Going From Spreadsheets to Smart Agents - Modernizing Supply Chain Intelligence

Welcome back to Data Driven, where we dive headfirst into the bubbling cauldron of AI, data science and the occasional existential crisis about digital transformation. In this episode, Frank chats with Itai Habber, CEO of Data Noetic, a company daring to bring order to the chaos of supply chain data. Forget dashboards and spreadsheets. Data Noetic is building an autonomous digital brain for supply chain operations. No, not Skynet. Though the temptation must be overwhelming.

From AI agents that flag delivery issues before they become disasters, to why your 3 month wait for bathroom tiles could have been avoided with better data orchestration, this episode is a masterclass in how agentic AI is moving from hype to hard results. So grab your headphones and your favorite supply chain KPI. It's time to get Data Driven with a dose of dry wit and digital

wisdom. Hello and welcome back to Data Driven, the podcast we explore the emerging ecosystem of AI machine learning, data science and data engineering. Now, my favorite is data engineering. Data engineer in the world is not here today because he is presenting at SQL Pass in Seattle this week. And I'm actually going to be at Microsoft Ignite this week. So Andy and I will be in the same time zone, but not the same city. But we must march on. So today I have with me the an excellent

guest. He is the CEO of Data Noetic and it's Itai Haber. How's it going, sir? Very good, thanks. Thank you very much, Frank. Awesome. Great to meet you as well. Thanks for scheduling this. And you know, we love, we love talking data. We love talking AI. When I saw the name of your company, I had to go back and relive some freshman philosophy,

Data Noetic. And I actually, not gonna lie, how to pull up ChatGPT because I'm like, I remember that means something around like not gnosis because that's more spiritual knowledge, but more intellectual kind of understanding. And it turns out it does. It's a, it's an ancient Greek word and it in noetic refers to wisdom, intellectual insight, and so on. So how do we get, what does Data Noetic do? How does it live up to its name? Right, okay, so the origins, I can't take credit for

the naming. That goes to our founder, Sandeep, who's been in supply chain industry for a couple of decades and has had the original idea of the

company. But I can definitely talk about what we are, what we're trying, what we're trying to do is basically Data Analytic was founded to become the autonomous digital brain for supply chain process optimization, automation and to jump to the kind of where data analytics comes from, it is in a way taking advantage of new developments in AI, machine learning, data science, etc. Which we might come to a bit later, in order to tackle a gap

that exists in a lot of organization at the moment. And the gap is currently between on the one side and the lots of transactional analytical data that exist in various places, data warehouses, data lakes, etc. And on the other hand, the same organizations have the process improvement initiatives, lean robotic process automation, et cetera. And those two things, the data that they have and the process improvement initiatives don't always

sync. There's no sync between them. And so data analytic aims to be, as I said, the autonomous digital brain for optimization and automation. By tapping into the data that exists in various systems in the organization, applying AI, agentic AI more specifically, or slightly more specifically, and trying to predict and suggest actions that can be taken, can be taken, sorry to, to improve things, and by so doing, orchestrating the data and making it actionable.

That's interesting. So, so getting to the brush tacks like how do you make it actionable? Like what, what happens? Do you, do you have like a UI where business user would use it, or do you have. Or do you enabled kind of data engineers to kind of work stuff and then surface it in a tool like tableau, power BI,

etc. Great question. The intention is actually to allow people who wouldn't necessarily be able to do all the data analysis on their own, so to kind of rather augment the ability of a business manager to take actions without necessarily having to rely as heavily as they might otherwise have to on the. On the business analyst that can go and query the Power BI or various other analytical tools that exist at the moment. And so to give an example of a prospect that we've spoken to recently, this

is a company without naming names. They are in the business of providing household. It's not appliances because it's like taps and syncs and things like that. And they had an order from a customer. Now one of the things that they care deeply about is delivering on time and in full for customer orders, also known as OTIF or otfd. On time, in

full delivery. And it was almost by coincidence that some process analyst has actually looked at the data and figured out that that particular custom order couldn't actually be delivered in full and on time because the particular item or items that they had in the order didn't exist, it would take too long to manufacture, to deliver, etc. Etc. Now that's great that they kind of figured it out ahead of it

actually happening. But that was the kind of exception that proved the rule that normally that information comes to light at the point at which the customer delivery order has already been kind of missed and on time, in full delivery was not met. Now what can be done and what data ethics helps to do is apply for example, what we call like the KPI guard, a KPI guard agent, which is basically an agent, think about it like a virtual assistant, a copilot as an example

that looks at the information that already exists. The information that the customer order has just been placed exists, the SKUs, the particular products that have been requested, that data exists not necessarily on the same systems. And here I go back to what I said about the lack of autonomous thing. The information about what exists in the warehouses

exists in some potentially warehouse management system, etc. Etc. And so by being a little bit more proactive on an ongoing sort of automated basis, it can flag the point that okay, this customer over here has just made an order for this

