What Leaders Need to Know About Strategy in the Age of AI

What do business leaders need to know about AI now and in the future?

AI will have a huge impact not only on strategy but on everything. That means that leaders need to get up to speed and become what I call hybrid strategists, who understand both technology and business. They also need to hire people like that and experiment with new technologies, processes, and business models. This is the time to increase your research and development budget so that you can try more AI initiatives.

Whenever there's a new technology, new opportunities emerge, and it's up to firms to discover how to manage them. That discovery means you need to explore, research, and hire people who can do that. Learn how to do it yourself. You need to get your hands dirty.

The largest companies today are almost all tech companies. All of those CEOs were trying things and gaining a deep understanding of the technology and its possibilities. There will be a new set of large companies given AI, and in 30 years, who will be the leaders of those companies? It will be people who understood the technology, who experimented with it, discovered the right way to use it, and saw that before others. You need to do that today in order to be a leader tomorrow.

What first convinced you that LLMs could perform some strategic tasks at a level comparable to humans?

In 2022, with the new round of AI developments, I started seeing that LLMs could do things that are useful in strategy. Maybe, I thought, AIs can do some of the things that strategists do. I thought, ‘Let's test that,' and I wrote a paper that was published not long after.

In that paper, there were two experiments. In one, we had real business plans that people sent to a business plan competition. We took only the first paragraph that explains the problem, deleted everything else, and asked ChatGPT to complete the business plan. So essentially, create a business plan that solves this problem, then enter the real and AI-generated plans into a business plan competition and see how they rank. It turns out that the AI-generated business plans ended up slightly above those from humans.

The other experiment compared how venture capitalists evaluated data from another business plan competition. We gave them real business plans. We have how the venture capitalists evaluated each one of the business plans according to 10 dimensions, and then we give the same business plans and the same rubric to another AI. We asked the AI to evaluate the same business plan and compared its evaluations with those of the venture capitalists. They are very similar, and evaluating a business plan is a complicated task. That paper showed that, at least in these two tasks that we experimented with, the AI performs similarly to humans.

In a more recent working paper, we gave AI questions we don't yet know the answers to. There are companies trying to raise money on Kickstarter, and we asked the AI to predict which will be most successful and raise the most money. We have 30 companies, and the AI will have to rank them. We will do the same with humans, then wait until all the campaigns close, and see who was more accurate.

The Kickstarter experiment is very interesting because the AI has seen essentially everything written about a topic and more previous Kickstarter projects than any human. It has a much more detailed understanding of Kickstarter success. The AI is likely more consistent than a person. It will read the whole thing. A person may be hungry, rushed, or having a bad day. All of those sources of inconsistency are not a thing for an AI. The AI will read the material very quickly and thoroughly, whereas a human may skip parts. We’re looking at AI-human comparisons in knowledge, consistency, and attention.

You argue that many processes were designed to accommodate human cognitive limits. Does the rise of AI mean these frameworks are obsolete, or do they still have a place in this era?

I have two answers for that. One is that humans will continue to have those limits. Our brain is not going to be upgraded suddenly. We are stuck with our brains. Whenever you need to communicate, to explain, to coordinate with others, anything that involves humans will continue to use simple models, because it needs to fit within our minds. But then, an interesting question is: What will happen with models that don't need to fit within our minds, models that are just for AI consumption?

My guess would be that those models would be more complicated. In fact, it already happens in all predictive models. For example, Netflix recommends a movie that you should watch next. Or when the bank predicts whether to approve a loan. The model that makes that prediction is one that a computer is running. There's no human involvement there, and those models are typically incredibly complex. So I would imagine the same could happen in strategy: models that only computers can run will be much more complicated than models that you need humans to run or understand.

If the underlying AI models are widely available, where can companies still build a durable competitive advantage? 

The moat is not the AI. It's the same as companies hiring equally intelligent people. Multiple companies, let's say, can hire Ross MBAs. But the company's advantage will come from the interaction of two things: hiring good people and using good AI, in conjunction with other factors. For example, to predict what movie you want to watch, it's not just enough to have an AI.

You need to have a database of all the movies that many people have watched and the movies that you have watched, so that you can be grouped with similar people, and then the AI will be able to say, given this similarity, a movie that you haven't seen that you may want to see is this one. If you don't have the data on movies that people watch, and what you have watched, even if you have the best AI in the world, you won't be able to make very good predictions. That's just one example that it's not all about the AI.

