Is Not Pursuing AI Like Burning Money?

A $150 million stack of $100 bills would rise to approximately 600 feet, or around 200 meters, and tower over the executive’s corporate headquarters. Burning each bill separately would take around 600 days. With the right amount of accelerant and appropriate stacking, I am sure it could be achieved much more quickly.

By the way, do not try this at home. In the US, and I am sure in many other jurisdictions, burning money can land you in jail. This was purely a thought experiment to push the executive’s thinking.

So what is AI anyway?

My involvement in AI goes back to the early 1990s, when I developed genetic algorithms to model complex chemical reactions. I even published a paper on my research findings in a peer-reviewed scientific journal.

I sometimes wonder where I might be now if I had kept the faith and practiced in AI through the various AI bubbles and winters. Perhaps I would now be basking under the glorious sun of the AI summer we are experiencing.

Instead, I bring more of a hard-nosed business focus to the dialogue on AI, while still, hopefully, understanding enough about the practicalities of implementing AI to be dangerous and useful.

Like metaverse, blockchain, and digital before it, AI can sometimes feel like the latest buzzword to use when hoping to pump up the value of your company. A few years ago, I read that only 25% of AI start-ups were actually using AI. Over the past six months, we have also seen how generative AI announcements can easily move a company’s stock up or down by 10%. Today, I saw a four-week-old AI company valued at $250 million.

AI is one of those small words that can cover a multitude of sins, just as the words “hedge fund” or “digital” can have broad and ambiguous meanings. The confusion around AI often stems from its broad and vague definition.

Some company executives have even gone as far as banning the use of the word AI, preferring instead to be specific about the types of AI applications being used, such as natural language processing, conversational AI, and cognitive decision support, and the critical business outcomes those applications are intended to achieve.

I personally find it helpful to be as precise and concise about the terms as possible.

AI performs miracles for most humans on the planet as they go about their daily business by mimicking a growing set of human cognitive capabilities at greater speed and scale. In this sense, AI has many similar characteristics to other machines that have come before. Machines augment and exceed the capabilities of their human makers, and indeed that is the point of a machine.

I am not going to get into the debate on superintelligence and its associated risks. That feels like a rabbit hole I do not want to go down in this post.

I have been pushing myself, though, on why AI gets and deserves so much special attention beyond the sensational headlines about the end of human civilization or a jobs apocalypse. For me, as a strategist and problem solver, what gets me excited is AI’s ability to augment and enhance human problem-solving capabilities.

There seems to be a limitless supply of problems that human civilization and businesses need to solve. What a wonder if we can apply AI machines to get more problems solved, hopefully without creating additional ones.

What does this all mean for business?

CEOs and business leaders are typically asking five critical questions about AI:

1. What does AI actually mean for my organization?
2. What is the value from applying AI to my business?
3. How much should I invest in AI?
4. How do I apply AI at scale where it matters?
5. Where should I start, how do I accelerate value capture, and how do I mitigate risks?

Let me share my perspectives on these critical CEO questions.

The business impact of AI is significant for companies that truly harness it

As I write this post, the ten most valuable companies in the world by market capitalization are Apple, Microsoft, Saudi Aramco, Alphabet, Amazon, Nvidia, Tesla, Berkshire Hathaway, Meta, and TSMC.

Notwithstanding the fact that Berkshire Hathaway’s valuation is partially driven by its substantial holding of Apple, eight of the ten most valuable companies in the world are either leaders in harnessing AI or, in the case of the chip companies, supply critical components such as the GPUs that power AI.

The top eight companies have collectively created over $10 trillion of market capitalization growth in the last decade. While it is not fair to fully attribute this to the impact of AI, that is an astronomical amount of value creation, roughly equivalent to half of US GDP.

So what about the rest of the economy? What happens if you are not a big tech player or an AI start-up?

We need to look at this question from several angles, comparing the large top-down estimates of impact with the grassroots experiences of non-tech companies trying to realize the potential of AI in their businesses.

Many industry analysts have produced top-down estimates of the annual impact of AI in the range of $4 trillion to $6 trillion globally, or around 5% of global GDP. With the advent of generative AI, we are seeing these estimates increase, and the logic behind that feels sound. Generative AI will allow more work to be addressed, so there should be more impact. Given how many unsolved problems there are in society and business, even a 5% lift in productivity can feel modest.

The problem is that historic forecasts on the impact and size of the AI market have often turned out not to be true. My friend Dr. Jeffrey Funk is an authority on this, and I recommend reading some of his papers on the topic if you want to know more.