particular items that you don't have enough of in the system. And given the knowledge I have about what is happening in the, in your, in your business up to now, you will not be able to meet the delivery timelines that you have just told me are your, your effective delivery timelines. And, and therefore I'm alerting you that hey, this is an

issue. So now you can either try and if maybe there's an option to move some stock from warehouse A to warehouse B that would allow you to deliver that, or if maybe, maybe that's not an option. Another option might be, hey, why don't you reach out to the customer proactively and say I need to change the delivery date because of so and so. That's another example. Well, it's easier to have that conversation early in the process. As someone who's done a lot of home renovations and

more than I care to. I remember it was from a major big box hardware store. I won't name them, but we had these really nice like tile set up. But it took them three months to get this tile. And the frustrating thing I understood, it was stuck in customs, right? Or there was an issue with the supplier that I can relate to. But the fact that I wasn't told, I had to basically go through I don't know how many hours on hold, how many people to talk to,

right? That to me makes me like whenever they, you know, we do another project, I'm like, if it's not in the store And I have to order it. I don't want to do it right because I, you know, I had to, I had to hold up contractors and stuff like that. It was, it was, it was very painful. Now if they had told me straight up that it's going to take three months to get this, you know, instead of the normal 14 days, I would have chose a different tile or found a different

supplier. Like, you know, and then like to this day, every time I walk into that store, it kind of taints my like Absolutely, absolutely. You know and I think the amazing thing is that the, it's not like the information didn't exist. If somebody cared enough to connect the dots, it would have absolutely been possible. Now the actions that could have been taken once

those dots were connected, there are probably different things that they could do. They could have actively decided, hey, we don't want to tell Frank that it's late because we're worried that he's going to cancel the order. Fine, you can do that. But you have to take the risk associated risk as that particular operator that you're going to end up having a very unsatisfied customer. You might get the order not cancelled. Small gain. Shorter, but might have very meaningful potential. Oh, any,

any other tile job at. Since like unless they have it in the building, I don't order it. Like I go with somewhere else. Right. So yeah, I mean granted I'm not a big contractor, although I think, I think what my wife's second career one might be becoming a contractor. I don't know. But no, like it totally. You know actually just as we're recording this this weekend we had a, our hot water heater like basically flooded our basement. Right.

So I have to go back. It's very relevant, right? Because I have to go back and I have to figure out what you know, tile I want to put down or flooring. And I'm like, my wife was like what if we go to this store? I'm like, no. Exactly. But you're right though. Like it is a short term gain. But even if they, I would have been okay if they told me honestly because I, they would be in the running for any kind of future work. But I guess

people don't think like that. And, and, and the fact by the way, even telling you in, in advance might actually you might really want that tile. And you would say you know what, fine, I'll, I'll wait those three months but I will reschedule my plumber or my Tyler or whatever. Right. So that I don't up being annoying or, or kind of frustrating. Another call it process that you have a few, which is for you is renovating your bathroom or whatever, you can

adapt accordingly. Maybe you say okay, I'll do, I'll check, whatever. But I imagine that also would impact, you know, larger projects. Right. If I was a real estate developer or whatever. Right. We actually had a previous guest that, that does, you know, basically optimization for construction jobs. Right. Because if there's a delay, the project manager would rather know that. And we're talking like massive skyscrapers, like sort of thing type GC and like the UAE

and stuff like that apparently. So some of the work that that company had assisted on. But like you know, a delay of a day is like millions of dollars, tens of millions of dollars in some cases. Right. So if I am presented, if I'm a project manager, I'm presented. Well, you know, if going, going dark on the customer right now, obviously me, Frank as an individual is probably a way less influence over a supplier than like somebody who's building

skyscrapers. But you know, I would at least have, I would be, I would as a customer be able to make an informed choice. Right. I could be delayed by one day. I could be delayed by four weeks if I can't avoid the delay. I think I know which one I would pick. Right. I mean, I think, I think everybody appreciates stuff happens. Yes.

And it's just about the ability to be more informed about it. So you can actually take the appropriate actions about it. And look, to use another example, spoke to another customer, this time not in household goods, more in pharmaceutical. And there again actually OTFD was on time in full delivery was

a key factor for them. And there was an instance where one of the key executives went to their customers and proudly presented how their on time, in full delivery of the pharmaceutical goods to the particular healthcare provider was 90 plus percent whatever they, they thought it was only to be then told by the customer. Well actually no it

isn't. When we are, according to what we know, it's like whatever 70, 80%, whatever it is, whatever it actually was that the numbers don't make, don't aren't relevant for, for the purpose of the point that there was a difference between what the pharmaceutical executive thought that their performance was and what their actual performance was as reported by the customer. Obviously very embarrassing for the, for the executive coming

back into the organization saying what the hell is going on? What's going on here? That must have been an uncomfortable meeting or two. Absolutely set meetings actually in the Organization. And what they figure out actually is that as they were kind of summing up, the amount of time that it takes to provide the full delivery was

being done by different departments. Now, for all sorts of semi valid internal reasons, various departments chose what components to include and what to exclude from what they reported as for the time that it takes to deliver. So for example, this one department that counted the amount of days that it took them to get the thing from point A to point B, they excluded credit checks because credit checks is not part of what the department did.

