AI is top of mind for all industries. How do you see the future developing?

What we have seen so far is the digitisation of the B2C (business-to-consumer) sector. Think books, travel, movies. Google to Netflix and Amazon, which have transformed the lives of consumers. The next frontier is digitising of B2B (business-to-business) industrial companies — tractors to automobiles and aircraft engines.

The fundamental difference in what we have been able to digitise so far in the B2C sector is what I would call pure information sectors —like Google, or physical products, say cameras, which have disappeared and been overtaken by mobiles. In B2B industrial companies, the physical product, like a tractor, will never disappear. But the value is going to migrate from the physical product to data and AI. Therefore, this is the most exciting frontier, because the total GDP in the world is $100 trillion, of which only 25 per cent is in the B2C sector, which has been digitised. And digitising this $25 trillion, think about the value we have unleashed.

Imagine if we digitise the remaining $75 trillion. Why has that not happened? There are three important reasons. You can digitise the B2C sector with one technology, through the mobile phone. But you cannot digitise a tractor with a mobile. You need sensors, computer vision and IoT. All those technologies are there, but their cost is prohibitive and it’s going to take time for the cost to come down.

Tesla has shown we can digitise automobiles. But lots of other sectors are yet to be digitised. The second reason is that in the B2C sector, 80 per cent accurate recommendation is good enough. Suppose Amazon gives you a recommendation for 10 books, and you liked only eight. It may be inconvenient, but you can live with it. But an airline cannot get 80 per cent recommendations from Rolls-Royce, because when you’re flying the plane miles high up, it would be disastrous. The third reason is that traditional AI models were capable of analysing only structured data. While industrial sectors generate structured data also, the bulk of the data they generate is unstructured and qualitative — say, an image or the sound of a machine.

But GenAI is exactly what the doctor ordered. When you put all these three together, one can develop large language models for the industry that are 100 per cent accurate.

Will the AI revolution be as seminal as, say, the internet revolution was in the early 2000s, or will it be even more of a game changer?

Without a question. It will be a thousand times more transformative than the internet. You see, we have seen three fundamental inflection points in technology. Web architecture was the first. It just decentralised information. Second is the development of apps. However, apps are fragmented — different apps for different uses. The whole agentic era we are entering into, it is able to learn across these various agents and then solve your problem. It can coordinate all the fragmented sources of information: some may be in my wearables, some may be in my finances or, say, information from a hospital. It is able to coordinate all of that and give me real-time recommendations.

Therefore, this is huge. This is bigger than the steam engine or the industrial revolution. This is definitely bigger than the decentralised web and the app architecture.

Is AI a bubble or is it for real?

A lot of people are asking if we are seeing an AI bubble, since we have spent, and continue to spend billions. AI has completed five years now, and what we have done is to build the infrastructure. Because all the large language models are essentially infrastructure. The next five years will tell whether it is a bubble or not. Only when companies build applications will it create value, and the investment in AI will pay off. Again, go back to the internet. We spent a ton of money on developing it. But the value was created by Netflix, Amazon, Airbnb, Uber… That has not happened yet in AI. So the next five years are important. If AI has to pay off, then leaders of companies have to step up and figure out intelligent ways by which to use the infrastructure to deliver more value for their customers. Then AI will pay off.

AI is moving at warp speed. Whereas you have a generation of managers who learnt in an era where things remained static. Accounting or marketing principles would only change incrementally. But now, things are moving fast. Is academia retraining itself?

Let’s talk about higher education. I believe AI is going to transform all industries. Even the current capabilities of GenAI can transform higher education, not to speak of how rapidly this technology is improving. We need to intelligently use AI to deliver more education for our students. Therefore, the point of view that I would like to advocate is, in education we must have a bifurcated approach to AI. By that I mean there should be some parts of our education that should be an AI-free zone, which is teaching the foundational knowledge. If you’re a computer science major, you should know how to code. Even though GenAI can do the coding, you must teach them how to code. It’s very important. Similarly, in a writing class, you should write by yourself. If you don’t do that, I think you are losing a very valuable skill. I think we should very generously use AI, but as professors we can still add a lot of value to students. Because things like judgement, persuasion, communication, connecting the dots… AI cannot do.

I’ll give you a quick example. I’m going to do an experiment where I put the assignment questions for my classes on ChatGPT and give the students ChatGPT’s responses. Read that, think about it and come to class. If I cannot add value for 90 minutes over what ChatGPT told you, you don’t need me. If you substitute the thinking for ChatGPT, I will expose you in five seconds because your learning is superficial. The point that I’m saying is ChatGPT is an information processing tool, but humans have tremendous ability in judgement, making choices, persuading people to accept your choice.

A foundation skill is knowing how to add. Recently, I went to a shop in the US. I was buying five shirts — each cost $55. On that day, the cash register broke down. So the guy behind the cash register, his brain stopped working. Because, from the beginning, he was given a calculator. So he never knew how to add and took a sheet of paper and was slowly adding $55 plus $55. I told him it is $275. He kind of looked at me, but then went back to his paper and pencil, slowly adding. After five minutes, it came to $275. And he looked at me and said, ‘you must be a PhD from Harvard’. I said, I am! But that’s beside the point. He didn’t know the basic logic of addition and multiplication.

Therefore, this is what I’m saying. We must separate foundational skills. Let them use GenAI for information processing. But there is a higher level of intelligence that humans have, which are AI-resistant skills, that only humans can do. Therefore, I would say higher education should embrace this bifurcated thinking in the use of AI.

Do you see companies retraining employees to use GenAI?

The companies which are leaders are absolutely doing it. A Honeywell, a John Deere, a Siemens are retraining their employees to be digital-first. It is very important to be digital-first and digital-savvy, absolutely. But not all companies are doing it. Therefore, these are the companies that are going to take the lead.