In a nutshell

Given our decade-long experience using AI as part of how we serve clients,[1] we’ve been getting questions about the nature of AI and how it is likely to impact us. So we’ve written a trilogy of articles to explain the why, the how, and what to do about it as a CEO.

This is the third and final article in our series, which focuses on what CEOs should do.

Note: This article is action-oriented, and thus the shortest of the three. The background reasoning has already been explained in the previous two articles of the trilogy.

The name of the game: cash + experimentation

Parts 1 and 2 of this series showed that:

  • Whereas many jobs will likely be taken over by AI, predicting which ones is tricky
  • Large shfits in economic structures are likely to occur
  • The best practices of using AI are not currently understood

Firstly, in such an environment, cash is king. This is because having cash is the most flexible way of keeping one’s options open.

Secondly, the experimental scientific method has proven itself to be a powerful approach for learning, when one is faced with a new environment such as the post-AI world. For memory, the experimental scientific method is: observe, make a hypothesis, test that hypothesis with an experiment (seeking to falsify the hypothesis). Then rinse, repeat.

Consequently, CEOs should experiment with different uses of AI, while expecting most of the experimentation budget to result in failed experiments.

However, experimenting is not enough. The hardest part is to put in place the planning, measurement and tracking infrastructure to turn what would otherwise be spray-and-pray spending into a series of true scientific experiments.

Workbook (feel free to skip)

Cash

  1. In a bad year (the “happens once a decade” level of bad), how much cash do you need to cover your needs?
    Answer: ________
    Multiply this number by 2x: ________
    The result is your cash buffer need #1.

  2. Respond to the “Experimentation” workbook questions below. How much discretionary cash is needed to cover the resulting experiments?
    Answer: ________
    Multiply this number by 2x (to take into account cost overruns): ________
    Add on the salary of a senior trustworthy full-time champion (as, without such follow-up, the cost overruns are likely to be more than 2x): ________
    Total: ________
    The result is your cash buffer need #2.

  3. Sum your cash buffer need #1 and cash buffer need #2.
    Answer: ________
    The result is your total cash buffer need. This is the amount that you should build up, all other things being equal.

  4. Does your company have this level of cash buffer? If, not, what measures will you take to build it up?

Experimentation

  1. What are 3 areas of your company where you think AI can improve efficiency?
  2. Choose one of these areas.
  3. How do you expect that AI can help? In order to express this as a testable hypothesis, use the folllowing format:
    If we do

then the result will be

which can be measured by

This is the concept of your pilot project (experiment).
6. Is your measurement strict? If not, adjust. Being too loose, so that success and failure of the experiment are not clear, is the biggest error committed in such projects.
7. Do the people involved understand that failure is fine? Most of these experiments will fail. The people involved should understand that they will be judged on how much was learned from the experiment (which is within their control), not on whether it succeeded (which they can’t control, and which they shouldn’t control, since failure is often a desirable outcome[2]).
8. Is the entire company behind the project? (See the white papers on the firm as a system, in the methodology section.) If not, modify the concept, including appropriate alleviatory measures.
9. Identify a senior project leader with the appropriate skills, and make sure they have powerful backing. It will be their job to flesh out the concept into a complete pilot project.
10. Repeat with the two other ideas.

Surfing the chaos

From the outside, societal dislocations seem to be all of one piece: in a crisis, things are going badly. In a eucatastrophe (Tolkien’s term for a good catastrophe), things are going well.

In truth, anyone who has lived such a dislocation will confirm that, from the inside, they are not at all of one piece. In the middle of harmful events, there are many opportunities for honest gain.[3] In the middle of an economic boom, there are many risks and failures.

A key competitive advantage in such an environment, is being ahead of the crowd. This is rarely easy (the crowd is often wise!). It becomes much easier, and almost easy, in catastrophes and eucatastrophes. All that is required is keeping one’s eyes open, and having a cash buffer. Many of the people who in normal circumstances would spot and exploit opportunities faster than the rest of us, will, during a crisis, either de distracted by another risk/opportunity, or have insufficient cash.[4]

Workbook (feel free to skip)