So very kind of, which is, which. Is my point of view, look, not a customer pov exactly. Which is fair enough for the department which maybe is

doing, actually shipping the thing from, from the warehouse to the distribution center, but they can't distribute it, they can't do it before the credit check is done. Okay, so for the purposes of their work, yeah, it's true that the credit check is irrelevant and they shouldn't be quote unquote punished or, or, or in somehow in some way kind of made to look worse than underperformance than they actually were.

But for the purposes of the customer, the fact that however many one or three or seven days have taken an additional days for somebody in the finance team or the procurement team or whatever to do a credit check on, the customer still added those exact same days, which would then manifest themselves into the amount of time that it takes from the point at which the customer in their view made, not in their view, in reality made the

order until the point is delivered. And that is just one example of the sorts of discrepancies that can create problems and where what I'm talking about, the orchestration that we're talking about, the data noetic system, the data knowing system that looks at the various components kind of dispassionately and practically and kind of is able to give the suggestions, in this case, it would be able to connect to the CRM that maybe captures the date at which the order is made,

the financial system that does the credit check and then the warehouse management system and the ERP etc that track the various other steps that go along the way and give you a complete and hopefully more accurate picture of everything that's going on. Right. So, so what do you think is blocking organizations from doing this? Right. Is it data silos? Is it the fact that when these data systems, particularly the larger, the enterprise, when these were built, supply chains were not as complicated

as they are today? Do you think, you think it's a combination of those data silos organizational politics. You did say, you did kind of allude that, you know, the problem was some of the problems are valid. I can assume the ones that are not valid are kind of ridiculous internal politics within that organization. Or is there something else I'm missing? First of all, just in terms, I think the answer is that it's probably a

combination. And just to correct any misconception, I'm not blaming the organization for doing something that's outright ridiculous. I think that when you check each individual action or decision on its own, it kind of makes sense. But when you edit and aggregate, it creates a situation where you have an executive going saying, our delivery performance is X, where actually the delivery performance is worse than. Worse than X. Going back to why that's happening, I think it's a combination of

probably most, if not all the things you said. It's a combination of silos. It's a combination of kind of people looking a little bit different, people for valid reasons having a bit of tunnel vision. Exactly. And also

there has been, up until relatively recently, it's been very hard to be able to orchestrate all those things, which is something that the advent of various forms of artificial intelligence and machine learning is manifested by large language models and the increasingly amazing capabilities that AI agent building brings on board. Those things haven't been around. And so being able to connect all those dots that once you tell a story after the fact sound obvious. Like, why didn't your

tiling company know that. That they're going to be delayed? And why didn't they tell you that it's going to take three months? And why did it take you 15 calls to. To figure that out? It's all, yeah, it sounds kind of obvious, but the reality is I don't think that anybody in this company, in the company that the retailer that or the company you're working with kind of set out, okay, how do we deceive Frank? That's not. No, no. Absolutely no. No. And if I phrase the

question that way, I apologize. That's not what I meant. I mean, I think that what you described with like each little, each little error added to one big massive compound error. There's a fancy word, there's like a fancy word to describe that in engineering of complex systems, right? And the classic example

is like the space shuttle, right? The issues that they had, like, some people knew that, that, you know, whether it was the, the what, the heat tiles, whether it was the O ring, Some people knew, some people didn't know they. How to communicate. It was there Maybe some other things going on. Maybe. But you know, but, but you know, honest mistakes can happen and honest little mistakes add up to

one big. One big honest, you know, mistake. I, I really doubt that this company was, you know, you know, had a picture of me on their wall and it's like if this guy calls, like. But exactly. But I mean, you know, just the same, like, it's still frustrating. Right. And that's a great point that you brought up like up until now with Agentic. You bring up a great point about Agentic. AI really would make this much easier because the alternative historically would have been doubling or tripling

the size of your data analytics team. Right. And even then that's not a guarantee. But I suppose you could say the same about agents. Right. Like an agent that is operating on bad data. Right. Could also do some serious damage. Absolutely, absolutely. I think that is why, look, when we talk about, to use data analytics, just an example, and you can extrapolate from that afterwards what we are trying to do, we'll kind of try and think about it a little bit like a brain. There's

a left side, right side. The left side for us is what we call Data Pro V. Looking at the data processes and actually process value. So we use principles of value stream mapping and we are, and we're relying, we're not trying to replace the systems that you already have. So you probably already have an ERP system in place and a CRM and various other warehouse transport, various other management

systems. So it's not about ripping and replacing everything. No, you've probably made a decent choice and they're probably doing a good job of managing the particular part of the process that they were meant to deal with. But the problem is that they were all provided as point solutions and they don't necessarily talk to each other. And so up to now, what you needed to do is to somehow connect the data points

