AI Leadership Insights: Trends, Use Cases and Predictions with Saeed Elnaj

Hello, and welcome to the AI Leadership Insight series. I'm Amanda Ani, and with me today I have Sayed Elna. She is the CIO at Relay Group.

How are you doing today? I am good. Thank you for having me.

Happy to have you on our show. Can you talk a little bit about Relay Group? What services do you provide?

Yes, so Relay Group is a system integration and management consulting firm with focus on healthcare safety and security sectors. We, we have over 22 programs with the Center for Medicaid and Medicare Services, as well as with N-I-H-T-S-A and many other agencies. So we're based outside of Baltimore with the some 38 employees around the, uh, the, the country.

Uh, we've been supporting health IT initiatives and security emissions since 2013, and our focus is on EVA innovative solutions that leverage cutting edge technologies with, uh, to help the, improve the quality efficiency healthcare, reduce cost, fraud and waste that ensure best possible experie user experiences are the on outcomes for our customers. Wonderful. So our topic of the day, the day is AI trends and predictions, but first I'd like to know a little bit about, from your experience, how is AI impacting healthcare, for example, or any of the other industries that you touch?

Yeah, so AI is going to impact not just healthcare, definitely healthcare is one of the areas that it will impact, uh, in a substantial way, specifically with drug discovery and so on, but also with fraud and abuse, and I'll talk about it, but it really is going to impact many business, uh, many businesses, many business processes and various verticals and industries. It's not just going to be one single way. I'll give you some interesting numbers and some facts about the predictions with ai.

So it's expected that this year, 2025, we will have some 750 million apps that will be built using LLM using large language models that will automate about 50% of what we call the digital workforce, uh, processes. So that's substantial. 5 to $4 trillion.

And we're expecting this, the, you know, when you look at this number, this is the size of the UK GDP. So there is enormous value that's going to happen, and this is going to be impacting definitely many business processes, healthcare being definitely one of them. I think we look at AI today as the next foundational infrastructure.

If we think about the internet, if we think about operating systems in the, in the past, AI will be that foundational infrastructure that will enable many solutions and we'll transform business processes and we'll transform even companies. We'll see major disruption, major changes, and we can talk more about it and, and I'll give you a very specific, uh, example, but, uh, I'll, I'll let you lead with questions. Yeah, Absolutely.

Well, uh, you recently had a Forbes article and it was discussing, uh, not only the great use cases for ai, but some of the kind of misplaced hype around ai. Can you go into a little bit more detail about that and what are some of the key points from that article? So the, the, there were a number points from that article.

One point is that, yes, this technology has merits, it'll transform, it'll do amazing things, and I'll talk about some of the, the, the innovations here. Very quickly, we will see what gen, what gen I specifically, when I talk about AI now in the context of Gen I, we're going to see innovations that are amazing. Multimodality is 1, 1 1, 1 feature, which is we initially, when l and m started, when this technology started, we were typing texts, chat bots were the main way of interfacing with these lms.

What we are going to see is multimodality where text, video, audio, images, and it goes by direction. Meaning I could be talking to AI and showing it things and it can respond in audio or video or text. So the multimodality, that's something that we will see more of.

It mean, a good example is Google Gemini that now actually was built from the ground up to be a multi-modality, uh, LLM. And interestingly enough, it can see two things at the same time. That's one amazing.

And even the builders of this technology did not realize that this is the, the way it works. So that's one we will see also coming up. And soon we'll see, uh, agent ai, a lot of agents being, uh, deployed, and we can talk more about it as, as a, as a whole innovation, uh, side of, of, uh, gen ai.

And we'll see also tooling and we'll see, uh, a AI on the edge and AI in chips and so on. So all of this, there's tremendous innovation. I think 2025 we'll see amazing innovations in terms of gene ai.

And we're already seeing, last week we saw, uh, open AI announcing a deep research as an agent. And we can talk more about the, the, so that's, that's from one side, but there's also hype around the technology. So Gartner, a very reputable it consulting company, has what's called the hype cycle, uh, of technology, which, and it looks at it in different phases.

