Is there an AI opportunity for Egypt?

Patrick FitzPatrick [P.F]: Gentlemen, I’m going to ask you to invent a time machine and warp to 2035. Not 2030. Kefaya 2030, we’ve all had enough of a Vision 2030 and a 2030 plan and yada yada yada. It’s barely three years away, so you’re in a time machine to 2035. 

We’re going to start with you, Ahmed. What is one thing that you want to have seen us build or do, that makes us regionally or globally competitive in the field of AI, that we need to start now?

Because if we sit here and twiddle our thumbs until 2033, it’s not going to happen. 

What is one big thing you think we should do?

Ahmed Ali Abdelrahman [A.A]: First of all, thank you very much, Patrick, for having me today. I think what we’re missing, basically, is that our digital transformation at scale wasn’t ready for AI age. If you look at most of the private companies or even the large corporates, I think everybody’s lagging in a different way in having ERP [Enterprise Resource Planning] systems in place, which basically is the foundation for an AI era, whereby you have to have the workflow fully digital.

You have [to have] the data sets in place and be able to basically leverage this. I think this is what we’ve missed. And I think it’s a common issue across all sectors.

P.F: Sherif, if you’re going to pick one thing, one target we should hit on that map in 2035, what is it?

Sherif Kamel [S.K]: First of all, thank you, Patrick, for having me. It’s really [a] pleasure to be here. If I want to go ten years ahead, allow me in 30 seconds to go back 56 years.

P.F: 56 years? All right.

S.K: And precisely I’m saying 56 years back because someone — many of us and others have used a term that he wrote 56 years ago. We repeat it all the time. 

And that is, “The illiterate of the 21st century will not be those who cannot read or write, but those who cannot learn, unlearn and relearn.”

Who’s that person? His name is Alvin Toffler. Together with his wife, Heidi Toffler, they wrote three books. One of them is called Future Shock, and he was talking about the fast change and how communities are going to be able to adapt to that change. That was in 1970. The second book was in 1980, that’s 46 years ago, and it was called The Third Wave. He was referring to the Agricultural Revolution, the Industrial Revolution, and now the information revolution and how it’s going to transform our life. The latest was 1990, and he called it Powershift, where the strength of society would not be based on the traditional resources, but on those who have control of the data, of the information and of the knowledge.

Again, these three books were written in 1970, 1980, 1990. It’s exactly what is happening today. So if I move forward nine years to 2035, the one thing that is the elephant in the room, that is extremely important in the readiness of the ecosystem of AI, is the human capital. 

Will we be ready? Because I’ve listened very carefully to every single speaker today. They all spoke fluently and accurately about the prospects. Many spoke about the infrastructure, which is extremely important. I’m going to come to that maybe later on in the conversation, but the key building block is: are we ready as human beings to make a difference?

P.F: So we need to invest in people. Bassem, what about you?

Bassem Fayek [B.F]: Kind of similar. I was going to focus on education, and I like the fact that you’re taking that long-term view, because when you take that long-term view, we start to think a bit more strategically: “Okay, if we really have to do one thing, what is that one thing?” I would say really, really to reinforce how our education system is enabling the next wave of talent to adapt to whatever hits us in our faces.

I look, for example, at one of the recent initiatives by one of the GCC countries. I think it was the UAE, who just introduced their AI curriculum, which is, in a way, to help the next generation of talent to become AI-ready. And I think that is something that does not necessarily require the huge investment [we] think about on technology or infrastructure.

It’s kind of a no-brainer, and it also plays to our advantage when we think about, “Okay, what is Egypt’s advantage today?” We say, “Okay, we have maybe some good ambition,” but where we can differentiate, or are already differentiating, is on talent. Let’s double down on that.

P.F: I’m going to go off script for a minute. I’m going to throw a question to Sherif, because both of you guys have hit on talent. 

Ahmed, you hit on readiness [at] the corporate level, which depends on people. It’s all about people at the end of the day. And I want to put a difficult question to you [Sherif]. 

My daughter called me at 2:00 in the morning last March, crying her eyes out, because this child, who hates AI, who won’t use AI, who’s been told by the University of Toronto, “If you use AI, you will get kicked out of university. You’ll fail the course and you’re gone,” submitted an essay to an AI detection service that said this is 85% AI. She wrote every word herself. “What do I do, Dad?”