AI requires other things, and this is a typical example. Data is something only available to some companies. Let's say an AI could be very useful with emails. It can respond to an email using all of your context, knowing when you have talked before with that person, what you have written to that person, and who that person is. Gmail can do that because it has all of your email history, but another company without access to your conversations and past interactions won't be able to replicate it. Even with a good AI, if you don't have the data, you don't have a business.

Will proprietary data become an increasingly important source of competitive advantage?

Data and networks are major sources of moats. Think about communities. You might use Facebook or Instagram, or other social platforms. Some other app could be equally as good, but if no one else in the world uses it, you wouldn’t want to use it.

Networks will continue to be big moats, regardless of AI. AI will be able to do things on top of it, like giving you better recommendations on what to watch, who to talk to, or what content to add to the network. But you still need the network, and it will be the moat. For other companies, the moat may be the data they have accumulated over their history.

For example, if you are a consulting company, you have your database of all your past projects. All that information can be used by an AI to create better proposals and better solutions for future projects. Everybody may have the same access to ChatGPT, but not everybody will have the same access to data and to users. That data is important, and we haven’t even talked about the complements to that, like brand awareness, factories for production, or a footprint.

Will large companies with this data at their disposal have a major advantage over new or smaller companies? 

Yes, but at the same time, many businesses haven't been invented yet. So in those, no one has the right data or the right product for that. And I think AI will be able to help the new entrants there. An entrepreneur with an AI is like an entrepreneur with a whole team of people, so they can come up with more ideas and implement them much faster than before. I imagine that AI will be helping entrepreneurs a lot.

How and why will large companies decide that AI is a necessary part of their future success?

I think the main pressure will be competition, so a big firm will see a smaller firm doing better and growing, and they will have to adapt or disappear. Competition will be the main propeller of change. The other thing that will start happening is that companies will help you with this transition. For example, there could be consulting firms that specialize in helping large firms use AI more effectively.

Things will become more like products that are easier for large companies to adopt. Also, more managers will deeply understand AI, and companies will hire them. So that's what I hope our role is as educators: to educate the next generation of managers so they are very proficient and creative in how they use AI in the firms they found or work for.

How significant a change do you think AI represents for business? 

It’s maybe the most important thing that we have seen in our lives in business. The internet was a big one, but I think that this is bigger. Before that, it was the PC. You see, all processes changed in organizations because of computers, and then because of the internet. But I think AI will be more transformative because, at the end of the day, anywhere there's cognition, or you can add cognition, it will be touched by AI.

How should companies decide when to apply AI and when to rely on humans? 

I think it will be different for different decisions. I think the decisions that are easier to automate are those for which you have a lot of data, so that you can do the back testing and you can see how well it could have performed in the past, and then when you release it out there, you have some very accurate prediction about how well it's going to go.

For things where you don't have very detailed data, you will need more human involvement. It's super important to go decision by decision, learning and checking. It will be a trial-and-error process. Having a human in the loop will be more important in those areas where you don't have that much data, and you cannot easily verify AI’s answers in an automatic way.

How will AI impact the role of humans in developing and implementing business strategy?

I don't think that AI will replace strategists in the foreseeable future. AI will complement strategists, and strategists will do analyses they didn't do before, at a much larger scale, maybe much more creatively, across many more conditions and cases. Currently, managers can focus on strategy when they have some available time to think, which is maybe 5% of their time. Now, you can let an AI think 100% of its time, or multiple AIs think 100% of their time about the strategy, and that could lead to a faster, deeper, or more creative strategy process.

One could think of an AI as being analogous to a consulting company. When a decision needs to be made, both can prepare alternatives for the manager. That doesn’t mean the manager stops thinking. The consulting company will spend a few months developing alternatives after interviewing people and collecting data, and then give you their three best options. They’ll provide the reasons for each option, and then you will decide. You'll have a meeting with your top management team, and you will still think about this deeply, but your starting points are those three alternatives from the consultant.

Now think about that same situation with AI. You will have AI do the first round of thinking, akin to the consulting company, coming up with the best possible alternatives, simulating them, and then giving you the best they can do. At the end of the day, because you are the responsible one, you have to make the decision. Are you willing to bet your career, your company, on this or not? You will be thinking deeply, but the starting point will be different.