yourself. But going back to what we're doing. So dataprov is about first of all capturing the value stream map as it matters to you, to your process, to your supply chain, capturing the KPIs as they matter to you. Because for you, maybe cost is the most important thing, maybe on time, in full delivery, various other things. And, and irrespective of what that thing is, you probably also have a quantitative measure for what is good versus bad. One company's own time

in full delivery should be over 90, another might be 75. Doesn't matter, but

it's kind of your stuff. So that's kind of the Data proofy side of what we're talking about. And this is where it's crucial that the data that we are able to connect to the, to the data and that the data is valid because. Absolutely right. If you're saying, you are absolutely right in saying that if the data is incorrect, all the conclusions you're going to draw on top of it are going to be problematic. So that's on the one side. On the other side we've got what we call

dnai. So the data analytic AI part, which is where at the most basic level we provide you with some sort of copilot, let's call it, which allows you to interact with it a bit like a consumer would interact with ChatGPT or Claude or whatever the favorite LLM model is, which is basically ask a question in plain language and it should be able to give you a contextually correct answer. And in our case, in the context

of your supply chain, your supply chain data. So it's not about data analytics, isn't about asking it a question like, okay, what does the word noetic mean for that you have Gemini and whatever other tools. But if you want to ask, okay, how much of SKU 1 to 3 have I sold from the distribution center

in Baltimore over the last six months? It should give you the right answer that would otherwise have taken you and put you on the queue for the business analyst to interrogate the various SQL or other databases and give you an answer maybe in a week. Or if you get to the queue. Get access to and dig through 30, 40 different dashboards or spreadsheets. Right, that's the thing. I see,

absolutely, yeah, absolutely. And so that's at kind of a. Call it a basic level, but then you can take it and not chop, because that basic level of a copilot requires you to proactively ask a question. Whereas what an agent can do is, and you can have actually a set of agents that do a particular job for you, like what I mentioned as an example, you can have a KPI guard you might want. So let's take the case of dashboards that you rightly said. Lots of organizations have various dashboards and

various systems. And then those dashboards get complemented by those spreadsheet dashboards which collect information for all sorts of data points and some manual intervention, etc. They tend to be, okay, a weekly or monthly report that somebody sees and kind of, it could be that a week or month after the fact that you have breached whatever key performance indicator you wanted to meet, you get to know, oh, My costs have just gone 20% higher than what I need them to be or something like that.

At the base, at one level you can say okay, let me have a KPI guard that tells me as soon as a part of my process has breached a particular KPI against a particular guardrail or a boundary that I set, I want a notification immediately. And you can choose whether the notification is a slack message or an email or whatever else.

You can go a level beyond that and say, okay, I want you to actually, on a particular part of the process or a particular KPI, I want you to try and kind of simulate or predict basically what's going to happen and tell me if you think it's likely that I'm going to breach a particular KPI. Those things are. There's a lot of kind of work that needs to go behind the scenes and lots of ifs and ends and buts etc that kind of need to take into

account. But in principle you can see, I think it's kind of exciting that the emerging and constantly evolving capabilities of Gen AI and either various types of models, be it LLMs or SLMs or VLMs, whatever is relevant to your. In our, in our data analytics example in enterprise context, allow you to do things that have up to now been either impossible or very, very hard. Interesting.

So does it help, Is it fair to say this helps with governance? Right. There were discovery, not necessarily governance, but kind of the discovery like what does the agent do? In particular, does it. How do you discover all these different disparate sources? Is it. How much of a degree, to a degree is it automated? So this is if I understand the question correctly and I may not have. Phrased it right, so. I'll have a go.

I think you write those like let me try and raise the question a little bit differently and you tell me if it was kind of. If I got the general gist and can I rely on the agent to kind of. I'll exaggerate a little bit. And can you rely on the agent or agents to kind of save me for any. From any possible kind of fire drill or problem that I might face is one way of asking the question or another way of asking the question is how specific do I need to be in what I'm

asking the agent to do? Am I kind of roughly on the right track? Yeah, I would say so. Like that. That's one of, the, one of the, one of the aspects of it. But the first one I was going for is I get your product, I sign up. What happened what's the first thing that happens? Do I talk. You do get together with the business. Like who orders the product? Is the cto, is it the CEO, is it the. I don't know how many companies have chief logistics officers. Like who, who,

who you sell to. Is basically, it could be. There's a number of kind of levels of, of buyers, and it could be any one of the, of the following from the Chief Digital officer. Different companies have different names for it, but could be chief Digital Officer, chief Information Officer, kind of somebody who's responsible for the. What historically has been called the IT side of things to the systems management. Or it could be the chief supply chain officer. Which companies have.

It could be the layer below that, but doesn't matter about the titles. It's still the same functions all the way through to. It could be the Personas, the managers. It could be the product manager of a particular product in the pharmaceutical organization or written organization that can use the capabilities that we're talking about. So that's a little bit in terms of the types of users and buyers that we, that we're looking into, that we're, that we're working with The.

