Reimagining business through AI

Welcome to Beyond the Benchmark, the EFG podcast with Moz Afzal.

Moz Afzal:
Hi everyone. So I'm very excited to have José Pedro Almeida on the podcast today. So José is actually introduced by someone who's a frequent comer on to the pod, obviously Nathan Furr, who is a part of the EFG Future Leaders Network, and José also works with Nathan on the Digital and AI initiative at INSEAD. So José, welcome to the podcast.

José Pedro Almeida :
Such a pleasure, Moz. Thanks for having me.

Moz Afzal:
No, absolutely. Maybe just to kick off, maybe you can describe some of the exciting things you've been doing and the complex things you've been doing within the healthcare space, and of course some of the initiatives you're doing today.

José Pedro Almeida:
Yeah, sure. So for the last 17, 18 years, I helped transform some of the most complex organisations that operate in the healthcare space from patient care to well, retail diagnostics. Some of those organisations, private equity owned more than a thousand stores, 80 companies, so pretty complex businesses. And basically what I've done there working internally leading AI and data for their CEOs was reimagining how things can be done. And that's exactly what I'm doing nowadays as well. It's advising large organisations, consumer businesses from financial services, insurance, retail, governments, you name it, on a new trend that we have among us and that requires any business to completely reimagine their business and operating models.

Moz Afzal:
So maybe we'll start there or how do you go about re-imagining a business? Because clearly digital has been around for a while and many companies have embarked on that journey, but clearly AI adds another dimension to it. So what frameworks do you use to get people to reimagine?

José Pedro Almeida:
Well, really high level, what I think companies need to understand and what I see across the industries even when I'm at INSEAD, teaching at the leading AI in digital transformation programme is there are no shortcuts to your AI and digital excellency. And if I had to frame two or three key points that any company needs to do, the first one is that they need to be all in on this. And that comes first of all from the empowerment of the CEO, which means that this is not something that you buy from a vendor or a consultancy company and it's done. It doesn't work that way. This is not an occasional pronouncement, data and AI needs to be at the centre, needs to be in the organigram of the organisation. The CEO needs to block agenda, allocate resources. A lot of blockages will come along the way from the business because the immune system of the organisation will kick in.

But mostly not being too focused on the ROI on the return on investment, but being more focused on the vision, like understanding that in some shape or form, this will bring speed to the business. The second point I would highlight is understanding that the organisation needs to have its own internal intelligence, its own soul. And that means that you need to hire some champions and ninja teams that know about these new topics, data, AI, software technology, and that are able to blend with your business, with your business units, work side by side with them, build these products with them, understand their pain points because only then you will have the change management process going on because otherwise they will fight the transformation. And lastly, when you have something to deliver, how do you democratise access? How do you empower all these people because habits are hard to change. It needs to be seamless, easily accessible in the platforms they use. And so these are probably the main secrets for any organisation in any industry that I would immediately share.

Moz Afzal:
So clearly that is a big challenge. First of all, getting the CEO of the organisation to understand that rather than, which I think is what you're alluding to, is avoiding this is another IT project that we need to do.

José Pedro Almeida:
Exactly, exactly.

Moz Afzal:
So we really need to come from the CEO to do that. In terms of the organisations that you are consulting to, what are some of the tricks they've used to try and get the CEO on side? Probably the first thing. And if the CEOs on side, what type of people are they typically recruiting into the organisation to ensure the strategy is fulfilled?

José Pedro Almeida:
Well, I would say there's no trick to onboard the CEO. If you look at a company like Walmart, which I follow for a decade, is really inspirational. Doug McMillon, there's no trick to onboard him. He is all in. He bought Jet.com when Walmart was trying to fight Amazon. Look at where they are today after that investment. It completely changed their culture. And so to be fully honest, I only work with CEOs that are already with that mindset because I know that I'm losing time if they are not already there, which means that they need to retrain themselves as well. The new CEO needs to also be an AI CEO. And so I think that's the main point as well. There's no shortcuts there. Do you have the empowerment and the vision or you do not and you will fail because when people see that the top management is all in, they understand that they need to change because the train is leaving the station. If not, they know that that is another project that will die very shortly.

