Two Realities: Lessons from History on AI Implementation

_{My day job gives me the opportunity to speak with many leaders in communications & marketing, and with multiple experts in the field of AI. You are somebody that I’ve previously had engaging interactions with on the topic of AI - my goal with this newsletter is to share the thoughts and insights I gain from these conversations. This time:}

When it comes to AI, we face two conflicting realities. First, the challenges of implementing AI often seem to outweigh the advantages that proponents like myself frequently highlight. Yet, history reminds us that the risks of ignoring technological advancements are very real, regardless of our preferences.

I received a lot of thoughtful feedback on my last post, where I argued in IMD magazine for a more fundamental and radical approach to how AI can transform businesses.

Several readers highlighted a crucial point: immediate, practical concerns frequently take precedence over the broader theoretical concepts. As one reader pointed out:

Others noted that the prevailing reality is that most companies default to maintaining the status quo. New technologies are often seen not as opportunities but are instead viewed through the lens of today’s problems:

Some shared their experiences in more stark terms:

Recent studies support these observations, showing that early experimentation with AI is not translating into significant business benefits despite the potential - the media currently enjoys reporting on studies with headlines such as “74% of Companies Struggle to Achieve and Scale Value through AI” . 1 2

The Reality of Business Longevity

It is sobering to consider that the average lifespan of an S&P 500 company is only 21 years. Great companies rise and fall more frequently than we might think, often overtaken by shifts in competition, technology and customer preferences. The AI transition is one of those forks in the road that will significantly impact this 21-year average.

This may be a slight digression, but bear with me while I try to illustrate the above point. From Bloomberg last week:

While much of the commentary has focused on the progress of the four-year turnaround Gelsinger embarked on after becoming CEO in 2021, the critical decisions and lack of foresight that led to Intel's precarious situation were made long before.

In early 2020, Intel canceled its Nervana AI processor project, which would have directly competed with Nvidia in the AI data center market. Even earlier, in 2010, Intel shut down its Larrabee project, aimed at competing with Nvidia in the GPU market. These decisions effectively excluded Intel from today's most significant growth areas in the chip industry.

In 2005, Intel missed another major industry shift when Apple chose the competing ARM architecture for the iPhone. Missing out on the high-volume mobile market made crucial investments in chip manufacturing less viable, causing Intel to lose industry leadership to competitors.

Ironically, ARM was developed in the mid-1980s after Intel refused to license their 80286 architecture to Acorn Computers.

The lesson here is that you cannot catch up with exponentials, and that choosing the wrong fork on a path that has an exponential on one side will doom a business.

As the cost of AI continues to fall and its capabilities grow through advancements like agent-based AI, the marginal cost of intelligence—one of today's major organizational expenses—is rapidly approaching zero. We are clearly at an exponential fork in the road.

Back to Today

The challenge today is that even if companies are eager to implement AI, the process is far from straightforward. The AI landscape is complex and crowded with competing voices and claims. A persistent issue is the lack of necessary skills, which is often cited as an insurmountable barrier.

Think back to how companies approached the Internet 20 years ago. From a 2004 observation:

Many teams are taking a similar approach to AI today, or worse, delegating it to their IT departments — an equally misguided strategy for managing an exponential technology.

Learning From Past Disruption in Software

In the early '90s, Microsoft’s Bill Gates faced the relentless challenge of technological disruption in the software industry. The tech world was undergoing dramatic changes with the rise of personal computing, the Internet, and the integration of technology into daily life.

Gates maintained two lists. The first was a list of the most pressing strategic challenges, ranked in order of priority.  The second was a list of the most capable and experienced people again ranked in order.

In many cases (then as well as in every organization today) the skills needed to solve problems at the top of the first list would best match people towards the bottom of the second list. This mismatch explains why websites used to be run by interns and why AI projects are often delegated to juniors.

The reason is straightforward: experienced people often stake their value on past achievements and double down on the associated skills and techniques. Beyond a certain career point, they are inherently opposed to learning entirely new things, as they perceive they have more existing value to lose than new value to gain.

This leads to an organization where value creation is fundamentally misaligned. Or upside down.

Gates' solution was to be acutely aware of this dynamic when making personnel decisions. He aggressively sourced talent from diverse industries, academia, and wherever Microsoft could find individuals who matched the strategic issues at the top of the first list, rather than merely using the closest match from the existing pool or assuming that experience = innovation.

Since the advent of the Internet, most technological changes have been incremental, with exponential shifts confined to specific industries or functions. While exponential change is often discussed, it is rarely experienced—few of today’s leaders have encountered anything akin to Microsoft’s challenges in the '90s. Exponentials and transformation have become something we continuously talk about with little experience or success in navigating.

However, it is difficult to imagine a scenario where scalable intelligence does not have a truly transformative impact across a broad range of industries and job functions.

We are going to have to deal with it.

A Ray of Hope?

So what’s the takeaway from this exploration of the past and future, especially given the very real skills and time challenges of today?

• Delegation of AI challenges is a real and present risk

• Hiring diverse talent (rather than promoting from within) becomes crucial

• Defaulting to old ways can be fatal - you can’t catch up with exponentials

Admittedly, this perspective may seem a bit pessimistic. So let’s consider the counterpoints and the potential AI offers in each of those terms:

• AI provides the potential to delegate much of today’s time consuming non value added work so that we can focus on the strategic issues

• Intelligence at scale can solve many of today’s talent bottlenecks

• We have the opportunity to leave the worst of the old ways behind

There is much more potential to be gained than lost. We just need to keep this in mind as we face the tough decisions of the coming years. It’s an exciting future!

Thanks for listening to me play out my thoughts,

Mark.