Sorry, that was the question about who we're dealing with. I think there was another problem. Yeah, no, no, that was really it. And then like, what's the first step? Right? Like, you know, say, like, you know, you or your sales rep have come to me and you explain it and I'm a, I'm a company. Whether I was like, oh, pretend I'm the executive that got kind of embarrassed by that thing, we need this today, we need this

yesterday. What happens next? Does the agent go out and search around for SQL Server instances and spreadsheets, or do you tell the agent, hey, I got my data here, I got my data here, I got my data here, and have at it. So there is a. I think it's probably before going into the specific process that we go through and kind of the steps. Yeah, sorry, I was just excited because this sounds. No, no, it's fine. It's

fine. It's great. I think it's worth maybe pausing for a second and doing a slight detour to talking about the evolving business models that are happening, I think, in the industry overall. And then I'll tie it back to how we're dealing with things. The, the

fact that AI is making the rapid progress that it is. I think, I think it's kind of fairly evident to, to everybody that we're talking about a, a fundamental technology evolution, if not revolution that we're, that we're seeing similar, if not at least as impactful as the Internet and cloud revolution etc and, and the same way that the advent of, of the Internet revolution or the cloud revolution has, has given birth to a new a paradigm of delivery which we all know is software as a service,

which replaced kind of client server software, the advent of AI is very likely to also usher in a different delivery model which is not so much going to be the software serve model, software as a service model, whereby there are those monolithic kind of systems that you more or less need to adapt to. Because the whole purpose of software as a service or the whole. One of the basic tenets of it was that

you kind of build it once for everyone and which means that everybody needs to adapt to you. These new models, there are different names being given to them, they're not all the same. But you might have heard of things like bespoke at scale or service as a software to kind of revert the SAS acronym or outcomes as a service. Those are all kind of different models that try and verbalize a changing a paradigm in,

in delivery of software in that context. Now, going back to how we're, how we're doing things, we are, we're not seeing ourselves as kind of charging on, on a, on a perceived basis. Not, not that I'm talking about pricing now, but the delivery is, is intended to be tailored per customer in the sense that when we get to you say you're excited

and you just bought the product, we will come. And one of the first things we you is a process discovery and data maturity assessment because exactly as you said earlier, if the data that you have is actually not going to give us sufficient information in order to make any decisions, we're going to fail. However brilliant the agents that we have are going to be later because the data is not going to be there. So we have to do this process discovery and data maturity. Then we have to

kind of connect to various systems that you have. We need to understand your value Stream, map your KPIs, your, your targets, ensure that all that is kind of

adapted for your, for your circumstances. And then we can start saying okay, here's maybe a library of a few agents that you can choose to use as is, or here's a sort of call it a canvas in an agent builder that you can take a few capabilities and build again an agent that's specifically tasked with addressing challenges that, that you have. So that that's kind of a slightly longer answer to your question about what happens next? Interesting, interesting. So it's almost like software

as an agent, right? You know, saw. I never heard that acronym before. But. So agent is a service. I don't know, there's different ways. No, that acronym has. Would be pronounced very awkwardly. Asian. So your website says you focus on, you know, the four main industries are pharmaceuticals and life science, Omni channel retail, consumer goods and what is fmcg? Fast moving consumer goods. Gotcha, gotcha. Okay. And logistics and supply

chain. But I guess any industry really has to rely on some kind of supply chain, right?

Correct, Correct. The reason why you're seeing the particular industries you just called out is that we think that the. If you have a relatively large number of products that you have to deal with in a relatively complicated supply chain, this is where the potential added benefits of having like proper orchestration and the assistance of AI agents and is going to be more pronounced if, if you have just one product in a super

simple process. Yes, you can benefit, but it's probably something you might be able to do kind of intuitive. Intuitively on your own or relatively LinkedIn system. That, that's why you're seeing the industries there which have the characteristics I said. I mean that makes sense. Right? Pharmaceuticals, life sciences, those are very highly regulated. People's lives are literally online on the trend of retail. I mean to compete in a world with Amazons and Walmarts, etc, you have to

be. You have to bring your A game. Right. And consumer goods, probably the same thing. Right. Because any consumer good or what is a fast moving consumer good? I. I've not heard that term yet, to be honest. I'm not sure where is the definition of what's fast versus not fast. It's just a. It's one of the definitions. I. Another term that I've heard for this is cpg, Consumer packaged goods. Okay. That, that. I know what that is. Yeah. I mean I would imagine something like food, right?

Yeah. There's a time component to a lot of foodstuffs possibly. Although I think, I think I am not sure of the look, not being an fmcg, I have not been in the FMCG industry myself, but I would imagine that they would refer to things like anything that you can kind of take and use quickly. A bar of soap. Oh, okay. That makes sense. Perish necessarily quickly. But it's going to use it. Within a week it's gone. It's. I think it also falls under.