How are you guys using AI to help prepare the next generation of leaders that we’re all going to employ? I was speaking with Kim [Fox] here in the audience earlier this morning. Our entire audio team [was] trained by her department. You go across our journalists, half of them have done a degree somewhere at AUC. 

How are you guys as a school, and you specifically at the [Onsi] Sawiris School [of business], using AI to prepare the next generation of leaders? Are they allowed to use it?

S.K: Absolutely. And I think the person who told your daughter, “You should not be using AI, otherwise you get kicked out,” is in trouble, and I mean it. Because that will mean that the students, those who had the opportunity to get exposed to a large amount of data, at the end of the day will be analytical, in a sense, to read it, criticize it, analyze it, and then share whatever thoughts they have, and obviously mention the references that they use. If they don’t do that, they are lagging behind big time.

In fact, if we go back just less than four years, this whole phase of Gen AI started in November 2022. So we’re not even [at] four years. Imagine the development that happened over the past four years. And just imagine 2035, as you mentioned. Where will it be? That would be a completely different planet, right?

So going back to your question, I want to just answer it in two different ways. I look at AI as an ecosystem where — again, I’m not going to repeat myself — the most important element of it is people. But then comes an extremely important element, which is the data. What kind of data are we using? Is it relevant? Is it timely? Is it efficient? Is it on point? And then the third, which in many ways we don’t think about often, is the legal and regulatory environment, the governance model. And then fourth — and because it’s number four doesn’t mean that it’s not important — comes the infrastructure.

But we tend all the time to get attracted more to the infrastructure. Do we have the latest technology? Maybe yes, maybe no. But can we use it? Do we have the proper governance model? Do we have the data? So that’s the model that should be taught.

Going back to your specific question about AUC and the Onsi Sawiris School of Business, and in general how AI should be taught: AI should not be taught as a separate silo. 

First of all, speaking of the education system, it’s more how you think and not what to think. It’s more how you learn and not what to learn. And specifically when you zoom in on AI, I look at this as two separate sets of skills. One of them is obviously the technology skills, the new platforms, the new tools that are available for us to use. But I think [what] is equally important is the human skills. How can you use the outcome that is coming out of those systems? Which is critical thinking, teamwork, analysis, data science. For me, these are equally important to what the technical skills are.

So what are we doing? Two sets of programs here. One of them is — as you know, AUC is based on liberal arts education. So there is a core program that every student gets access to. Over the last two years, the entire core curriculum has embedded AI in its courses. You go to the second level, which is the schools. Each school on its own has its own plans and so on.

Are we embedding AI in all the courses? Absolutely, yes. Are we there? We will never get there, because there is always an adaptive [element] to that. 

P.F: It’s a moving target. 

S.K: Yeah. It’s a living story. But it’s two ways. And I think, talking to my colleagues around the world, there are sets of rules and dos and don’ts and manuals that have to do with how the university is moving, where they are. But there is also the organic growth: one faculty versus the other, one group of students versus the other. And that’s exactly what’s happening at AUC as well.

P.F: I think there are two great ways to become inundated with AI slop. The first way, to Ahmed’s point, is to have bad data and turn a model loose on bad data. The second way is to tolerate people feeding you AI slop. And, Bassem, you brought something up on the call as we were preparing for today that I thought was particularly interesting.

Tell us, how do you screen people for their ability to use AI? 

I receive job applications all the time. Person looks great. Person sounds great. Person appears in front of you. They can’t put two words together to communicate with you. 

How do you use AI in job interviews?

B.F: For those familiar with consulting, typically, to get into a consulting job, you do a number of what we call “case interviews”. So you go in for an interview for an hour, you’re given a case, and you’re required to solve it in an hour, basically, maybe even 35, 40 minutes.

And we use pen and paper, right? So you kind of put a framework, you put a structure, you do some math. No calculator or anything, because we’re really testing them: How good are they with numbers? Are they able to structure a problem and, and, and…And also their communication skills and what have you. This has been, let’s say, the standard practice for quite some time, since I got into consulting maybe 15 years ago.

In my own humble view, this obviously requires a change, at least to some extent. So what we’ve introduced in Egypt — and I’m glad to say this is something that we started; it’s not necessarily that this is a company-wide direction that we need to fit in. We kind of took the opportunity to test. Given our scale, we can also take some of these, let’s say, calculated risks.

