Hristo Hadjitchonev: 0:03

if we have to compare what is going right now with the people that are more afraid about that technology or more reluctant or have any, some doubts about is it working or not? I would say that it's, it's not a question to be men versus machine. It's a question about men versus men with the machine.

Tom Raftery: 0:25

Good morning, good afternoon, or good evening, wherever you are in the world. This is the Digital Supply Chain podcast, the number one podcast focusing on the digitization of supply chain, and I'm your host, Tom Raftery. Hi everyone and welcome to episode 348 of the Digital Supply Chain Podcast. My name is Tom Raftery and I'm excited to be here with you today sharing the latest insights and trends in supply chain. Before we kick off today's show, I want to take a moment to express my gratitude to all of our amazing supporters. Your support has been instrumental in keeping this podcast going and I am really grateful for each and every one of you. If you're not already a supporter, I'd like to encourage you to consider joining our community of like minded individuals who are passionate about supply chain. Supporting the podcast is easy and affordable with options starting as low as just 3 euros or dollars a month. That's less than the cost of a cup of coffee and your support will make a huge difference. in keeping this show going strong. To become a supporter, simply click on the support link in the show notes of this or any episode, or visit tinyurl. com slash dscpod. Now, without further ado, I'd like to introduce my special guest today, Hristo. Hristo, welcome to the podcast. Would you like to introduce yourself?

Hristo Hadjitchonev: 1:45

Thank you very much, Tom, for the invitation. My name is Hristo Hadjitchonev. And I'm an expert of a lot of things. I have been software developer, engineer, three times entrepreneur. So right now it's called serial entrepreneur, things like that. But so I have my graduation in maths or more especially the maths behind the AI. Way beyond, I mean, before this term to be invented, let's say, or to be became a buzzword around. And I have a lot of experience implementing such things dating back into year 2000, really.

Tom Raftery: 2:21

Okay. And what are you doing now?

Hristo Hadjitchonev: 2:24

So right now, I'm I was invited and I'm helping a company called Transmetrics, which is in the logistics sector. First, initially came here and helped them to to speed up and boost their forecasting and AI platform to meet the customer demands in particular in logistics. And currently going further after that successful transformation. I'm helping the company to introduce the new version, brand new version of their core product Netmatrix version. I'm working on Netmatrix version 2 generally as a product project manager and generator driver and setting the, setting the scope of all that.

Tom Raftery: 3:03

Okay. Obviously, since the launch of ChatGPT last year, AI has become the, the hot new thing. As we, as we mentioned before we turned on the, the recorders, suddenly everyone's an expert in AI. So, you are actually an expert in AI. You've been at it close to 25 years now, maybe even longer. What are you seeing that are, you know, are kind of, The best use cases for AI in supply chain right now.