That makes sense. McG as an example. That makes sense. Okay. Wow. It's fascinating stuff. And like you Know, I think one of the big concerns is about AI of late. Right. It's always fascinating me how the, the tech news cycle works. Right? It works and it finds something to grab

onto. It's like a little like, it's like a toddler basically. I have a three year old and you know, when he gets his mind on one thing, nothing else in the universe exists, you know what I mean? And I think the tech news industry like. Right, like so, you know, earlier this year it was agentic this, agentic that. Right now the last week or two it's all been about oh my God, we're in AI bubble. Are we in an AI bubble? Are

we like, it's almost like so. But I think that, you know, one of the things you kind of pull back with the concerns about AI bubble is the concern of how do you add value. How does AI realistically add value to organization? I would imagine that when you get your product installed and everything's working amazingly, you probably have pretty quick ideas in terms of how much time is saved in terms of analysts, how much more effective people can be. I mean, is

that something you see? Yeah, I think, look, there's a lot to unpack. What you said we can go back to the bubble and tech news maybe later. But in terms of the tangible results that you can get, it's. Yeah, I mean it depends on again, going back to the value

stream map and the KPIs that that matter to you. If you care for example about cost, you might find that transportation cost per unit is particularly relevant for you and for various reasons because it's been, actually because it's been hard to analyze the data, to collate, collate and synthesize the data from different sources to orchestrate it. Basically you haven't been able to achieve

the reductions that could have been achieved in transportation. So you can end up finding you got 5 or 10% improvement there. If you care about asset utilization, the same thing can be said for inventory turns and, or days inventory outstanding. Again, you can buy better orchestration of the data and looking into it, you might find improvements that are 10 to 20%, etc. Etc. And so almost every KPI that you, that you look at, there are bound to be

improvements that you can make. Some of them can translate immediately into capex savings or cost reduction or revenue enhancement capabilities, etc. Some of them are going to be a little bit more, I was going to say qualitative, but let's go back to the example of what I said earlier about the OT3 for the pharmaceutical companies, the fact that the executive came with the wrong number to the customer,

I'm sure there is a value to it. So they would like to have the right number if they have the right number, as opposed to the wrong number. How, how much exactly does that quantifiably? Well, there's trust. It's a trust issue, right? Exactly, exactly. It's a trust issue. Like if they're wrong, you start wondering if they're wrong about that. Yeah, I agree. What else are they wrong about? Right, exactly. And so all I'm saying is that some things will be very easy to

translate immediately to cash, to dollars. Some things definitely have value in dollars, but are not as easy or obvious to make the connection. But on the whole, there are, there are so many different places in which you can, you can see additional value here that it's just, I mean, the opportunities I think are, are endless. We can go back if you want. We can discuss. Oh, no, I think it's great because, like, you know, I, I've been, I've

been poking around agentic AI. I've been fascinated by it. But when it comes down to breast hacks, as people would say, it's hard to figure out what exactly would be a good objective source of value. I guess what we're saying is there's objective value, like hard cash numbers, hard dollar or pound numbers, because you're in the UK as well as, you know, kind of that soft kind of subjective stuff, whether that's trust, whether that's, you know, et cetera, et cetera, both are important.

But I think that, you know, if we do get into a situation where people are going to tighten their belts or the hype wave is going to crash, having hard numbers, yep. Is always, always good to have the hard numbers. Right. But I mean, I would imagine that, you know, and again, you're right. Like, you know, I would. Even within the same organization, I would imagine, like there are different metrics to track, right. Like, you know, whether it's time, whether time to fulfillment, cost

of transportation. Right. There's probably some kind of ecological things too, right. Like, you know, you know, we use this much fuel versus that much fuel, which again does tie to cost. But I think that there is a number of different. It's. It's. I think that over the last, say, 20 years, companies, supply chains have gotten orders of magnitude more complicated and the demands of a business environment have gotten orders of magnitude more

complicated. And the people, the headcount for the departments that would figure stuff like this out have not grown by orders the same orders of magnitude. And I think that AI, far from being this job taker, could actually solve a lot of these problems that people don't have the job anyway. Right? No, they're, you know, they're not gonna hide, they're not gonna go out on a hiring spree and hire like a thousand people to kind

of sort out the stuff. Right. They're gonna, they're gonna demand it of the existing or even less people. Right. So this is, yeah, I, I, I. Think that this is, look, we can, we can talk, let's

talk a little bit about the, the, the potential hypo bubble and, and let's talk about the, the jobs. So, so the bubble talk of the last how many days or weeks. In a way, irrespective of whatever my personal opinion is, it almost doesn't matter if there is a bubble or not. Because first of all, let's be clear, I think my understanding, at least when people talk about the bubble, they talk about the financial valuation bubble. So people will ask, okay, is Nvidia really worth 5, should it be worth

$5 trillion? Yes or no. And even if you subscribe to the notion that no, it isn't, and it's actually how much overpriced and so instead of 5 trillion, it should be worth 40 less, first of all, still worth 3, 3, 3 trillion and still a lot of money. And second, again, irrespective of how much Nvidia in particular is worth or open AI or any other company, it doesn't matter. There's a completely separate question as to whether or not the underlying technology that it is part of the, of

the industry that enables AI. Is that hype? Is it hype that AI is actually never going to achieve anything meaningful? I think that is a completely separate question and I don't hear anybody, and I would disagree with anybody who would say no, no, AI in itself, the actual technology, the actual capabilities are a bubble. It's actually meaningless.