So a candidate usually does five interviews. We said, “Let’s take one interview.” That’s the interview I do, where the candidate knows before coming in that they would be given a laptop. They are given a similar problem to solve, but also a much more complicated data set that they need to, let’s say, process. And then they are given an hour to solve it with an AI model or any of the, let’s say, LLMs, be it Claude or ChatGPT, and they are required to put a presentation together. I come back in the room after that hour, and then I ask them to present their outcomes.

P.F: Okay. So they are using the AI to prepare what they’re presenting?

B.F: To solve the case. Now, obviously, it’s not that simple. At the very beginning, I still, for example, do the 15 minutes where “it’s a fit” questions. I still see how they think about structuring the problem. And then I leave them. I come back and I start to challenge them based on what they are presenting, and based on what they have not necessarily addressed with the tool.

And it’s amazing, really. It’s amazing how you see the difference between someone who is using the tool to enable and, let’s say, accelerate and advance their thinking in how they address a problem, versus someone who has kind of switched off their thinking and outsourced it to an answer given by the AI model. And I think that’s where, to some extent, education comes in, and what I really like to hear is how they are integrating that: not as a course, but in their day-to-day work.

Again, the idea is not to outsource the problem solving, but to enable us to solve problems that we were not able to solve before in a particular time, or time domain. And I was very proud to see how those who were successful in these interviews how they are performing on the job.

P.F: They’re doing well?

B.F: Yeah, it’s absolutely amazing. Now, this is something that we initiated in Egypt, and we started to share the lessons learned with other offices globally. So that was also something that feels at least… again like we can build an edge with these capabilities, obviously.

P.F: We talked, gentlemen, earlier on the prep call about opportunities that sort of took two forms, Ahmed: software and human opportunities. Bassem’s BCG X, and then things like AI data centers and value engineering with existing assets.

So I’m curious. First off, you said, Ahmed, that one of your portfolio companies is kicking the tires on a data center investment, now. What made that interesting to them? And then I have a couple of quick follow-up questions on that.

A.A: Infrastructure definitely is a main, I would say, challenge in the equation here. Of course, if you look at the existing market capacity today, we’re talking about probably 15 to 20 megawatts, which is — if you compare [it] to the announced projects in the pipeline, we’re talking about four mega projects that would basically take this number to around 400 to 500 megawatts, which requires huge investment.

So in terms of the infrastructure challenges, I think this is one of the core challenges. The second part is, again, the level of adoption and readiness at the portfolio level. In fact, one of our portfolio companies is working [on] one of those projects because of the key selling points that Egypt naturally has in terms of connectivity, the submarine cables and stuff.

But apart from the connectivity and the location, I think we have another major challenge with this: the power supply. And this, again, comes into play given that the existing grid capacity is around probably 50 to 60 [gigawatts]. If you’re talking about the pipeline that is planned or announced, I think we’re talking about a major investment in the power grid in terms of generation and transmission as well, because of the need for compute power and cooling power.

So I think this is one of the core areas. The other area that would come [as the] third challenge is the off-takers, or the hyperscalers. And now, listening to the previous sessions, we have multiple sub-issues around these off-takers. First of all, data sovereignty, which is something that I think the government is working hard [on], trying to basically find some solution for.

I just had a discussion with one of the key panelists here about a new model that is called Neocloud, which is basically an aggregated layer that sits beneath the hyperscalers and basically can aggregate a number of data centers or capacity and resell it again to the hyperscalers. So this is something that could be a solution for the existing lack of [presence of] the major hyperscalers.

I think we’re in a very early stage of, deployment, [from] what we’ve heard so far. As I said, 400 to 500 megawatts of capacity [is] being built at the moment. The only committed part is around 20, which, again, relative to the existing capacity, I think is quite moderate.

The last challenge is the funding, which, again, is something that nobody has a clue about at the moment. We’ve heard lots of numbers. We’re talking about USD 4 billion to USD 5 billion worth of investments for these announced projects. This is something I think [is] new to the market. Will it come in [as] FDIs? Will it come from local banks? International banks? I mean, the funding mix is not clear yet, which puts us, again, in a situation where there are lots of questions that need to be answered as far as the infrastructure and security and sovereignty are concerned.