  1. What are three business developments that you are afraid of? Or that, more generally, will have a high impact on your company?
    Example: 1 – Tariffs disrupt my supply chains. 2 – AI makes it easier for a new entrant to reverse engineer my product. 3. Rising cost of energy kills my bottom line.
  2. Pick one. What are three business advantages that you could gain if, by being attentive to your surroundings, you manage to catch it just when it begins?
    Example: let’s pick “AI makes it easier for a new entrant to reverse engineer my product.” When that starts to happen, everyone will join the bandwagon. Hmmm… These companies will all be competing, reducing the sector’s profits. But, being new companies, they’ll also need a lot of infrastructure which takes time to build and which I already have. I can pivot from selling my product to selling to the competition access to the infrastructure that they need.
  3. Same question for the other three.
    Example: let’s pick ‘Tariffs disrupt my supply chains’. If tariffs get extended to my raw materials, what is likely to happen? The cost of manufacturing my widgets will shoot through the roof. I’d have to raise prices. But I can’t do that, customers would switch to the competition. But wait, the competition would have the same problem. The winner will be whoever has the most efficient operation. So now I need to stop focusing on increasing the number of SKUs to satisfy smaller and smaller niches of the market, and instead focus on operational efficiency. But will that be enough? Probably there will be a period of consolidations. There always seems to be, after a crisis. So I’ll need to build a cash buffer for that time. If I’m lucky, I’ll be on the buying side and able to buy less efficient operations on the cheap. If I’m unlucky, at least when the buyer comes to my door I’ll be in a good negotiation position, with both a lean operation and dry powder (cash).

This is a good morning exercise to do from time to time, as a reminder to stay alert to the opportunities hiding in difficult events. Note that it doesn’t have to be about AI.

A lense for situational awareness

A pre-requisite for surfing the chaos is sutational awareness. This section proposes using lenses through which to view unfolding events and exercise situational awareness.

Geopolitical risk

All other things being equal, AI:

  • Will, in the long-term (but not the short-term) increase the geopolitical risk of disadvantaged countries
  • Will, once robotics has caught up, decrease the value of cheap labor from disadvantaged countries

Workbook (can be skipped)

The exercise here is to remain, over the coming years, mindful of these evolutions and how AI and other factors influence them.

Also, be mindful that the timing and effects of geopolitical changes are hard to predict, and that betting on them is likely to be counterproductive. Famous economist John Maynard Keynes, who was also an investor of legendary stature, lost several fortunes by investing based on his geopolitical expectations. He was completely right about the geopolitics, but he lost his investments due to timing effects.[5]

The recommendation here is: just stay sensitive to the big trends, bear in mind the possibilities, and be ready to take advantage of them if they materialize. But do not bet the company.

Cost structure of your company (a.k.a. Dutch disease)

When hydrocarbons were discovered in the territorial waters of the Netherlands, the citizens were pleased and excited. Surely this would usher in a new and golden age!

But instead of a new and golden age, what happened was that the non-hydrocarbon sectors of the economy, and in particular the manufacturing sector, were badly hurt. Indeed, the new hydrocarbon industry voraciously ate up skilled manpower and raw materials, paying higher rates than the other sectors of the economy. The other sectors were hit hard by the increasing scarcity and price of the labor and raw materials that they depended on.

This came to be known as “Dutch disease”.

With AI, we should expect to experience Dutch disease. All of the inputs that your company pays for which are also inputs to the AI sector, will become more expensive as they get hogged by the AI companies. Hence, there is value in reviewing which of your inputs are also inputs to AI companies, and therefore likely to become more expensive.

Workbook (can be skipped)

  1. Is energy an important input in your cost structure?
  2. Are processors (especially GPUs) an important input in your cost structure?
  3. Are you dependent on financial markets for funding?
  4. Are you dependent on banks (or worse, one bank) for funding?
  5. Is your firm established in a location where one of the giants has created a data center?
  6. Among the skillsets you need in your staff, are there the specialties that AI firms also require?

If the answer to any of these questions is “yes,” you may find yourself, in the future, in competition with the AI sector for resources. This would push up your costs.

For each “yes” answer from the above, answer the following questions:

  1. How can I diversify my needs? (For instance, for financing develop relationships with more banks. For employee skillsets, find ways to use a broader pallete of approaches and skillsets, so that if one type of talent becomes hard to hire, you have alternatives.)
  2. How can I increase efficiencies, to be less sensitive to such an evolution? (E.g. develop economies of scale.)
  3. How can I use AI to reduce my cost in other areas? (E.g. use AI to increase the productivity of my data center operations specialists, who are at risk of being snapped up by the AI sector.)

Epilogue: What will happen to the CEO’s job?

Certain styles of CEO management can already, as of today, be done by AIs. Consequently, CEOs will need to focus on the strengths that AI cannot emulate.