As I said earlier, I think it's going to be at least as meaningful as the advent of the Internet or mobile telephony and the combination of which have enabled things like, like Uber and Airbnb and a lot of, I mean to name just two of like many, many applications that have made our lives different. No, that's true. Right. You know, when you look back and I'm old enough to remember the.com boom, right, the.com and the. Com

bust, right? And a lot of the things that these.com startups in the late 90s promised have come true. Right. I can. Pets.com didn't optimize their supply chain. Right. The cost of getting you dog food, they hadn't figured that out. But obviously Amazon does. I get my dog food 90% of the time as an auto delivery. Right. Because they can use. And it's not so much the technology aspect of it. Right. Because HTML hasn't really changed radically in that

intervening time. The backend systems have changed in a lot of ways. But you know, for the end of the day, I mean it was really the process. You know, Amazon built out a whole delivery network and worked out deals with other delivery companies. Right. So it is now, you're right, like it is now possible to do that. Right. It is now possible to call an Uber and you know, get, you know, get a car. But like the whole notion of, I think a lot of that

relied on, you know, having smartphones. Right. Because now it's easy to order stuff versus in the olden days you had to sit down at a computer, you had to wait for it to boot up, see the loading screen and then you had to dial, click to Internet, connect to Internet. Then you had to hear the screeching of the modem. It was a five minute process to get online, plus the page had a load. Now it's just a lot of people have broadband or certainly faster than dial up today. You know, it seems much

more feasible to do that. Like if I need dog food I can just, you know, even when I'm talking to you, even though I shouldn't because it's kind of rude, I could click open another window and say click order now. Right. But I think that's a long winded way of saying I agree with you because the, the promise of E commerce, the promise of the Internet has been fulfilled right now though how we got there was not the way that the startups in the 90s kind of thought. Right. But yeah, sorry, go ahead.

Yeah, so, so we're very much agreeing on the fact that the underlying capabilities and new capabilities that AI in its broader term will enable, I don't think anybody's questioning that. I think very few people know exactly how it's going to work out, but I think there's wide consensus that it's going to be fundamentally, it's going to have fundamental, a fundamental impact. Now part of the fundamental impact that people worry about is about jobs, which is what you said earlier

about oh my God, is AI going to take everybody's jobs? Etc. And look, again, I don't I don't know, I'm not, I'm not a prophet. There are valid arguments as to why AI might be, might be risky for some jobs. My white might be disruptive

to all sorts of jobs. But if, if we want to take the optimistic view, which also is, has a lot of valid arguments for one of which being every technological evolution or revolution has impacted certain jobs, but by and large created more opportunity, more jobs, more advancement than it, than it created, to use my favorite example is just because I can't remember where I came across it, but the word computer used to mean a person that did computations. Yes, that's right. And, and lo and behold,

computers. Now when anybody says computer nowadays, they don't mean a person doing computational because we've got machines that do that. So if you want to find a job as a computer, good luck. Right, right,

right. Gonna be quite hard. But does that mean that kind of nobody can find a job anymore? Absolutely not. Is that, is that a promise that, that, that's exactly what's going to happen with AI? No, but at least the trajectory up to now has been one of, I don't know to call it like a positive, positive trend going forward.

And I think, I personally think that the, exactly as you said, at least some of the capabilities that AI builds and the examples that we talked about about what data analytics does and being able to give you this KPI guard or various other agentic capabilities, I'm not necessarily seeing it, in fact, I'm not at all seeing it as kind of taking away anybody's job. I don't think that the business analyst that currently can't deal with all the tasks that they're being given

is going to be replaced. I think that they are going to be helped by both them. So they're going to be helped by it. And more importantly all the business managers who up to now just wouldn't deliver what they needed to deliver because they didn't have access to the business analyst. Now they have access to an agent that is able to. Yeah, I wish Andy was here because Andy has a really good anecdote about how

DBAs used to be. Like you would have a database administrator and typically you had, it was a one to one relationship. Every database had one DBA and then sometimes a backup if it was important enough. Right. But now the job of a DBA is they realistically manage dozens if not hundreds of databases. Right. And that's because of the cloud and automation and things like that, even before AI. But the job of a DBA still exists. Right. It just looks really different. And I agree with you.

