What does AI mean for your team?

Nick Edwards [N.E]: Thanks for being here. Thanks for joining this panel. 

Depending on the headline that you all have read in the last 24 hours or so, AI is coming for all of our jobs, or it’s creating millions of new ones. It’s slashing payroll costs, but it’s also driving up the price of human inputs. There’ll never be another entry-level job, but experience is going to fetch a premium price. It’s solving workflow problems and it’s creating entirely new ones. 

So how do you build a winning team when you’ve got everything like that happening everywhere and all at once, those contradictions?

Todd, I’m going to start with you, if I may. Deploying AI to a team, rolling out technology in a global organization, as I remember all too well, is a real test of human endurance at times. 

So, what has deploying AI really taught you about its impact on team dynamics, on workflow, on process, and on unintended consequences?

Todd Wilcox [T.W]: There are different types of impact. Where I see a real big, immediate opportunity is in fraud detection [and] financial crime, where you’re dealing with masses and masses of data, and those types of things run in the background. It’ll have an impact on the nature of the work, so not the people. Where I do see changes is [in what] one of the previous panels talked about, the ground-up AI. 

Just to give you an example: every day we get thousands of calls to our call center and we capture all that data. The call agents type everything in. And we’ve always struggled trying to get to [the] root cause of issues or operational situations. 

We were able to build in-house an AI model to really look at all the data that we’ve gathered over the last five years and really start to say, “What’s the root cause? Why are clients calling? What’s the impact?” and map it back to our processes as well, because we’ve put all the processes in that AI engine as well.

In the first meeting we had to look at all that, everybody was wowed, me included, because it really showed you something we’d been asking for, for years. I’d been asking for five years.

So suddenly we get to a stage where, ah, we know a process that we do back here is triggering 5% of our calls down the path. What happens, though, and we pull the team together to really start actioning some of this, you get caught in the awe of the AI output and the AI engine, as opposed to, “Let’s get to the root cause and let’s get to solving the problem.” And there’s the danger, I think, of getting so caught in the data and the AI mystique that you forget that we actually have a job to do, right? And that focus. 

The other is that you still challenge each other. That’s my big concern as a bank, and as the head of a bank: I need the team to challenge each other on the approach we’re taking, on what we’re doing, and there’s a real danger that we just take the AI output as the direction to go without that real check and challenge.

So I think accountability still remains there. The nature of the skills we need around the executive table is even more: challenge each other, challenge AI, and don’t just take the output for granted.

N.E: Khalil, can I ask you related to that, does that sound familiar to you in your business, that sense of challenge and accountability?

Khalil El Bawab [K.B]: Indeed, it impacts. It’s not just about making decisions. It’s about which decisions you want to make, under what controls, and you need to control it for sure.

And accountability is quite important. It’s the core of using AI in the day-to-day business, or the quarterly, or the strategy and so forth.

N.E: Omar, when you’re thinking about the way in which AI is deployed, particularly in the businesses that you invest in, how important is it that the client companies, the invested vehicles, really understand how they’re going to deploy it? Do they just think it’s a great idea, or do they really have to understand it for you to get interested in them?

Omar Khashaba [O.K]: Well, I think it depends on what layer of AI you’re speaking about. People are talking about AI as if it’s this sort of strategic imperative in and of its own right. But really, AI is going to become like the internet. It’s a commodity. And so your question is tantamount to someone asking, “How important was the internet in building e-commerce companies in the late 1990s or early 2000s?”

And the answer is it’s incredibly important, but it’s sort of the diffusion layer. Unless you’re working at the bleeding edge of AI research and you’re building an AI-native company, it’s an applied level within the organization. And of course, it’s incredibly important.

N.E: Are your teams trusting of you when you bring AI and you say, “We’ve got this great AI tool that we’re about to develop”? How trusting are your teams, Todd, that they think, “This is good for me”?