CEO styles that AIs can already emulate today:

  • Big picture thinking
  • Designing grand visions

CEO styles that AI can “fake” well enough to look like it can do better than a human CEO:

  • Designing strategies and soon, business plans
  • All skills whose impact takes time to appear, or for which it is difficult to measure whether they were successful or not

CEO styles that AIs can almost emulate today:

  • Deep-diving / Debugging issues: currently AI has a bias, when it hits a difficult problem, to try to shortcut the solution and sweep any issues under the rug. This appears to be the intention of the AI firms. The AIs are trained on purpose to behave like this. Indeed, an AI that always went to the bottom of things would result in a massive bill for its users. This would result in some very angry customers! However, this issue is not insurmountable. AI coding tools have already partially addressed this issue via breaking down requests into multiple prompts. Further development of AI will probably find even better solutions to this issue.
  • Talking to people face-to-face. Holographic AIs have already been produced[6]

CEO styles that AI cannot emulate:

  • Decision-making in areas in which it has not been trained and cannot generalize, provided success / failure feedback is swift. (If not, then, in this area, AI will be able to fake capabilities it does not have.)
  • The human touch. Caring about people. Even with holographic technology, people will continue to be sensitive to whether they are interacting with a human who cares or a human who does not care or a holographic AI – at least for the forseeable future.
  • Membership of cliques and coalitions. Unfortunately this could lead to a greater frequency of leaders who are in their position due to who they know, or politics, rather than their competence
  • Moral compass: AIs have political correction, not integrity
  • Responsibility: originally, the concept of CEO was tightly linked to taking personal responsibility for the management of the company. Over the last few decades, this aspect has been progressively deemphasized, with mechanisms such as liability insurance and compensation plans that produce mighty payoffs to CEOs who drove their companies into the wall. With the rise of AIs, we can expect that taking personal responsibility will return as a major aspect of what makes a CEO. AIs are just algorithms, and cannot take personal responsibility. CEOs who shoulder personal responsibility will be irreplaceable.

Consequently, we’d expect tomorrow’s CEOs to tend to fall into the following categories:

  • T-shaped: this sort of CEO is the “universal man” that came out of the medieval cultural renewal. He knows a lot about a lot of topics, and his knowledge is not superficial. In addition, he also has a very, very deep understanding of one topic. Thus, he combines breadth with depth. This is the ideal profile for being able to do the sort of generalizations that AI is by construction going to find challenging (c.f. The Limits of AI for more on these limitations.)
  • Members of power networks
  • People with outstanding people skills
  • People placed in a position due to their moral compass. For instance, when creating a governance system (be it that of a company, or that of an entire regulated sector), much can be accomplished by digitalization. This digitalization forces participants to follow specific procedures, thus cutting out a lot of the opportunity for corruption. However, such a system is only as good as the people who design, maintain and enforce it. A person of unimpeachable integrity will still be needed at the top.
  • People who take personal responsibility

What about employees?

The wise CEO will try to find ways (e.g. training programs, insurance plans, mental health programs to alleviate the effects that AI can have on employees and their loved ones, etc.) to care for his employees during the AI transition.

This is due to three reasons, of which two are invalid and one is valid:

  • The Caring CEO is one of the types of CEO which AI cannot replace
  • Howevermuch importance AI ends up taking, the people who make up organizations will remain key to its success.
  • It’s the right thing to do

The first two reasons, while true, are invalid, in the sense that:

  • The Caring CEO will do it because of the third reason, not because he is putting on an act. AI will soon be able to put on a better act than any of us.
  • The CEO who mechanically takes care of his employees is also doing something which AI can do better than him

Thus, only the third reason, doing it because it’s right, gives the CEO a key competitive advantage. Which, paradoxically, is not the reason he is doing it.

  1. Example 1: in the context of the acquisition and turnaround of a near-bankrupt company, cash flow was the most important consideration. As a result, the ability to peer forward into the future, even if only by a few days, in order to predict future revenue, constituted a decisive advantage. We achieved this using AI.
    Example 2. In the context of negotiations with our client’s suppliers, we faced a difficulty. There weren’t all that many alternatives to the suppliers that we were negotiating with. In order to strengthen our negotiation position, we needed to find more potential suppliers capable of manufacturing the same products, but who were not advertising the fact. We used AI to identify other factories which had such capabilities.
    Example 3. During the COVID pandemic, we created a system to dynamically update the relevant policies, based on a cost-benefits analysis that was updated in real-time to take into account the latest incoming data (such as covid cases inthe organization and where they appeared).
    ↩︎

  2. It would be a pity for a bad idea to succeed in the pilot phase, resulting in its being rolled out to the whole company. ↩︎

  3. Example: when Lebanon’s economic collapse led to a collapse of the already atrocious electric infrastructure, a booming solar power industry emerged. ↩︎

  4. Example: as of this writing, the price of gold is quite volatile. The futures price is higher than today’s price, which reflects a market consensus that the price of gold is likely to rise steeply. But then, isn’t this an opportunity to make a quick, easy profit? No, for several reasons, one of which is that those who have easy access to this market already have their cash fully allocated. (Please do not misinterpret this to be saying that gold is a good, safe investment. Gold is an interesting but risky investment, due to factors such as duration mismatch. We are not giving investment advice, here.) ↩︎

  5. Especially on FX. ↩︎

  6. See here for dated examples. In the 2025 edition, one of the presenters was an impossibly beautiful holographic anime heroine. ↩︎

Leave a comment