Moz Afzal:
So in your experience, have CEOs gone in which I've seen in other places, hired a McKinsey or a BCG?

José Pedro Almeida:
All the time all the time.

Moz Afzal:
And is that a successful model or not?

José Pedro Almeida:
It might be a successful model for a single use case, niche use case, but the point here is that this does not transform the entire organisation. And we are at a point in time where organisations need to be transformed in an enterprise-wide way horizontally for every business process. And so what's funny is that quoting one of the CEOs that called me recently, and I'm working as the fractional Chief AI and Data Officer, his words were “I want you to work with me because you tell me the truth and I'm tired of hiring all these consultancy companies. I spend tonnes of money and nothing happens.” And the answer is always the same. This is hard work. There's heavy lifting to do. Your company is completely fragmented. It moves at its speed because it is fragmented. And so what we need to do is build a separate layer, which I would call a digital representation of your business that's built sideways. And even BCG calls this the digital, data and digital platform approach where we will start to have an abstraction of that complexity, but we will have access in a single place. And that is absolutely crucial because these new AI, they need context to bring value and that context is siloed in all those systems that you have hired. And so that's why I really advocate for this enterprise-wide approach with the top management on board, a single shot doesn't change anything.

Moz Afzal:
Right? And then the others, I guess as part of the implementation of that strategy is measuring success and KPIs to measure success. Maybe you can, because obviously, CEO or maybe the CFO will look at cost or look at profitability or some of those KPIs. If the non-traditional KPIs that one might think about when you are trying to implement I think a parallel digital AI based organisation.

José Pedro Almeida:
I will be very controversial in this topic. And by the way, this was one of the latest discussions in these business schools with all the industries saying, my CFO is asking for that Excel spreadsheet where I have the savings. By the way, that's all in my perspective. That's all a lie. You are lying to yourself because obviously what they tell me as well is that they fill those Excel spreadsheets because they want the exco to be happy, but in reality, nothing is changing. And so answering your points and going to the start, this cannot be ride driven. This needs to be vision driven and there needs to be risk on the table, which means that you need to be willing to invest completely blindly and saying probably there's a probability of throwing this money into the garbage, but I will hire the teams that have at least success in their past that have transformed organisations.

And so there's also probability of success, and we will just start, I was listening to Jamie Dimon recently. He was interviewed at Stanford and his words resonate with me, which he says, I don't care how much this costs, we need to do this, just do it. Just start. And that resonates a lot with my career. So going to the point of what kinds of impacts do I enjoy measuring, one of those impacts is usage. It's not on the P&L. We were doing 5,000 business analysis per year because we were doing it in Excel and suddenly two years later we are doing 500,000 because these AI and these analytics, they come to us and we move much faster and we can measure that. That is one of the ways of measuring this transformation. The other one is that things happen almost naturally. Those outcomes, whether it's for the top line or the bottom line, they will come naturally.

Just an example on the hospital side that we have transformed, it had a huge platform that was managing 560 billion data points. What that means is that you could ask more than 500 billion questions to a structure like ChatGPT and it'll answer you in a flesh of a second, any department, what is the value of this? How much is this worth? Just an example. They could immediately see where were infected patients, which in hospitals is one of the worst problems worldwide. End of story, two years later, hospital acquired infections dropped 30% just because of they have information faster. What does that mean for that P&L, 30 million annual savings just for this organisation? But we didn't start the transformation with that expectation. That came naturally.

Moz Afzal:
That's a really, really good example.

So the CEO has bought into the AI strategy. He's all in, right, exactly how we described it a little bit earlier. What are the next steps in what type of people should the CEO be recruiting to ensure that that strategy is implemented?

