AI and Data – Techstrong AI Podcast EP41

Hello, and welcome to the Techstrong AI Podcast. I'm Amanda Ani, and with me today is Benet Samuel. He is the CEO of Zarus.

How are you doing? I'm good, Amanda. How are you?

Doing well on the show today. Yes. Happy to have you.

Well, can you share a little bit about Zetas? What services do you provide? Sure.

Zetas is a new type of, uh, data platform, which basically, you know, the, the fundamental capability is to automate information management. So where, um, you know, clients are trying to bring together data from different aspects of their business, um, and usually they wanna move it all into one big data lake or data store, um, and they have armies of people doing that. Um, we automate that process.

We make it, um, much easier. We actually use AI to connect to data across the business and in the cloud, um, and then transform it so that it's in a business kind of language, if you like. Uh, and then present it to either AI or BI or reporting, um, applications, uh, across the enterprise.

Uh, so the main takeaway in terms of our capability is we bring down the amount of work effort, the amount of pain it takes to integrate data for business uses across the enterprise. Great. Well, data is always at the root of many issues and company concerns, harnessing the data and storing it, getting the most out of it.

So what do you see as some of the biggest barriers or problems that companies have when it comes to this? Yeah, sure. Um, and, you know, the, this whole thing that I'm about to say is exacerbate, uh, exaggerated, I should say, by, um, the whole AI movement, right?

Um, where everyone wants to build an enterprise ai, they, uh, every CCEO out there, every board is, is trying to create competitive advantage by deploying AI across their front end, their back end, across the entire operating, uh, layer. Um, and the biggest barrier we are seeing is that clients, um, haven't even resolved their data for reporting or bi let alone be ready to be data-driven, AI enabled companies. Um, and so what, what that means is organizations are faced with, you know, the, the speed of data that we've never seen before.

So, you know, data across the organization being created by apps, by every customer interaction that's happening, uh, by their, their own channel partners. Uh, and so first it's the speed and the volume of the data, and then just the, the, the different shapes of data, data, the different types of data, whether it's videos, pictures, transaction data, um, the complexity in terms of the variation is, is huge. Um, and what usually happens is clients look at that and say, well, let's just focus on one project to fix one use case.

And if we fix that, then we'll get onto the next project. Um, and, and the speed of that internal approach where you're trying to manually bring your data together using the old, what's called ETL process, you know, extract, translate, load it into a downstream database, that that speed is slower than the amount of data that's being created and the speed at which, you know, data, new data is coming into the, into the company. So we seeing, uh, clients who are in a constant state of, um, you know, um, disarray for if you like, you know, as they fix one data problem in, in one part of the company, there's a whole nother stack of new data sets or, or use cases that, um, require, you know, a restart of, of the whole project.

So, so data in different shapes, data in different speeds, messy data, um, and, and, and the slow process that we think is the state of the art at the moment, uh, in terms of integrating data. All of that is, is in my mind, the biggest barrier to ai, um, adoption and maturity in in enterprises. Yeah, absolutely.

AI certainly does change things quite a bit. Everyone wants to implement AI and there are so many use cases promised with ai. So what solutions would you give to business leaders?

Yeah, so it's, it's really, uh, exciting for us. We've actually built a platform that enables you to do the end to end process in the one platform, in the one pane of glass, if you like, right? What, what we enable is you can use the tars to connect to all your data and create a single view of your raw data layer, right?

Once you do that, we use AI to transform it into a high quality business, language based, uh, data view, um, that then in turn can be deployed for many different, uh, uses. And one of the things we do within our platform is automate the deployment from the data layer into the AI framework. So, uh, some, some of your viewers may have heard of this RAG framework, this, this ability to actually intervene, um, uh, uh, within the generative AI process to intervene with your own private, uh, enterprise data.

So what we've done is we've connected the, our data, um, platform. We've connected that to a fully automated, um, rag capability, um, to be able to within one process or one set of pipelines, if you like, be able to go from your raw data that's at the source, transform it, deploy the rag, deploy the AI avatar. So we actually have avatars that you can deploy, um, so that that end-to-end capability.

Uh, we are not seeing many, um, uh, competitors in the market. That ability to create the single view of data ability to create a high quality, um, dataset, put governance around IT business policy in terms of how it gets used, and then moving that all into an ai, uh, framework that ultimately ends up in a chat bot or some kind of AI experience. Uh, one of our experiences is this avatar, to give you an example, we actually have an AI mortgage broker that's running within a large, uh, financial services organization.

