What does AI mean for your company? 

Joseph Marks [J.M]: I want to start by getting down to really the ground level. I'm hoping you all can tell me two things your company is doing with AI: one internal and one external. Mirna, can you start us off?

Mirna Arif [M.A]: Thank you, Joseph, and good morning everyone. First of all, I'm genuinely super happy being here today and obviously sharing the stage with such prominent leaders. And I have to start by saying a huge thank you to EnterpriseAM, not just for bringing us together and creating this fantastic forum, but really for finally changing the narrative.

I think it's been such a long time that I've been asked, "What can AI do?" And I think we've all heard enough about that. And we're finally talking about what AI means. So what does it mean for your company? What does it mean for your people? What does it mean for you as leaders? 

So to answer your question and the two examples, they apply both internally and externally in my case.

The first obviously is my new best friend, Copilot. I'm literally tied to my phone, and that's not a joke. I live in that thing. I'm not sure about all of you, but I think we all juggle a lot of meetings, both external and internal, virtual and physical. And I'm known to be very punctual, but for some strange reason, whenever I have calls, it's always this Teams tick: "Your next call is starting." And I'm like, "Oops," and I have to chat, "Sorry, I'm running a bit late with my previous call." And it used to be very embarrassing joining that new call and everybody giving me that look, and now we have to repeat what we've been saying for the past 15 minutes.

But now, genuinely, with Copilot, I just join unembarrassed, because I ask my Copilot, "What did I miss? What have they been talking about? Was there any action addressed to me?" And I can capture all of that and jump straight into that meeting. And I think that's a huge productivity gain.

And moving on from just the productivity, from what AI can do for me as an assistant or a copilot, there is also agentic AI, and that's where AI becomes a collaborator and coworker. 

I use Cowork extensively now, and I have no idea how I was living before Cowork, genuinely. That's where I can actually rethink our entire operating rhythm of business and our workflows, and I can have my own agent [take on] part of the workflows. And that's equally applicable to us and to customers, obviously, depending on the industry.

But I think the best thing about AI today is that we've genuinely moved from the fascination stage and the experimentation stage to really creating business value. And I think Patrick said it really well at the opening: it's not just a defining technology. It's really a defining business model change of our generation.

J.M: Can you talk really quickly about one thing you're doing externally.

M.A: Externally, for a customer, let's take financial services, for example. AI has been a huge player in financial services. One area is fraud detection. There are tons, and I see Todd [Wilcox of HSBC Egypt] nodding, and I hope they're using that. I know that you are. So there are a lot of use cases with Copilot and specific agents detecting and automating a lot of the capturing of fraud cases and communicating with customers.

There is tons we do on customer experience, whether it's in banking or hospitality, and across industries. And that's really what's agentic and the frontier.

J.M: Hossam, what are you doing internally for your company and what are you doing for your clients?

Hossam Seifeldin [H.S]: Let me first give a brief about Capgemini, because we came to Egypt probably three or four years back. 

Capgemini is an organization where we are a leader in business and technology. But if we look internally, we are at the intersection point between technology providers and the clients. We serve all the different sectors globally.

And this is where we have been helping clients for many years to realize the benefits of technology. So now that AI is being infused in every single discussion and every single project that we're having, we're using it internally because it relates to talent, and we use it internally because it relates to how we are going to implement it with our clients.

So let me give you two clear examples. One of them is we have several teams, because we serve IT, we serve business process outsourcing, and we serve engineering and research and development: three different practices. So let's take business process engineering, which is what we call Intelligent Operations: anything around HR, finance operations, supply chain.

We have some clients that give us a service desk operation. So anything where you have issues in networking, or in your specific applications, or in your mailing system, they call our teams. We have a service desk operation, and we have what we call an “AI buddy” that listens to the call, summarizes the call and suggests to the agent, the human agent, immediately, while they are actually having the conversation with the client, the most likely solution, or the steps to solve the issue at hand.

But the AI is not only doing that. It actually takes that knowledge base and ensures that it is ready for the next call, or for the next issue that appears in another call with the same client. 

We're talking here about what we tag as "zero touch": zero human touch. And before I came on stage, I wanted to see the percentage. In one of the clients, we were able to achieve 56% zero touch which means 56% of the traffic is not handled by humans. It is handled by AI agents.