José Pedro Almeida:
Yeah, so first of all, I think that at the exco level, you need to have a Chief AI and Data Officer for sure, which is someone that is able to blend AI data, digital because you'll need all these pieces working together. And the function of this person is first to influence the senior leadership team because the CFO, the COO, the CMO, all of them are too busy. They are not thinking can they do their job differently. The second role of this individual is to think how can we rewire the entire organisation, each business process, how can we do this differently? And the third role of the individual is igniting the culture, inspiring people, getting everyone on board, getting this learning hackathon so that everyone starts experimenting these tools so that everyone starts testing, starts getting into the train. This takes a lot of time, effort, but this is surely needed because people need to change.

And the second angle is that besides this type of talent, and then you'll hire a data platform area, an AI factory area, infrastructure for all of this. So it goes top down. But then you need to focus on those technology foundations as well, which means that again, you need to build this digital layer, this digital abstraction of the business that takes a lot of time, a lot of effort, but eventually what you manage to do is that you have in a single place the semantic layer of the organisation where you operate as a single business, even if you have 80 or 90 companies behind in that layer, you seem to operate as a digital and a single business, which means that your online customer and your retail customer is the same, is connected to each other. All the heavy lifting was done and now you are ready to start plugging AI from marketing to finance to operations, you name it.

Moz Afzal:
So the plumbing is pretty important beforehand, right? Before you can even implement anything. I guess that's a much more digital, traditional digital platform implementation issue. As you just described.

José Pedro Almeida:
There's no GenAI strategy without a data strategy.

Enterprise-wide. Which doesn't mean, this doesn't mean that you don't have opportunistic AI use cases. For instance, team in marketing. There are tools where you can nowadays, I was just, by the way, I was just in an airport going to speak to a financial services group, and in five minutes going to marketing use case, I was able to generate a video commercial campaign for them just with a picture and a prompt going from text to video commercial. So these opportunistic AI use cases exist, but this does not transform the whole company. So you need to do both.

Moz Afzal:
Right. Which I think is a very, very good point. And do you see that being coordinated through the, I guess the AI CEO or the AI exco member? If you like running in parallel, you've got sort of I guess change your organisation and then you've got a tactical piece. Can those run in parallel?

José Pedro Almeida:
Yes, they can run in. No, no, for sure they can and should, I think they can and should because building that digital representation of the business takes years because most businesses are lagging behind. So they will take years. They have not even started and they are selling that they are AI driven just because they have activated ChatGPT for their employees. This is not being AI driven. This does not change the way you operate. So this will take years. But you can do these opportunistic AI use cases and one of them is just use the tools that are available. If you don't know what perplexity is, if you don't know what Google AI Studio is, if you don't know what Gemini Deep Research is, you are a walking dead. And that's the message to send to every employee. Even in your personal life, you need to be using this stuff.

Moz Afzal:
Yeah. So digital plumbing. Next step then is the GenAI. And obviously we talked about tactical and structural. What about GenAI? And you've described it as a bit of a game changer.

José Pedro Almeida:
It is.

Moz Afzal:
How does that then build on from a GenAI perspective?

José Pedro Almeida:
Well, I would say that it is a game changer because for the first time and doing the parallel for the last 20 years of doing AI for the first time, there is an entity that understands our world and this entity. While in the past we had to explicitly tell computers how to do things. They were merely tools for us. And to build an AI model was a cumbersome work of a huge data science team that would take one two years to go into production because you had to do all that engraft work of placing data into a table, all of that. Now you have this generalist that has crawled the entire information of the world and that has compressed this expertise and that you just need to unzip that expertise for your use case. And that is a profound change because one thing I think people are not aware of is they are too focused on the chat interface.

But what people need to understand is the digital species that lies behind that. What does that mean? That means that this entity is multimodal. It is able to go from video to intelligence, from sound to intelligence, from image to intelligence. It is multitask, it is able to do any task. And another thing that I would highlight that people need to be aware of is that as you scale data and compute of these large models, they get these emerging capabilities, which means that they are now starting to programme, which means that you will be able to generate a programme for your use case on the fly. And they are starting to use our tools natively, which means that in two, three years time, they will be able to do a task that, a digital task that any human now does. So imagine that these models will click any of your enterprise softwares, download the reports, and do those 50 or 60 clicks that you nowadays do, they will do this for you.