And that AI mortgage broker broker actually acts as an assistant to all of the human brokers and takes away all of the pain in terms of understanding the bank's policy, understanding all the different rules in terms of, uh, product and lending. Uh, we have an AI uhlin, uh, clinical assistant, um, which basically does triage, um, uh, supports, um, uh, human nurses in terms of, um, that triage moment and t and taking away, again, some of the assessment, uh, need for lots of data. You know, um, when you're making an assessment on, on someone who's got a, a cardiac issue, for instance, there's a ton of data behind, um, that, that decision, including the cl the, the patient's own data, we take away that, that, that admin pain, that that, uh, knowledge, uh, requirement that's getting bigger and bigger using these ai, uh, agents.

Um, so it's really interesting for me, I started a data management company, uh, you know, 10 years ago, um, a company to, uh, automate data management takeaway all of the, the, the pain in terms of bringing data together in organizations. But where we've ended up is these avatars that are assisting humans in terms of AI capabilities, AI avatars that, that are doing role-based sort of assistance of, of humans. So we have, we have risk managers, AI risk manager assistance, we've got AI property managers, um, AI network, um, um, uh, assurance managers that are all based on the client's data, the client's data layer.

That's so amazing. So it's offering people the ability to make more informed, wiser decisions very quickly, quickly. Absolutely.

And, and, and beyond that, be, be able to move from just that chat bot kind of ai, like everyone wants to build a chat bot, but the real aspiration is how do we move from chatbots to, you know, these, these assistants that are actually role-based, that are a lot more mature. And to answer your question earlier, um, you know, what, fundamentally what do we offer clients in this space is, is not just, you know, we've got a set of AI capabilities that we can bring. We, we actually have a workbench to build any kind of enterprise AI outcome.

So usually what we do is we sit with the client's team and we work out what is the one, number one, number two or number three AI capability that if we could build it within six weeks, would revolutionize their company, change their business model. And that's, that's fundamentally our promise. We say to clients, we'll do an, do a Titas AI bootcamp, um, six weeks later, we will give you a minimum viable product of an AI avatar or an AI outcome that could fundamentally change your business.

Um, so, uh, it's an exciting time, uh, for zita and, and, and, you know, where technology has gone. Yeah, no, that's something I hear a lot from different business leaders that everyone's trying to harness AI and jump on the AI bandwagon, but sitting down and really brainstorming where is there a real need? Where would the use case best be applied is important.

Yeah, no, no, totally. Uh, and I see that, you know, that ideation thing happening every day in these boot camps that we're doing. Um, we do like a one day bootcamp or a three day bootcamp depending on whether the client has access to their data or not.

So if we've got access to their data, it makes it better and faster and easier, but we can still do a one day bootcamp. And in that bootcamp, seriously, you know, the, the, that concept of there's no bad idea, you know, e everyone comes to the table, people from the business side, sometimes they bring in, um, customers and clients, uh, and technologists. And I've seen it happen, you know, in front of my eyes where, you know, something that's looks like just the most ridiculous idea actually ends up being the main game.

And is, is actually the revolutionary, uh, idea to, to change, um, you know, that business, uh, in fact, this cardiac nurse project that we, um, we are doing, uh, where, um, what was an example of how in a, in a bootcamp, someone said, you know, the, the biggest problem we have is we treat every heart patient the same. And, uh, we, we, we treat their hypochondriacs exactly the same as, you know, someone who's actually having a heart, heart attack. By the way, I'm a hypochondriac.

Yeah. So, so we treat them all the same. Uh, and wouldn't it be wonderful if we could find a way to differentiate whether this person's having a panic attack or this person's having, you know, a, a genuine heart attack, um, and that whole it, it was, it was actually said as a joke, um, but it actually became the main thing that we focused on, and we ended up building, uh, a really valuable capability.

That's amazing. Okay. So once they get past the idea stage and they've implemented the ai, another con concern I hear a lot is how to track return on investment.

So what tips do you have there? So that's a, that's a great question. So our, our whole, um, model in terms of going from a bootcamp to understanding the ROI to actually coming up with a minimum viable to test a thesis, that whole MO model has gotta be, um, financially based and value based.

'cause because, 'cause what I've seen is a lot of these AI projects, in fact, um, you know, the, the reference is Gartner, A lot of these projects are failing. Um, 85% of, um, AI projects are failing. And quite commonly the ones that fail are the ones that are focusing on the technology.

I mean, I'm a technologist, some would say, what's he saying? But actually, actually my advice is, is to really understand the business side and understand where the money comes from. Like, and, and when I say understand where the money comes from, that, that's across every spectrum, whether it's a hospital or whether it's a bank.

It's, it's all about, you know, does this thing give back time? 'cause time actually equates, you know, as you know, equate equates to financial value. So this thing, does it give give back time?