I have faith in the trend line. Right. Historically, every aspect of automation has created more jobs over the longer haul. And my only concern is irrational. There's irrational exuberance. Right. But there's also what people don't talk about as much as irrational pessimism. Right. And that was, I

lived through that in the dot com bubble. That's the part of the dot com bubble I remember the most because it was the most difficult where it was, oh, you know, the Internet's just a fad and like it's over and it's like, you know, we, we don't laugh enough at the, the people who said that. Right. You know what I mean? Like, you know, so to your point, right, like, you know, is, is Nvidia worth really worth $5 trillion? Is it worth. Who knows? Today it could be like worth seven. Right. I haven't

checked the markets, but. But it's certainly not worth zero. Right. It's certainly not worth like I can easily see kind of the way the clickbait machine works is, you know, we go from they make the money on the roller coaster right up and they make the money on the roller coaster right down. Right. Irrational exuberance, irrational pessimism. And that's kind of the, there are dangers to both. But the part that at least traumatized me more was the, the way down and how far

that kind of went in the other direction. That's the only thing that would keep me up at night. Yeah, look, there's no,

there's no guarantee, I think that we are, it's possible, let's phrase it not in double negative. It's possible that we are in a financial bubble and it's possible as a result that at some point there's going to be a correction and that correction might be a short and sharp kind of decline or it could be a, a long and steady decline. It could be all sorts of things. Things. And whichever, if it, if it does happen, whichever form it takes, it's going to carry with it

something. All of that is under rival if it goes down that path, irrespective of whether it does that or

not. In the same way that happened with the dot com boom and bust of 2000, the late 90s and what then happened in 2000 and a little bit afterwards, the boom and bust, the financial boom and bust and the absolute roller coaster that the NASDAQ had and the implication, the financial, very real financial implications that it had for people didn't change the fact, as we've said and agreed, that the Internet, or the Internet, which is what created the to begin with, has

fundamentally changed the way a lot of things operate the same, absolutely the same will be true of AI. I have no doubt about that. I think that in the same way that the Internet has had a lot of positive impacts and some impacts that people would argue are actually

not that positive, I'm sure the same will be over. And hopefully again, as with the trend up to now, hopefully we can all, if everybody does everything they can in order to maximize the positive and minimize the negative, we have a very, very, we can be very, very optimistic about the future.

And, and, and to give a few examples, I mean, AI in theory holds the promise of helping us solve really, really hard problems like climate change, like the world hunger, how do we feed the population, how do, how do we manage resources in a, in a, a, an earth that is limited in size, with a population that keeps growing, how do we fight cancer, et cetera, et cetera. Those are all things that theoretically AI should help us kind of increase manifold. Yeah, so there's, you also have to think too.

Like the cognitive load that we have today could be reduced. Like if you're a business analyst and you have a book next, you used to have a book next to you. How do I do this in SQL? Because they don't want to wait for the data people, right? How do I do that? How do I do I go look and I do, I go, I do a Google search, right? And you know, Stack Overflow would have a billion different answers. Well, two or three different answers and then 100 people, anytime

you post a question would chomp on you. Like check through the existing answers rather than. Whereas now you go to ChatGPT. How do I do this? And it gives it to you right now. Is it always accurate? You know, obviously there's some rough edges there, but for the most part, you know, if you run into a problem in an unfamiliar space, it's a lot easier to get an answer now than it was before. It's a lot

more time efficient. So if you think about the cognitive load that now can be shifted to actual other more pertinent problems. I don't know. I see that as a net positive. I think we both agree, which is cool. That's always nice. Brilliant minds do think alike. I know we're coming at the top of the hour, so where can folks find out more about data noet and about you? The obvious places. So for data analytics, best place to start is our website which is datanoetic AI. So data D A T A N O E T I C

A I data analytic AI. The website about myself. I'm on LinkedIn so I spell my name it A Y. Like Italy without the L and last name Haber H A B for Bravo E R. You can find me on LinkedIn, you can find our website and happy exploring from there on. Awesome. I think this is great. I think it's the reason why I bring up the bubbles and is I. Think that. What your company optimizes will give people the hard numbers to kind of splash cold water into the irrational pessimism.

That's what I think. Because I think if you're an AI company today, start thinking about gathering those hard numbers. Right? Because those hard numbers are going to be. I mean that's how Amazon survived. That's how any of the survivors of the dot com crash, they had the hard numbers to prove it. Right. And most of the major Internet companies today, not all of them, but a lot of them had the survivors. There were survivors in the dot com bust, right? Absolutely.

And they came out stronger for it. I think Amazon being the most notable. Amazon being the most obvious. Google was started kind of in that era and they're a major player. And so I think that, yeah, just remember, you know, the sun always rises, right? No, I think again, we're probably violently agreeing. If you're able to deliver real tangible outcomes,

you'll weather the storm. You'll, you'll be able to, you'll become indispensable as people find Amazon at the moment at a consumer level and actually AWS as an example at the business level. So yeah, absolutely. Well, I think on that thought, thanks for having, thanks for coming on the show and what a great conversation. We'd love to have you back sometime and we'll let our AI finish the show. And that's a wrap on

another illuminating journey into the dataverse. Huge thanks to Itai Haba for joining us and proving that AI isn't just about generating cat poems or pretending to write your emails. It can actually prevent multimillion dollar supply chain nightmares and possibly, just possibly stop Frank from reliving tile related trauma. If you enjoyed this episode, be sure to subscribe, rate and leave a review because somewhere an AI agent is judging you

based on your podcast engagement. Until next time, keep your data clean, your models lean, and remember, in a world full of dashboards, be the agent of change. Bailey signing off with perfect on time in full delivery.