So why did the forecasts fall short, and where is there hope for fundamental value creation?

On the negative side of the ledger, some of the big quests, such as my favorite one, autonomous driving, have sucked in huge amounts of capital, more than $100 billion by some estimates, but remain perpetually five to ten years away. I will explore why this is the case in another post. At a smaller scale, enterprises often struggle with AI application, and I have read recent work showing that the average return on enterprise AI initiatives is probably below the cost of capital.

Turning to the positive side of the ledger, when you get a targeted and successful application of AI, the results can be nothing short of spectacular. In fact, one could say that effectively applying AI to a business is one of the few legal ways available to print money.

For example, a mid-sized US retailer used cognitive voice agents to improve the sales process. The business case was compelling: a $20 million investment, payback in three months, and a 10x return on investment in 12 months.

This case is also a representative example of impactfully applying AI to transform a business by driving increased sales, increased employee capacity, and improved employee and customer experience.

The top-down estimates of the impact of AI on business are almost certainly conservative. Companies should probably be doing ten times more impactful application of AI than they are actually doing today.

So why are they not? What drives the disconnect?

In a recent McKinsey survey, CEOs were asked how much impact to their profit they could attribute to AI. The answer was around 5%. Why are most companies still barely scratching the surface in harnessing AI for business impact? Around 70% of companies have some kind of AI strategy and are deploying AI, yet only around 10% of companies report getting tangible value.

Remember those piles of burning cash? They are real.

Getting back to the mid-sized retail example, the executive team was initially hesitant too. I was intrigued that any company would not move at utmost speed around such a no-brainer business case. It turned out that the business case was not the issue at all.

The biggest issue was concern about the impact on the human experience, both employee and customer, and consequently on the brand. The AI needed to be welcomed by employees and customers as a positive.

In retail in the US, there are over 600,000 open job vacancies, so anything that frees up capacity and creates a better employee experience is critical. Customers also benefit from higher levels of service, either from the AI or from freed-up employees.

How about other companies? What gets in their way?

There are three common challenges that tend to inhibit transformational impact from applying AI at scale.

First, AI initiatives often lack alignment with the business strategy. This leads either to AI with vague aspirations or AI carried out more as a science experiment than as a business transformation.

Second, the business value is often insufficiently articulated. This prevents organizational focus on what is most valuable and makes it difficult to secure the investments needed for success.

Third, companies often focus narrowly on models and tools, rather than putting in place the fully integrated set of components and capabilities required to drive business impact at scale, including culture, operating model, and skills.

So what can CEOs and business leaders actually do about this? How can they effectively harness AI and capture outsize rewards?

There are three things every leader should do.

1. Forget AI, at least for a moment, and focus on the North Star

It is important not to create an AI strategy that only chases shiny objects that do “cool” things but fails to deliver tangible impact at scale.

CEOs often ask me, “How can I apply machine learning to my business?” That is often the wrong question, or at least a misdirection. It can lead to an AI hammer looking for a nail, or to random acts of AI.

AI is far too scarce and precious to apply randomly to any business.

The first thing CEOs should do is to forget about AI for a moment and refocus their AI efforts on the North Star of their business strategy. They need to decide which critical business problems and opportunities AI can make a significant contribution to.

For example, a mid-tier bank was seeking to drive geographic growth. The starting point was going back to the business strategy. The bank had scale in only two of its markets and decided that AI could help mitigate its scale challenges in smaller markets. Applying AI to improve efficiency and effectiveness in areas such as anti-money laundering was a no-brainer and created a robust business rationale for deploying AI in support of the growth strategy.

Finally, few, if any, businesses can cost-cut their way to glory. Yet so many AI initiatives appear to be singularly focused on efficiency. That is like trying to play a symphony using only one musical note.

It is critical that AI initiatives move beyond a narrow cost focus and strike an effective balance between growth, innovation, cost, and risk mitigation objectives.

2. Base the approach to AI on the company’s unique starting point, assets, and legacies

The stories from digital-native and AI-native companies are so compelling and seductive that many CEOs respond by saying they want to become tech companies. To a certain degree, they are right.

Large companies will need to become adept at developing and deploying scaled digital, data, and AI products to remain relevant and survive. However, investors are betting against incumbents’ ability to reinvent themselves in this way.

Look, for example, at Tesla, one of the ten most valuable companies in the world and a company worth a multiple of the automotive incumbents it is disrupting. This phenomenon is true not just in automotive, but in practically every industry where there is technology-driven disruption.