T.W: I think right now everybody is interested and we’re learning together. Remember, in the financial services sector here in Egypt, we’re still a little bit in our infancy, because the regulatory and legal framework is still being defined. So everybody is on a bit [of] a pause. We’re testing [it] on the small projects and we’re learning together.So I think there’s a growing interest and a growing awareness too of how broad the use can be. 

Am I seeing people concerned about the roles at this stage? No. What we are doing, though, is really getting people up the curve — and you’ve heard it from many panelists — on their knowledge of AI, how it works and how to work better with AI.

We’ve launched something very basic, which is Productivity Suite, which is a Microsoft project that we’re now starting to use for our papers, our decks. So I think we’re [at] early stages. What I am concerned about is less so the jobs, because the jobs will change, they will transform. And we know that that will happen.

When I started in banking a zillion years ago, we used to use factor tables to calculate mortgages. And then when we launched financial calculators that everybody used, half the bankers didn’t make it, right? They couldn’t transition from the tables, which was a booklet, to the calculators.I think if you’re not able to keep up and you’re not willing to learn, you’re going to get caught. 

The other [thing], though, is that AI is really at its best in banking doing the routine transactions or the routine analysis. So it’s reconciling, and it’s going through all the data and doing a summary.

That’s also part of how you learn — in banking. So you know, our relationship managers, again when I was a new corporate RM [Relationship Manager], I did the analysis of financial statements myself, right? I did it on a spreadsheet, but I did it myself. I did the industry analysis. 

That now can be done by AI. But you start to lose some of that core knowledge and core capability, [so] we’re going to have to reinvent how we train new bankers.

N.E: Okay, Khalil, I know you’ve got a view about that.

K.B: So the way we train new members in the staff needs to change, because the basic things are going to be done by AI. So trusting it 100% is… okay, we trust it in the collection of data at this point in time. We trust it in making suggestions. But do we trust it in deciding? Not at this moment in time, not in our line of business.At the end of the day, we’re entrusted to invest people’s money, so we need to be very cautious with the decision-making. 

As I said before, [we should] not [let AI] take decisions on behalf of us. And we always need to check the quality of the suggestions, the quality of the outcomes.It’s still too early to decide on that. 

Needless to say, regulations are not covering it yet, at this point in time. For instance, client data is highly critical. As per the current regulations, client data cannot be on the cloud. It could be on a local cloud. It could be in-house. And that’s very sensitive, and we need to abide by the regulations. 

But that’s an opportunity: you need to have rules and grow with it, go in line and basically synchronize with the regulator, who is, in our humble opinion, very open to those developments. 

To serve your clients, you need to be using AI tools to help them analyze, let’s say, stocks, funds or whatever. But do you trust AI to take the decision at the end of the day? At this point in time, it’s [the] client’s decision. Give them tools to analyze. Give them tools to think about the information. 

Ages ago, finding the information was a challenge for us. So you would be successful if you had the information. Currently, I have pity on this generation, because they’re flooded with information. The quality of information, or the quality of data, is the key here.

N.E: Well, that leads us very neatly, I think, because you’ve both mentioned trust a little bit. 

Omar, you’ve got a model that you were telling me about, and how important it is to keep humans in your judgment loops. So you’ve got two things that you use that sounded really interesting: judgment loops and what you call the Open Brain. Can you tell us a little bit about that and how that builds trust in AI as a member of the team?

O.K: Sure. Just for context, Open Brain is the system that I’ve created to externalize my own judgment. But I’m going to link my response directly to Khalil [El Bawab]’s last answer. If you listened to some of the panels earlier today, most people were using AI to prepare for meetings or draft better documents, et cetera, and I think that is barely scratching the surface.

The real leverage of AI is in what I call building systems of judgment, right? And in order to get into that, you have to ask: what is judgment at an organizational level? A lot has been written about this, but the best ideas out there are that, effectively, you have this context graph that is an institutional memory of how decisions get made within an organization, and that ultimately that is what is going to be key to enabling AI systems to really power us into the next era.