So initially when the technology is launched, it's very high and it's peaking at its, uh, and the peak of, uh, inflated, uh, uh, expectations. So there's a lot. People talk about it a lot and so on, but really the actual value from the technology is still hard to measure.

What we see is a lot of software companies building, uh, AI and gene AI into their tools and so on at times, really to detriment the user experience just to be part of the show and to be part of what I call with the innovation theater, right? So, so there is that, that form as well. We want to be in, in that part.

Uh, the, so there is also the hype about the value. Some of it is also self-serving. Nonetheless, I think all of this does not mean that the technology would not, it has to mean this value.

I think if we look at 2025, and we can talk a little bit about EEPs seek and the innovation that they brought to today, today, um, in the last few days, what, what it did, it really highlighted the ability of, of in how innovative we can be. Fast tracked AI probably about five years. It, it showed us that this technology can move at a very fast speed and it could lower the cost of ai.

While society is completely embracing AI on many levels, there's also still a lot of hesitation and concerns about security. And you brought up deep seek. So that's a good talking point, right there is, I'm seeing two sides of the coin with deep seek.

I'm seeing how amazing this technology is and all the innovation there, but I'm also seeing it's been banned. There's security problems around it. So what can you share as far as your insights with the hesitation about different forms of ai?

Yeah, so, uh, good question. I think there with deep seek, there's a lot to banzo, there's a lot to unbundle. One, one important aspect about deep seek, is it an open source?

So they shared their, the full code, everything about the model itself, it's available online. You can download, you can download it into a laptop and write it on a laptop. And in fact, what, uh, Microsoft Azure and AWS, what they did, they took the open source, cleaned it up, and made it available to their customers.

So you could actually run right now deep seek as an enterprise, as a company, or as a startup, if you wanted to run it, you can run it on these platforms, on these cloud platforms, and it would be safe, secure, and so on to the, to a limited extent. So, and a and we need to still, and this is a separate topic, but we still need, as enterprises, we still need to think about all of the guardrails that need to build around this technology. We need to build guardrails around, uh, toxicity to prevent toxicity, bias misuse, and, uh, hallucination the accuracy, like making sure that the, these systems can deliver accurate results.

So going back to deep seeq, yes, there have been issues and I wouldn't choose the open source one that is deployed and um, uh, and operates in China. Definitely not. There are all kind of questions and issues there, but if you wanted to run it as an enterprise, you can run it right now in a very safe way on these two platforms.

You can actually even download it, clean up the extent that you, you can have provided that you have the right, uh, skills to do so then you're able to, to run it safely. So there are always around that there are, uh, different, different aspects of it. I think the, if you ask me the most important thing about deep seek, it's accelerated the speed of innovation.

Again, some estimates by maybe five years. And I think if there's a lesson learned here is that open source is still a key player and that we as enterprises, as as government agencies and so on, we should not sit on the side and do nothing. We should look at it e evaluate these different lms, these different technologies and determine which ones are the most secure, the ones that I could put the right gare around there and deploy solutions.

Deep sake, pretty much, uh, a commoditized LMS made AI expung, uh, exponentially cheaper. And this is very important. This is from, from a a, uh, a user perspective or from an enterprise and a government agency perspective, this is, this is very important.

Cost is no longer an issue. Talking about government, do you have any advice for, um, how business leaders can make sure they're complying as new AI regulations come into play? 'cause I know we have a new administration and there was the new ai, uh, executive order among other things.

So what advice do you have for business leaders? Sure. You know, regardless of administration, I think the foundations and the fundamentals did not change.

You still, as a government agency, you need to build a God base. You need to make sure that you are compliant with the vice government regulations and, and protocols and industry protocols, whether it's Nest, whether it's SMA and so on. Whether it's, uh, from, from our perspective, one of our important customer is, uh, CMS, uh, center for, uh, Medicaid Medicare services, and we have, we handle a lot of very, uh, PHI and a and a and health, uh, information, uh, the data.