Digital-native and technology-disruptor companies have the luxury of developing their business models unconstrained by legacy, optimizing them to fully leverage technology. Many disruptor businesses are engineered around AI rather than human workers. They are also swimming in the data that powers AI. In fact, many of their business models could not exist without large amounts of data and AI workers.

It is highly challenging, to say the least, for incumbents to follow disruptors step by step and beat them at their own game.

Urban planning is a good metaphor. Rome cannot easily become Shanghai. Rome and Shanghai have different starting points. But many aspects of a modern city, such as highways, mass transport systems, fast mobile broadband, and contactless payments, are as beneficial in Rome as they are in Shanghai.

Successful incumbents will also find ways to partner with disruptors. For example, a large European bank created a marketplace for fintechs to securely provide value-added, AI-enabled products and services to its customer base. The bank brought its customer base and associated data to the table. The fintechs brought more innovative ideas and execution speed than any incumbent bank could possibly muster.

Incumbents will also be forced to master the human side of the human-technology partnership first, if for no other reason than they have many individuals in their organizations who will need to work effectively with AI. The ultimate real scarcity, and therefore advantage, is how to harness this human talent in concert with technology.

3. Remember that losing 20 pounds is not the same as buying a treadmill

I wish losing weight and getting fit were the same as buying a treadmill. If it were, I would surely have the physique of an Olympic athlete. However, most treadmills, like mine, gather dust in the corner of a spare room. If you are like me, you also have to stop drinking the milkshakes.

The latest AI shiny object can have the same allure and promise as a treadmill, or even a slimming pill, promising outsize gains with little effort. Unfortunately, the same rules that apply to human bodies also apply to human organizations.

To succeed, it is necessary to adopt a human-centric approach. It may sound counterintuitive, but value capture with AI is not primarily a technology challenge. It is just as much a human challenge.

AI requires the right people and processes to succeed. I often call this the “surround sound” that needs to be put in place to ensure that an integrated transformation can actually drive the expected value and steer organizations away from the rocks.

For example, a health insurer recently worked on fraud detection algorithms within its claims department.

This is a great use case. AI can help harness the power of data to deliver superior insights. However, insights only have value if humans trust them and know how to act on them.

In this case, the claims assessors did not know what to do when the algorithm returned a fraud flag. They did not trust the black-box algorithm to second-guess them, and they had no incentive to follow up on claims flagged by the new AI tool.

Not surprisingly, no value was delivered. What could have been a worthy application of AI was disbanded, the chief data officer was fired, and the organization became discouraged and disillusioned with AI.

It does not have to be this way.

I love it when someone comes to me with a novel and valuable AI use case, but I always make sure to clarify the value realization logic: what the better insight will be, who must make better decisions based on those better insights, and what it will take to ensure that those decisions are actually made.

The end-to-end value realization process must be thought through and addressed to avoid unexpected and negative consequences.

Finally, as AI adoption increases, the notion of AI for good, and AI that is transparent, unbiased, trustworthy, and explainable, will become an even more critical license to operate for business. Leading companies should get ahead of this before being forced into compliance by regulations such as the EU AI Act.

Final thoughts

AI and its application to business is such a fast-moving field that I fear the half-life of my predictions must necessarily be very short.

I will produce a separate, and now obligatory, perspective on generative AI. Quick plot spoiler: my perspective is that generative AI is a novel and valuable technique that will typically be applied in an ensemble with existing AI. It will allow organizations to go further, potentially with more flexibility and speed given its adaptability.

At-scale adoption will be predicated on attractive business cases. Given that in most cases generative AI will be layered on top of existing AI, we will need to look at incremental dollars in versus incremental dollars out, something that is not routinely measured.

As we read every day in our news feeds, generative AI can potentially get you into trouble, or even jail, quickly if you do not know what you are doing with it.

I am an aviator, and you do not go flying a plane without a good understanding of how it works and what its limitations are. I think the same approach to AI application would be wise.

I would like to leave you with a thought from one of my favorite philosophers, Karl Popper:

“Our knowledge can only be finite, while our ignorance must necessarily be infinite.”

I have always found this a humbling reminder not to become overconfident in what I think I know.

Equally, what applies to human intelligence must also apply to human-created artificial intelligence. While I am hugely optimistic about the largely untapped potential of applying AI to business, I hope we all avoid becoming too dazzled by the next AI shiny object. In the process, we must not neglect to realize and respect AI’s limitations. There is a ditch on either side of the road that we must avoid.

I have enjoyed sharing my perspectives on the future of AI in business and look forward to hearing from you.