Now, if you think about software in the previous era, what you had were systems of record, right? You had your Salesforce for customer interactions, your Workday for HR, your Oracle or SAP for your operations. But what those systems of record captured was the “what”, right? They didn’t really capture the “why”, the decision rationale, the decision traces across an organization.

But now you have technology that is capable of exerting judgment much like a human would, which is non-deterministic: not just to say yes or no to something, but to say, “It depends on the context.” But if you don’t provide that context, that’s like asking a lawyer, or a judge, to rule on a case without access to precedent.

The problem is we haven’t been capturing that historically. Look at our own meeting minutes, at least [in] the organizations that even take meeting minutes. Typically, meeting minutes will capture outcomes [and] agreed-upon action items. They won’t capture the key arguments that led up to a specific decision.

To cut a long story short, there’s a lot of technical nuance here, but it’s important to start thinking of organizations as world models, meaning they are state representations of how various interrelated parties behave. 

There’s an analogy in robotics, if you think of self-driving cars, we expose self-driving cars to world models that effectively simulate how objects move through space and time, and how they behave according to the laws of physics. Organizations are world models that are powered not by physics, but by decision dynamics.

And that’s what experienced employees have that new hires don’t. They don’t have a different cognitive architecture. They just have a better understanding of how the organization as a whole exercises judgment: “We tend to give this kind of exception, this sort of discount. If this customer churns, it’s problematic, and so we’re willing to go all out for them,” right?

So I anticipate a future where, as the models commoditize, companies are able to create a new competitive vector of organizational intelligence that compounds.

N.E: Todd, you have a look on your face that suggests that you’re digesting that.

T.W: I would say in financial services, that would be a huge challenge. And I’ll tell you that when you’re running an organization, especially when you’re running a bank, I answer to the depositors and the shareholders, but the buck stops with me and my team, right? So that decision on what we’re doing, where we’re going — it’d be difficult to defend to depositors that the bank failed because of our AI model and the judgment.

So there’s that accountability part that I think would really need to be clear [for] the executive and the team. And I think this is true of any organization. If shareholders are paying the executive, they’re paying the executive to make decisions and be accountable for those decisions.

What I see is that check and challenge of the executive around the table: the chief risk officer, the chief financial officer, your head of sales. They all bring diverse experience to the team, so that the decision you make today may not be exactly the same as the decision you’d make tomorrow, or the decision you’d make if you were sitting in Vietnam as the CEO of HSBC.

So I’m still a big believer. I think AI will give us data quickly. I think it’ll enable us to see patterns we haven’t seen before. It’ll help us make, hopefully, better decisions. But if you’ve got poor governance and no accountability, it simply leads to making bad decisions faster.

N.E: Does it change the type of people that you try and hire?

T.W: Well, if anything, Nick, the importance of decision-making and accountability, and having that deeper experience, becomes even more critical, because now they’re working with much, much more data, refined, right? So they’ve really got to be on top of their game. And that’s, again, what we were talking about: my fear is that [the] next generation doesn’t get the groundwork to get to that stage.

So we’re going to have to work on new internships [and] new development programs for those people, so they’re ready.

O.K: [Can] I just very quickly come back on that? The reaction is understandable, especially in a highly regulated industry like financial services. And I also agree 100% that accountability needs to be front and center. That’s part of what we all experience in daily life and are frustrated by, what you would call “AI slop”, right?

Which is effectively just, “Here you go”: something that AI put together that is not really cognitively accessible to the reader because it’s been condensed so dramatically, and it doesn’t follow how humans engage in [a] chain of reasoning. 

But, Todd [Wilcox], I don’t think it’s a switch you flip.

I think it’s a spectrum where you expose the AI to your decision-making, which is fluid. So it’s not a state in time, because the organization changes how it makes decisions over time. But then you incrementally decide which parts of that decisioning to surrender to AI, with humans in the loop, with your checks and balances.