Technology used to be an asset that depreciates with time, and AI is the first thing that is actually appreciating with time, because of the knowledge that is being generated and the intelligence whenever the system operates. So this is an internal thing, because we use it internally, and we use it externally as well.

Another good example, quickly: a client in Europe has a marketplace, and every day they have suppliers uploading price lists and products on their marketplace. And usually you had to ensure that the description, the tagging, the information conforms to the EU regulation, and that was a manual, human intervention. An AI system that has already been developed looks at the data and the description and ensures that it conforms to EU regulation.

You take it to the next step, which is taxes as well. Every ingredient has a specific tax. Anything that relates to data, anything that relates to manual, repetitive work, we can all agree that it can move easily to an AI agent.

J.M: There's an interesting phrase you used: "human agent". I think that really gets to the change in thinking about how work is delegated between computers and humans. Ahmed?

Ahmed Heikal [A.H]: We live in a very difficult time, and people need to understand that. It's the best of times and the worst of times, and this balance needs to be walked very carefully. 

There are three things that are happening at the same point in time in technology: you have AI, you have robotics and you have life sciences.

And the confluence of those three and the intersection of those three will cause an enormous amount of upheaval in many sectors. My role is not to make sure that I understand every single implementation of a technology. My role is to give symbols and signals to people that this is an important tool that we need to adopt, and to put the safeguards and the guardrails [in place], making sure that I have, within the boundaries of our enterprises, a good environment to foster that, but also that the ethical issues and the legal issues are properly safeguarded.

So let me try to give the audience four particular implementations that we're doing. And I thought to bring with me today two of our CTOs in the companies to give particular examples. 

One is maintenance: how you use data in making sure that maintenance is properly done. Today, this is USD 5 million per day in profit that I forgo. Every day this refinery goes down, I lose USD 5 million. So I don't have the luxury of doing it wrong. It can happen, don't get me wrong. But if a refinery goes down, it needs seven days to come back up. So I lose USD 35 million in those seven days. So I better try to get it right, and improve my chances of getting it right.

The second thing is regarding automation systems. The third thing, which is happening very fast and is very important, is the role of AI in agriculture. We own a company called Dina Farms, and at Dina Farms today you can have drones flying, and from the color of the crop you can detect either certain things that are missing, fertilizers, etc., or certain pesticides that need to be applied. And that is very clear. Instead of having 18 tons per feddan of…you can raise the productivity to 21 or 22 tons per feddan, and that is all hitting your bottom line.

The fourth aspect, and it is the one that gives me the greatest grief, is cybersecurity. Because with AI, imagine that good agents are people who can have access to the technology. Cybersecurity can kill you. And we've seen three incidents in oil and gas recently where some of the companies had to pay huge sums of money in order to unlock [their systems]. 

So if we have a couple of minutes, I would like to give our guys here [the floor].

J.M: Just one minute each, because we want to keep things moving.

Ramy Harfoush [R.H]: Thank you, Dr. Ahmed. I'm Ramy Harfoush, head of digital transformation at ERC. 

I would like to very briefly present one of the agentic AIs we're implementing today as a pilot at ERC. We actually have a few initiatives on AI. And as Dr. Ahmed said, ERC is an oil refinery. We produce oil products, and we are very keen on the availability of our equipment, because it hits our margins dramatically.

So what we are doing is agentic AI for predictive maintenance. We have trained this AI agent with machine learning on the healthy and unhealthy signatures of our machinery, mainly process data: pressures, temperatures, vibrations, flow, etc. And the AI agent is currently sitting and monitoring that data, looking for signatures that preceded historical failures.

If one appears, it goes and looks into our plant maintenance system, picks up historical work orders, checks the spares that were used, checks the availability of those spare parts in inventory, creates reports with the lead times if those are not available in the inventory, picks up root cause failure analysis reports, puts all of this in one package and sends it to our maintenance and reliability engineer.

They are then able to validate all this data and accordingly plan a shutdown, or plan a scheduled outage of this equipment, before it actually fails. This doesn't only save time, it keeps people on top of things.

Ahmed Imam [A.I]: Good morning. I'm Ahmed Imam, ASEC Automation CTO. We have developed two applications at ASEC Automation. The first one is EMS, the energy management system, especially for the cement industry, specifically for the cement grinding area, which consumes about 40% of the power of the plant.