P.F: Would you write an equity ticket for a data center?

A.A: Yeah, definitely. We’re looking into it, and we’re actually considering [it]. But again, it’s a very early-stage kind of investment. We’re basically putting bits and pieces together. We’re compiling a consortium for that because it has too many stakeholders in place. And once it’s ready, I think we’re definitely going to be putting money in it.

P.F: Bassem, you told us BCG chose to build BCG X here — that’s your tech and design arm — because of the quality of the people, the ability to serve clients globally. I like the model, obviously. I’m one of many people who sit in this audience who builds things with my hands. It’s wonderful to build things with your hands. 

Why does BCG survive AI?

B.F: Sure. Maybe just a bit of context. So BCG X is our AI arm. This is basically data scientists, engineers and so on and so forth that provide AI solutions to our clients, which is a parallel business line to our traditional consulting work and other lines of business. And we decided to do a BCG X hub in Egypt. The idea of that is not only to serve the local market, but maybe the bigger opportunity is obviously the regional market. And we’ve done exactly the same thing, actually, in Morocco, which is a hub that services not only Africa, but also Europe and the US. So that’s kind of what we’re looking to do.

Now, if I understand your question correctly, your question is: how would BCG  X survive the AI… 

P.F: Survive AI itself. So, not a future where the LLMs eat away at that bespoke element of the business: understanding the client’s problem, building a solution to that [problem].

B.F: So the question goes not just [for] BCG X but all of BCG right.

P.F: In a way.

B.F: Look, interesting to note that 40% of our top line today is tech- and AI-related projects. Four zero. That’s globally. In Egypt, you can imagine it’s a bit behind. So that’s one side of the discussion. And I think there was one of our competitors, who is traded on the stock market, [that] two quarters before had a very big hit. But recently it started to pick up again because the narrative is also changing.

I agree with your point that there’s, let’s say, some form of a sentiment that, “Now, with an AI tool, I can build a deck, a PowerPoint presentation of a strategy, whatever.” Little did someone know that the analytical side of the work we do is maybe 10% to 20% of what needs to be done. 80% is convincing client counterparts that this is the right solution. This is aligning interests and incentives. And it’s a lot more on the human element.

So what I do agree with is that definitely there’s a massive productivity gain. So, let’s say, in our industry, what’s really important is being able to embed these solutions and how we make an impact, and also change the narrative in how we do the work: from “I’m providing you with an ‘answer’” to “I’m helping you deliver impact.” That’s a very, very different discussion. And it’s been the case for the past two to three years. LLMs, at the end of the day, are in the business of tokens, let me put it this way. We’re in the business of making an impact [that hits] the bottom line. And that’s a very different value proposition.

P.F: So Sherif is building the graduates. Ahmed is going to invest in the data centers. You’re going to employ those people and grab the tokens and put them to work. What are the sectors in Egypt where, in 2035, AI will have made a difference in our lives? I’m going to ask you [Bassem] that question. You’re the consultant. What sectors will AI make the biggest difference in our lives in?

B.F: We’ve done recent work on that specific question. Our conviction is [that] where there’s an opportunity is energy, healthcare and — I would also like to think — education, where there’s a big, big opportunity for Egypt. That’s kind of at the application layer. So obviously, we spoke about infrastructure, we spoke about LLMs and so on. If we think about the actual application of these models to create an impact and to create value, these would be the [sectors].

Now, by the way, financial services is kind of an obvious one. Even if we think about the more basic forms of AI that do not necessarily require all the crazy GPU stuff that we talk about, there’s a lot more basic AI work that can be done in financial services, and we are far, far behind on [that]. 

So obviously the discussion is a lot more advanced. But if I think about financial services specifically as a sector, there’s a lot more work to be done before we start talking GPUs and stuff.

P.F: We are running out of time. I think Bassem has hit a super important point that I’d like you all to take away from this conversation. And it’s that when it comes to AI, the people who are building at the application layer probably have the biggest opportunities. If you are building something that a business needs, that a human needs to do their work, that’s where you have the most chance of not being commodified and becoming just another also-ran. 

I’m really excited about the application layer. I’m glad you brought that up at the end.

Gentlemen, thank you very, very much for joining us today. And I will hand the panel back over to Mr. Bassiouny.