And over time, as you develop more trust and confidence in that judgment, you’re willing to then have the humans move upstream to other parts of the decision value chain. I think it will happen. AI is making decisions. It’s just about how you make sure that, a, those align with your organizational judgment, and b, how you create the checks and balances. Because 100%, accountability is going to reside with the humans.

N.E: Okay, Khalil?

K.B: I agree and disagree with Omar [Khashaba]. I agree that AI is definitely the destination. It is going to be the only future. Yet, we’re not saying we’re not believers in AI, we are believers, but we’re treating it as a new hire that we need to train. We make sure it’s controlled. We make sure it has accountability for us to fully automate.

So we’re not in disagreement. As you said, and Todd [Wilcox] said, indeed, we are a regulated industry. 

At the end of the day, when it comes to the financial sector, in case, God forbid, [of] litigation, who’s going to be sued? It’s the managing director or the CEO of the company — not the machine — who puts the controls on the machine.It’s you at the end of the day, for whatever glitch happens. So it’s very sensitive. 

We’re not saying we’re resisting. We’re definitely going for it. But we need to be aligned: we have risk management, we have controls, we have regulations. So it’s happening, but probably in our industry the pace could be slower — that’s [the] disagreement.

N.E: Okay, so regulation, highly regulated [or] lightly regulated, might depend on the pace at which AI becomes part of your team and the types of tasks that you give it to do. But ultimately, it sounds like you’re saying as well [that] if you now have a new type of new hire in the market, it’s how you approach the training of that new hire and the kind of scope that you give that new hire work to do. 

Okay, so no change there. 

We’re running tight on time. I’ve got a last question for all of you. Based on your current business projections for revenue and costs, at the end of 2027, do you expect that AI adoption will see you with more or fewer human staff than at the end of 2026, and by what percentage? So thinking out from now to the end of next year, more or less staff?

T.W: I think it’ll be the same, but I think some of those jobs will be different, right? We’re going to shift the jobs. And because the regulation is still in development, you’re not seeing the banks make big, dramatic moves, right? It's iterative moves. So we won’t see a revolution in banking here today. 

Globally, I think we’ll see some shifts in jobs through HSBC, but it’ll be very much shifts away from that routine processing to more upstream.

Omar [said it], and I agree with that: you can [move] upstream, and some of that decisioning of the lower-level tasks for us. 

N.E: So that’s a no change. Omar?

O.K: No, I agree. I think that’s completely right. You’re not going to see a disruption in terms of the overall percentage of employment going up or down, but I think that there’s going to be dislocation, meaning the nature of jobs is going to change. 

It’s tough to be a 20-something-year-old in today’s world and have to figure out what sort of skills you need in order to be competitive in the marketplace. But my instinct is that the market will continue to reward a,  generalists, but also people who are highly analytical and logical, and who are able to think in terms of systems.

N.E: Brilliant. Khalil? 

K.B: We need to be agile. The team needs to be agile. We need to adapt to the new line of business. 

I don’t see [us] reducing. On the contrary, because one major component in the financial sector is the digital transformation, which is happening — digital onboarding and so forth — which gives you access to ample clients. So I don’t see us not growing our team, because the reach is much higher.

So you need to serve those clients. You fully automate, you use AI and so forth, but still there are some decisions that need to be taken by humans, and it also depends on your client segmentation. 

Gen Z, Gen Alpha probably don’t want to talk to a human being. When you go to high-net-worth and ultra-high-net-worth [clients], la [no] you’re definitely going to need the human element. If you them them technology, probably not. 

So the human interaction in our business is still needed, but we need to be more efficient [in using] all those tools and drive growth for the business.

N.E: So that sounds like you said that you might be hiring some more people. 

So, some more people, no more people, no more people. 

Okay, so there we go. If anybody in the audience is thinking about where the jobs might be in the next couple of years, Beltone is where they are.