technology and semiconductors to property, energy, manufacturing and cooling equipment,” he said.

But not all these firms will be financed in the same way or carry the same level of debt. A downturn would therefore affect different parts of the ecosystem unevenly, making it harder to determine where a crisis may begin and how it might spread.

Kande stressed that he is not predicting that AI will cause the next financial crisis. Rather, the connections formed as AI spreads across industries have made the old model of a crisis contained within one sector less applicable.

The broader paradox, he said, is that countries are pulling apart politically even as their industries become more closely intertwined, with AI connecting technology, energy, property and manufacturing. As a result, “any future economic shock may no longer be contained within one recognisable sector”.

PwC, for example, now thinks less in terms of separate sectors such as technology, energy and financial services, and more in terms of ecosystems that meet fundamental needs: how people obtain computing power, feed themselves, move, build and access healthcare.

But AI’s widening reach also creates new challenges for the economy. While the technology can remove repetitive tasks and help employees become more productive, companies must work out how junior staff will gain the experience needed to assess AI-generated work later in their careers.

Asked how junior employees would develop into skilled practitioners if AI takes over the repetitive and manual tasks that traditionally provide their grounding, Kande said there is no established playbook for now.

Companies will have to experiment with different approaches and accept that some will fail, he said.

At PwC, for example, AI has not replaced staff, who regard the technology as a “digital colleague” that helps them work better, faster and more safely. While AI could take over repetitive tasks, Kande rejected using it simply as a way to cut jobs.

Still, the uncertainty reinforces his wider argument that the businesses and economies that benefit most from AI will not necessarily be those that always get it right, but those most willing to try.

“If you don’t know what the future is going to be, create it,” he said.