Does it, um, alleviate a specific problem that's, that's driving cost? Um, so don't forget the business fundamentals in terms of architecting these AI projects. It's very tempting to, you know, listen to your tech, your tech team who have come up with some great, uh, eye catching, um, AI capability, whether that's a bot or a talking head or, or some sort of digital human.

It's very easy to fall in love with something that actually doesn't make any money, save money, or save time, you know? And so in our bootcamp, it's bolted in, we have a, um, a, a part of the bootcamp is, is an ROI modeling, um, uh, section where we really, we really ask ourselves the hard questions. Are we enjoying ourselves with some technology?

And are we getting the business to do something, you know, inverted commerce, something sexy, or are we focusing on something that's really gonna turn the dial in terms of, uh, business value? Uh, and it's fundamental, it's back, it's back to business fundamentals, um, as opposed to playing with technology. Well, AI is developing quite quickly.

So if you look two years into the future, do you have some prediction for AI that you could share? Well, you know, the first thing that comes to mind, there's a few things. One is we are seeing, you know, just about every day, two or three new LLMs or large language models that are being, or s SLMs, small language models coming out.

Um, I, I think over the next two years, what we are gonna see is organizations have a bit of a penny drop. That the idea of your business ip, your tat, uh, what's the word for it? You know, your, um, that that commercial knowledge you have that made your business is at risk, right?

In, in playing with all of this stuff. And we've already seen some, uh, early court cases and challenges happening to do with LLMs, public LLMs being accused of taking, um, private ip. So because of that, I think, um, more and more organizations are gonna say we need to build our own, um, language, language model capabilities.

We need to keep AI close to the, to the, to the chest and within the company's fences. Um, and we like that at Zaras because everything we do is within the company's control and is inside the company's fence. And all of the IP belongs to the company in terms of the business ip.

Um, so over the next two years, I think we're gonna be seeing a lot of small language models being developed by, by domain experts, by, by enterprises that are focusing on banking, telco, retail, who are experts in their domain, building out AI capability with the help of technology companies. 'cause I think the Penny's gonna drop that this is, you know, potentially, um, uh, an existential threat to some industries, to some companies, um, if they let that IP bleed out of their, their business. Um, so we're excited about it 'cause we have a model for any organization without the technical expertise to build their own AI capability and own it.

Uh, so that's one big thing. The other thing that, um, doesn't worry me that excites me is, um, a whole new framework for AI away from large language models. I think a new, um, a whole new, uh, set of capabilities we'll see in the next two years that will head a lot more towards this idea of general intelligence and true ai.

Um, you know, in my mind, what, whilst it's really impressive and it's changing every industry, it's, um, what we are seeing today is still a neural net technology that was discovered, you know, tens of years ago. Uh, and it's just that we have the compute to leverage it properly and train massive models. Um, I think with, um, the, the speed of compute, um, even this whole idea of quantum computing coming to the, the, the, the, the front of the industry, I think there'll be a whole nother type of framework coming out that will shock us all again.

And we, you know, maybe not the next two years, but I think in the next five years we'll see, uh, general intelligence come along, uh, and then that'll be another big game changer. And we'll have to rethink what business is all about and how humanity copes with it. It sure will.

There will be a lot more to think about at that point. Yeah. Well, if there was one key takeaway you could leave our audience with today, what would that be?

Uh, I think that the big thing is your biggest barrier to, uh, competitive advantage in this new world of ai, uh, and becoming a data-driven company is your data layer. Your data layer is messy. I've never gone into a company bigger than a, than a corner shop, uh, that, that hasn't got a data challenge.

So your data is messy, your data is in different shapes that you've got different types of data. It's scatter all over your enterprise and outside your enterprise. The sooner you can find a way to automate getting your arms around it and getting a single view of that data and transforming it using automation and using AI to transform it, the quicker you will get to those com that competitive advantage set of use cases and capabilities.

Uh, so the focus should be on, not just let me find a, you know, an army of people and bring them in and hopefully they'll fix it. The focus should be on what is the right platform and tooling. And, and let me tell you, Zetas is not the only one.

There's a whole new set, uh, of technologies and information management and a whole new set of data platforms beyond what we think is the standard like Snowflake, Databricks and Teradata and Oracle and all of that, they've all become, for me, they are legacy systems. There's a whole new wave of technology that automates information management, automates data layer development. Go and find them, talk to us.

We are one of them. And, and, um, and save yourself a lot of pain and accelerate yourself towards those high value capabilities and use cases. Wonderful.

Well, thank you so much for coming on our show and sharing all your insights today. Thank you so much, Amanda. Really enjoyed it.

Me too. And thank you to our audience. Stay tuned.

There's more.