This system reads from the control system, from the SCADA [supervisory control and data acquisition] system, and tries to manipulate the data and learn from it. Then it gives the best solution, the best parameters and best values, or set points, for the operator to enhance the specific energy consumption.

J.M: Thank you so much, guys. And I think those are two really interesting ones, because those are two things that humans can do, but it seems like computers can do them a lot better: predictive maintenance and how you manage your energy systems.

Mirna, you talked a lot about Copilot earlier. I wanted to go a little bit deeper on how AI has changed what you do in an average workday or an average workweek. 

What work do you do differently? What do you delegate to agents that you would have done yourself, or delegated to humans in the past? And do you delegate to human employees differently, knowing that they're enabled by computers?

M.A: I think that's a very important perspective that we all need to ask ourselves. And I always ask myself three questions. I start with: what can AI accelerate for me? So what would I use Copilot or my AI system for? And that's the typical productivity, whether it's summarizing documents, preparing me for meetings, gathering the different emails and chats and just telling me, "What are we talking about? What are those hundred people saying in these long meetings?" Or preparing me for a customer meeting, giving me briefing documents, etc. So that's more on productivity.

And there is a very important thing I do with AI as well. Whenever I get those emails that I'm sure everybody gets, you want to write that genuine answer and get your frustration out, and then you tell Copilot, "Please make it more professional and make sure I don't get fired." So that's a very important prompt I use.

The second question is on agents. So what can I actually delegate to agents? And this is not just a simple question of what do I give to an agent. I think, Ahmed [Heikal], you said it really well: what are the guardrails, the permissions? How am I reinventing my workflows? How am I redesigning the whole business model? Because it's not just an existing task that I'm going to give to an agent. It's how we are changing the way we work, and which part can be given to an agent, within the right level of permissions, of course.

And then the third, and in my opinion the most important question: it's not just what can I give to my employees or delegate to my employees, it's what must remain human at the end of the day, because there are so many tasks that are just meant for humans. I'll give you the example of meetings.

Yes, AI can help me very well to prepare for a customer visit, and give me the briefs: what we've sold them, what complaints they had, what they're happy about, what more we can do, and so much more. But then when I go to this meeting, I will always have the account owner, the country manager, along with me.

Because I need the human ingenuity. I need to be able to read the room. I need the person's own judgment. That's very human. That will always remain human. 

We need to be able to make tough decisions. I will never delegate to an agent or an AI a difficult conversation with a team member, for example. So there are so many things that will remain human, and that's super important to distinguish.

J.M: Hossam, how about you? How is your work different, and how are you serving clients differently than you did?

H.S: First, I do not delegate to the agent. I mean, so far I keep this close to myself, but I use AI as a kind of assistant from a productivity perspective. And I like the two examples that were mentioned here, because they relate very much to the stage where we are in the lifecycle of the deployment of AI in the industry, which is mainly around productivity — speed and productivity.

Those are the two angles that we observe now in the majority of the projects that we are seeing. The next stage, which I think has started now, is around business process re-engineering, which is how you apply AI in order to change the business process itself, or to change the business model.

So in that regard, as I said, we are at the core of the implementation. We have clients coming to us and telling us they are not able to utilize the technology, because the technology is great, but there is a business model and there are business processes [that need] to start considering all the safeguards that we need to put [in place]: the cybersecurity, the data, the regulation, by the way, and the laws, and so on.

This is where we are putting lots of effort into ensuring that our teams are well educated on AI technologies across the whole spectrum, be it Microsoft, Claude or OpenAI. We have agreements with all the partners in order to ensure that we are well trained. And how do we infuse AI in the different projects across the sector, be it project management, coding, implementation or modernization? 

We have two things that are critical, which are around modernizing for AI. And this is clearly what we see. Every organization has legacy: inconsistency in data, legacy applications, undocumented code. You cannot run an AI system on such a platform. You need to have the right foundation in order to get the right outcome. 

And I think that's the main challenge that we are seeing. If you are a startup, it's much easier to apply AI technology, because you can build your own foundation. 

Yes, you will face issues integrating with other systems and so on, but you can actually apply the foundation on your own, because you choose it. But in organizations like ourselves and the enterprises that we are serving, the legacy is making it difficult to really get the maximum benefit out of AI.