Moz Afzal:
So I guess thinking from a practical perspective and trying to draw an analogy here, you essentially got lots of, I guess GenAI, I guess the word we use a lot, today's agents and.

José Pedro Almeida:
Exactly.

Moz Afzal:
The skillset set that you have as the individual working in the organisation is like a conductor. So you are bringing all of these, I guess, music together into what you want to try and achieve. What are the skill sets you think are really important for the individuals who's running that task?

José Pedro Almeida:
Quite simple. The skill sets in my opinion is just being prepared to change and being prepared to learn fast. This is not a common skill.

Moz Afzal:
I know I was going to say it's quite hard,

José Pedro Almeida:
But just being aware of everything that is happening. Are people using X.com to know the latest version of Gemini Deep Research? What is coming out because it comes out on X.com, first place. Are really people looking at that and testing in their daily work, daily life. How can this make my life better? That is the skill. And by the way, I was listening to Morgan Stanley AI journey a few months back and they were telling that most of their teams are liberal art majors under 25 years old. They are not AI focused people because when you are on the end user front, you just need to be willing to change, to test, to experiment. Because the model, these models, they do most of the work for you. You don't need to fine tune anything, you just need to leverage the capabilities that are already available.

Moz Afzal:
So I guess just that sort of being open, being curious, being able to experiment, being able to be able to try things and fail. I guess failure is going to be also part of that, but not fall into the stigma that failure actually sometimes brings as well. So I think that's probably I guess a super important point. Maybe can you give some tangible examples of what you've seen where companies have been able to use some of these agents in an effective way?

José Pedro Almeida:
So by the way, agents is not yet the things that all companies or major companies are using, we are not there yet.

That capability only came after the DeepSeek reasoning models. They're starting to be reliable. And so you don't see large companies using them at this stage, but I would say that consumers can use them already. And I would focus again on Gemini Deep Research, on ChatGPT Deep Research. I was showcasing an example where I wanted to know the term deposits that are available in Europe over 3% that would require me to go into every company and every bank website and analyse all this stuff. And I just sent that to Deep Research. And what that does, it unleashes all those agents. They crawl all these websites, it crawled 469 websites in five minutes.

It analysed the content and it produced me the ones that have the best term deposits, the maturity, all of that while I was having dinner. And so what you are starting to see in some organisations is that they are partnering with some players that are really good at doing this internally with your internal data. Some companies that we could reference as Glean, Box, Cohere, a lot of companies that are already good indexing your internal information and they are starting to plug in these DeepSeek reasoning models to search for information the same way I was doing in the web, searching in your internal documents. And just to give an example of Morgan Stanley again, which it's farfetched clearly, but they saw their internal documents, which were only accessed 20% of time to go up to 80% of time. So you see the difference in usage and being able to use it to use the capital, the inflect capital of the organisation.

Moz Afzal:
That's quite scary about the deposit rates. So I might talk to you a bit later on to see how I can use that too, but I think that's a really good example of how you can use that information I guess for our personal daily lives, but indexing that in the organisation. So actually maybe I want to pick that up a little bit. You talk about indexation or indexing your internal documents, right? Yes. It's a very tedious task.

José Pedro Almeida:
And it's a science and it's a science,

Moz Afzal:
Yeah. Maybe give some examples of companies you've worked with who've done it well and maybe someone who's done it really badly and why? Because I kind of know a little bit about this. For example, hospitals, and I know you have a lot of experience here. I've been traditionally very bad at indexing scans and this sort of stuff. How easy or difficult it is and what are the big failure points?

José Pedro Almeida:
Well, I would say it's one of the most difficult parts is indexing that information and then retrieving that information.