So all of that data needs to be controlled and managed and compliant with HIPAA standards. And so that does not change. That had to be there.

These laws did not change. And we need to make sure anytime we implement these solutions, we absolutely make sure that those cadres are built. Another thing to look at it is evaluating LLMs in ways that which ones are the most secure, which ones are the more, uh, compliant and so on.

And if areas of no compliance, we, we train it, we, uh, we provide it with additional information to make sure that we put the guardrails and prevent the misuse, the bias, the hallucination, the accuracy, et cetera. So that's, that's, it does not change things from my perspective. What it changes is actually the, the administrator, this administration is continuing on making sure that AI is still a very powerful tool that needs to be used.

So in that, in that perspective, there are no breaks on this technology. I don't see any change in that respect. And ai, as we know, is advancing very rapidly when it came onto the market a few years ago from there to now.

Wow, what a difference. So what do you predict for six months to a year from now? Yeah, that's the good question.

Um, so if you ask me like early January, be prior to deep seek, I would say, okay, the classical ones, multimodality definitely. So right now, if I am a government agency or of IM an enterprise that I want to implement this technology, I would think in the user experience in terms of multimodality, it's not just typing, it's talking to the technology. It's actually ar vr, putting on a, a goggle and being able to see the data and manipulated by moving and gestured, it's being able to, to, to view it in videos is being able to provide it with videos or images and so on.

So it's really all these, it, it's pretty much becoming almost like us humans, you, we, as we are talking to another human being where we're providing information and expecting answers. So multimodality is, is definitely going to be a big thing that's gonna happen. Agent AI agents, I think I'll, I'll give you a a good example with, with the research that was just released last week, deep research, basically sympathizes knowledge and creates new, new knowledge.

You give it prompts, you tell it what is the problem that you're trying to solve, and you just start, go research it, and in minutes it can solve a problem that fix humans hours and maybe days. So we are going to see a lot more agents. It is still in the early stage, uh, but like deep, deep research, it's be, we're looking at it and as in engaging a PhD level kind of research, uh, ability.

So that, that's, that, that's pretty amazing. The other area that we see is software coding. We're going to see software coding being transformed.

Software engineers are going to be different. It doesn't mean that we will need less. In fact, we might even need more.

There's the, uh, there's what's called the Jns paradox where the, the cheaper technology becomes, the more it's being consumed. And we're going to see the same thing here. So with, with software coding, there are about 24 companies right now, or a software, a code generation tools that are available to us.

One of them is rep. You can actually, on your iPhone, you could, you could, um, uh, type a, prompt a tool, generate an application for you to, so I expect that in, in, in the next few months, we'll see this accelerating at a very high speed. So that's another area.

Um, the, the, uh, AI on the check celebrity systems is one amazing company that actually took the, uh, the, um, Meta's Lama LLM model and put it on a chip, the sizes of a dinner plate. So, and, and we will see also other variations, like with Deepsea, now that it's very light model. You can put it on a small chip and it'll probably be what we call it, AI on the edge.

It'll probably be in our, uh, in our smartphones and in many other, um, uh, end user devices. So there will be a lot of innovation. I think the, the key to it is, you know, this is kind of what my advice would be, is not to sit on the side to look at use cases.

We've developed actually a full methodology within Relay Group on how to select use cases, what makes sense, business outcomes, do you have data, et cetera. And we see this with our customers. CMS being one example.

So having the right approach using a, a well tested methodology like the methodology we have, and looking at use cases, evaluating them and experimenting with them until the product is mature and solving a business problem and generating value. Wonderful. All right.

Well, if there was one key takeaway you could leave our audience with today, what would that be? Don't sit on the side experiment and use experts to support you with the, with the journey With ai. All right.

Thank you so much for coming on the show and sharing your insights with us today. Thank you. Thank you.

Have a good day. All right. And thank you to our audience.

Stay tuned. There's more.