So modernizing is going to be the first [step].

J.M: Does that vary a lot industry by industry, the problem of legacy data? Do some industries have much better—

H.S: Of course. There are some industries where it depends on the implementations. Like at the beginning, when we were discussing ERP implementation, right? It depends on whether an organization was applying ERP as is, versus a complete business process re-engineering. That makes the difference. So it depends on the industry and the regulatory framework in different countries.

In Egypt, there is discussion around the laws [and] the regulation that is required. So it's not just a matter of industries. It's where you are in the technological lifecycle within a specific organization, and the complexity of your IT organization.

J.M: Ahmed, you told me a story about bringing some scholars over from New York University to talk to all of your executives. Can you tell us that story, and why you thought it was important to do that?

A.H: I sit on the council for NYU, so we have access to all the university professors from the different campuses that they have. We meet twice a year, once in New York around graduation days and once in Europe. And one of the meetings we had was with the university professors surrounding AI. They tried to give us, as a board, a sneak peek into what's coming.

I met with the professors, I was impressed, and last year I took every[one] across the layers of the different companies that we own. I took them for one week to Marassi on the North Coast in May. We were maybe 250 people with those professors, and we stayed for an entire week. 

That's, again, one of the signals that I, as the chairman of a company, can give: this is important. If I am taking one week off for this, then it must be important.

Everybody should try to get the best implementation. And there are two things. There are the local flows, the local processes. But also, sometimes you may find that you are able to commercialize an agent, and that can bring some value over time.

J.M: We're running short. I want to go kind of lightning round now. 

Which industries in Egypt do you think are going to benefit the most from AI in the next couple of years, and which are in the most trouble? Start with you, Mirna.

M.A: I honestly wouldn't put it this way. It's not one industry versus the next. I think it's within the industry: it's one organization versus another. I think AI transformation is not a technology transformation. It's really a business model transformation. So it's which companies, organizations and entities are really going to go through that change within each industry.

But then I think the regulatory framework, like Hossam mentioned, is going to be a key enabler, because some organizations are very ready. They want to transform, but then regulation would prevent them from doing so. So it's a coordinated effort that needs to happen between the policy, the regulatory framework, the industry, the skill sets of the employees, the training. It's a full thing.

But if I really have to pick a few industries, then I think there is huge potential in financial services. Huge. There is a very big potential for tourism, for sure. Agriculture is definitely huge, and it will generate a lot for the country. 

But in a nutshell, it's not one industry versus the next. It's really what we're doing within [them] as companies and entities.

H.S: From our side, I think all industries will benefit, and all industries will have the same issue. I think it's a challenge of, as Mirna was saying, how organizations are going to make the change, the culture change, the shift into adopting the technology. At the same time, from Egypt's perspective, the public [sector], the government, can really get a big boost when it comes to using AI.

If you take the education sector: the challenges of having enough teachers, the challenges of building schools, the CapEx that is required. And you think of AI, again, you change the business model. The idea used to be going to an institute, to schools, to get the knowledge. The knowledge started to be available widely, and the agent can replace some part of the teaching. Of course, there's still going to be the social part that is required for the kids to go to school and so on. That's totally understood. But the business model can change.

The other [point], because I think this is important: what we have is the talent. Our strengths in the country are the talent [and] our location, but also, for instance, languages. And that goes back to the call center business. Our strength is around languages, and AI is going to somehow eliminate that advantage when it is widely [used]. So that's an industry that will have lots of pressure going forward.

J.M: I just want to give Ahmed one minute before we're out. Are there industries you would not invest in now because of the AI change, that you would have before?

A.H: It has profound implications, but again, not only because of AI. It's the three things that I mentioned in the beginning. Robotics will have a humongous change. 

For example, there was a pendulum shift [of] certain types of industries moving from more mature markets to emerging markets.

Let's take the textiles industry. What happens in ten years' time, when there are going to be robots that are able to do exactly what humans are able to do, and whether this will result in a shift back? Because the unit labor cost of robotics is going to be a fraction of that of similar people in emerging markets. And then you're talking about [robots] that can work 24 hours a day. So the productivity gains are going to be enormous.

And that is going to take us into a totally different shift. And of course, there are certain classes of industries where I would be very, very hesitant to invest.