And these are two separate processes. I would say it's still an ongoing science information retrieval came with Google. A lot of techniques that are being used nowadays are similar to the ones that Google uses. And that's why I was also advocating for some companies that are ahead on this because they have the leading AI researchers that have 20 years of experience doing information retrieval. Because the mistakes I see sometimes is that some companies think that they only do retrieval of augment generation and the embedding, so going a bit technical and that it is done, they plug this, they get a kid to do those embeddings in a few lines of code and it is done. And it's a complete mistake because you need to have a lot of evaluation metrics of how that information is being retrieved. Is it being accurate, is it valuable? All of that, because it is searching, sometimes what engines are doing is that they are searching in a high dimensional space.

This is a probabilistic search. And so sometimes they find the information, sometimes they do not. And you need to have this cycle of feedback loop and evaluation metrics built in into your structure. That's why I'm talking about the AI factory. This is not only plug and play and sometimes what I see the companies that are ahead, one example Deustche Telekom working with Glean side by side is that they partner with these players that are really good at indexing and retrieving information instead of asking their internal teams to do it from scratch because this is really an expertise topic that will take you 10, 20 years to be an expert.

Moz Afzal:
Yeah, I think one of the things that certainly has been going through my mind, is it better to start a completely new company from scratch to be able to be digitally and AI native right from day one? And if the journey you're going to do say, I don’t know, a three year or four year project or five year project as existing company or a brand new company, is there an advantage here of being from scratch?

José Pedro Almeida:
Yes. But I think that approach fails. And it's not only my experience, it's some of the leading books and references on digital transformation. And the reason is that that will kick in the immune system of the organisation. And so what I think works best is that you bring the talent inside. People start to know that talent. They take coffee, you know, at the cafeteria with that talent. That makes a huge difference. And again, I would highlight the transformation of Walmart after they bought Jet.com. There's a very interesting book that is called Winner Takes it All, and it is about that process of how they brought Jet.com talent inside the clashes that happen throughout the transformation of trying to be two companies at the same time, the traditional one and the digital one. But eventually these two ones merged and look at where they are now. They are for me a reference for decades of how to operate on top of data and AI and to really transform customer experience.

Moz Afzal:
No, that's a good one. So Winner Takes All. So we'll definitely make a note of that.

José Pedro Almeida:
Winner Takes All.

Moz Afzal:
So really thinking about maybe stepping back to where we started is the CEO and the AI CEO mandate as you call it. How do we start thinking about those internal AI capabilities and actually start thinking about that process that we need to go through?

José Pedro Almeida:
Well first you have someone in your executive team that you'll hire and that you'll place there because everyone will be too busy to focus on this. This needs to be strategic. And then you need to tell everyone that this is at the centre and that this is not IT. This is something that will be built a data, AI, digital department, call it whatever you want. And we will have to start, we will have to allocate the resources, which is one of the first stepping stones. There are never resources to this because this is not one more AI projects. This is the future core of our organisation because one thing we didn't mention that people need to be aware of is that we are going into a future where we will have a hybrid workforce and the other part of the workforce, which will be digital, which are these agents.

It's just a function of having more data and more compute. And so we need to build those capabilities, we need to have those teams, we need to hire those people and we need to invest in cloud infrastructure, in the data platforms, in the software licences, in the teams. And also we need to invest in the time of every business unit to partner with them to block those agendas because all of those will be too busy. They don't have time. But if it comes from the CEO, I can guarantee you they will allocate time. And this is my experience until today, every company that I've been working on that is becoming successful. They operate based on these principles. And when I send an email to the CEO, he immediately unlocks the blockages, that sometimes the blockages are people not wanting to participate or unleash the information that they own and that they have owned for 10 years. But now this information is available to anyone and mostly to these AIs to bring value.

Moz Afzal:
I can already see a bunch of characters in my own organisation that would certainly have an impact there. Okay, so now take us to the future. AIs implemented, GenAI becomes very normal and day-to-day piece of our existence and our lives. Maybe walk us through what the future might look like in, I don’t know, 10, 15 years from now. And as me as the Chief Investment Officer for EFG, what do I need to adapt to?

José Pedro Almeida:
Well, I would probably first say maybe five, eight years, 10 years maximum. I'm pretty sure that most companies won't have all these softwares. The world of software as a service and all these enterprise softwares in my perspective is coming to an end. That doesn't mean they won't exist, but their central role nowadays where we have to enter 10 programmes and do a hundred clicks to accomplish a task, that will end. And what I mean is that the interface between the workforce and the company will be a single UX and that UX can be called through voice, or it can be called through a chat interface. Or in video where I have a call with a 4K definition AI agent that looks just like me, that I think that is a friend nowadays because I see that face every day and it is human and I just ask and it goes out.

It goes through all those softwares because if you think about this, this is 80, 90% of the job of anyone nowadays is just clicking things, passing things from email to Salesforce or to Workday or if I want to know how many vacation days I still have left, I just ask. I don't have to go to the HR system to know that. And so in terms of productivity and how we will work, I think this is the main point. And then it depends a lot on the industries. Just to cite an example, if you look at Tesla, which I think one of the main examples of operational AI in production, they have completely shifted their architecture that they were relying on 300,000 lines of code inside each car to take full self-driving forward. It never happened because it is impossible to predict all edge cases. And now the new Teslas with full self-driving, 13, they are running on a single neural network that just has cameras, watches your walls, and it takes actions and it drives for you.

And by the way, it drives a thousand times better than a human and that scales as a function of compute. So when they scaled from 100,000 GPUs to 200,000 GPUs, the network became a thousand times better. And so if you think that this will happen at Tesla that they are clearly ahead of any industry think how your company will be speed up by just plugging in more compute, I think that's the world we will live in. We will plug in more intelligence. I just want more marketing. It's just, it's just money. You are just hiring more marketing compute. I want more customer service, more customer service compute. You just unlock and unleash more agents that will answer your customers. And so in 20 years you'll probably see areas in the world where there will be only machine zones, there will be no humans on a factory. Construction site will be done by robots because the technology is getting there. It's not science fiction anymore.

Moz Afzal:
Yeah, I guess the vision and I think one KPI that I'm certainly very interested in is seeing the first billionaire, a single person business.

José Pedro Almeida:
Exactly.

Moz Afzal:
I mean that's really going to be, our eyes will be open, single man, billion-dollar net worth company would be, or market cap company would be something that from what you are suggesting could be feasible in a GenAI agent world.

José Pedro Almeida:
Absolutely. And by the way, it's not my quote, Satya Nadella “intelligence now comes as a log of compute”, which means that that individual, that billionaire will be alone. But in fact his organisation can have a thousand employees, but they are not human.

Moz Afzal:
Yeah, yeah. Well pretty utopian society, but something that we will certainly think about. But certainly this has been a very, very interesting discussion. Certainly has given me a lot of things to think about both in my organisation and within my teams and how I will gently prod our CEO to think about the future of AI. So José Pedro Almeida, thank you very much for coming on the podcast. It's been a real pleasure and hopefully we'll have you again on very soon.

José Pedro Almeida:
For sure. Happy to do so and such a pleasure and it's time for everyone to board the train.

 

This podcast is provided for general information only and assumes a certain level of knowledge of financial markets. It's provided for informational purposes only and should not be considered as an offer investment recommendation or solicitation to deal in any of the investments or products mentioned herein and does not constitute investment research. The views in this podcast are those of the contributors at the time of publication and do not necessarily reflect those of EFG international or new capital. The companies discussed in this podcast have been selected for illustrative purposes only, or to demonstrate our investment management style and not as an investment recommendation or indication of their future performance. The value of investments and the income from them can go down as well as up, and investors may get back less than the amount invested. Past performance is not a guide to future returns, return projections or estimates, and provides no guarantee of future results.