Hristo Hadjitchonev: 3:37

Supply chain is let's say it's a very huge domain. So you could you could take a look at the supply chain, like, the companies that are in fast moving consumer goods, it's retailers, which are purely digitalized retailers and et cetera, et cetera. So, logistics around all that as well is included. So you have a quite different perspectives around, around that, I will start with the very, very simple thing which is a little bit more like, fundamental, more more philosophical, let's say, and I will say, what is a prediction? And let's start from there. Let's say that people do not recognize, but actually. Probably 99 percent of the things they are doing on a daily basis, everything is actually a form of prediction. If you start looking how the kids start developing and knowing what is the world, they start doing some tries as an example. They have a cube and they try to put that cube in a particular place and they figure out that the cube actually fit or do not fit. And after that, they, they have already that knowledge and they know, I actually, they do prediction. They, when they take, took an object, they could predict where that object could fit. And going even to things like when you go to the work every morning before we do have Google Maps, after everybody was trying to predict which will be the best way, best routing, okay, based on experience. Here is a traffic jam, here is not, et cetera, et cetera. And I would say to the level of right now you're asking me things, but you have some form of prediction that probably I'll give some, some valuable information around that. I'm hoping so. So, to my understanding, and it's, again, it's a little bit philosophical, actually, everything else is a form of prediction. And everything is a form of prediction based on information we have from the past. Without matter is going to be done by human or by some other instrument or AI or whatever. But somehow the concept is that you learn from your past based on the... information you gather and experience that you have. And after that you manage to forecast this information with some cases for if for the future. And if I'm exiting from this purely philosophical thing, I would say that any business that that is probably the most and core part of the successful business. So either it's prediction, which, product is going to fit to the client needs. Either if you have your product is the prediction, what is the number? So generally that's a qualitative prediction. And the quantitative prediction is what amount of number of product you need for next week, next month, whatever, to satisfy the current demand. All this is form of prediction. So if we turn back in the past. It's generally the same. I mean, even, even if you go into the Mediterranean Sea into the 2,000 years back and things like that, they start predicting, okay, they're moving goods around where probably they have a demand. Nobody will just put some some wine in wine in Greece, moving that to Spain or to Germany without knowing that there is a market for that, which is, again, this form of prediction. Sure. Supply demand chain is a very sophisticated thing of the modern world to do everything to make, to, to, to give access to the, all the people around the globe to what they need. I would generalize like that. This fits everything again, fits to the, to this term, terms. I mean, you have to predict what your client needs and in what quantities, when, where. Simple questions, actually. And what is what's interesting is that you have, again, returning back, you have retailers. I mean, supply demand chain, but in supply demand chain, from one perspective, as an example, we have retailers. From another perspective, you have the logistic companies that are operating in that space and serving to the needs of the retailers. But even the retailers there could be very different. As an example, fast moving consumer goods, more traditional fast moving consumer goods, they have locations, they operate with stores. Probably, if they're big, they have warehouses and they have a huge, complicated supply demand chain including thousands of suppliers serving to probably millions of customers in some cases. I mean, when I'm talking about big plants like like, Tesco, like Kareve, and et cetera, et cetera, et cetera. I, on the other spectrum, even from size perspective, you have just a local store, which, again, is having the same need,. I mean, they, they need to know with what they have to satisfy the customer need. I work for one of the biggest platforms here in the Balkans about that, and they have a warehouse just because they supply about 15,000 SKUs. That means from diapers to vegetables, to beer, meat, fresh everything. So they generally compete against the, your local store or any, any store, but purely that they have pure logistics with them and they have huge warehouse. So for them, the prediction, what to have in their warehouse is crucial because operating with 15,000 SKUs, you have something like a lot of them are short shelf time SKUs, like fresh meat, as an example. It is 2-3 days. You have to be a very precise into the prediction about that because otherwise you're going to waste or otherwise you're not going to have enough meat for your clients. And the local store is optimizing generally the same, but not virtually having the warehouse there. And all of this has logistics around. So, electronic shops and any digital form of the shops right now has a massive impact on the logistics because you, you, especially after the, after the Covid crisis and the logistics has a little bit of a different meaning. So let, let, let's make the comparison between what is, as an example, what factors are important for fast moving consumer good, and to compare it with logistics. So, for fast moving consumer goods, obvious things are weather, customer patterns, and weather in terms of atmospheric weather. I'm not, I'm not meaning anything else. Seasons. A lot of seasonalities, but as well, you have impact from special days, like, as an example, here in, in the Balkans, we have a celebration around some of our as an example, Easter, there is a huge increase of consuming lamb. And this is really for few days. I mean, and when I said few huge, that means up to thousand percent in changes of behavior for just two, three days. So we have to, you have to, I mean, they know, but you have to still keeping the trends. And if you multiply that across many SKUs, that, that became a very complex structure of trying to, to know what is going to be next day, next week, and et cetera. And this is pretty the same with logistics. Because the logistic, again, the weather factor affects, I mean, especially the seasonality, and in logistics you have, because currently we are seeing a huge trend in logistics there is a huge increasement. It's almost more than half of the transportation actually is coming from any form of electronic trade. Generally people that are going to e shops. They are buying, and somebody has to deliver. So. So if we compare what is the, what is the challenge about the retailers and the logistic companies? And the difference with them is the fact that logistic companies, they operate currently, they operate with huge amount of goods that are delivered because of retailers. That was generally not exactly into the past. I mean, into the past the people were exchanging things and of course, companies, business to business and were exchanging. But right now it's a massive boom of of business to consumer deliveries based on the, based on the sales that was purely virtual, purely in the internet. And this massively increased after the COVID. I mean, the trend before that was huge, but right now. I mean, 50 to 60 percent of the packages that are transported actually are generated from generated from that. And this changed a little bit the structure for the logistic because they start having things like, I mean, huge changes into the into the transportation needs around special days. A good example is Black Friday. I mean, and you know that Black Friday is not just a Friday. It's could continue, I mean, the lasting effect is about a month. And another good example is Christmas as well. I mean, all days the people were going to the shop to find a present for their relatives and things like that. Right now, okay, probably they're browsing there, but majority of them, they're getting deals from from online. And after that, it should be delivered. Yeah. And so let's say the similarities are, as I said, again, you have impact from seasons, you have impact from special days and et cetera, et cetera. But what is the difference? And here I want to open a little bit of a bracket to explain that prediction is just part of the equation, because after that, any system has something called decisioning. So, as an example, you could predict something, but that doesn't mean that this, you apply your prediction in the real life. So generally you predict that as an example, tomorrow there will be a thunderstorm, but the decisioning is okay, the chance of that thunderstorm is about, let's say 30 percent and my decisioning is, do I'm going to get an umbrella? Or I will say, I don't want to have an umbrella for all day because 30 percent is okay. Relatively. I'm not scared. Might be not, but I will be discomfort to have an umbrella with me. But if the probability is not 30, but 80%, probably I'll get the umbrella. Let's, let's, let's transform that in, in retail sense, in retail sense of is a very good example is the packaging. So let's just imagine that we have a crystal ball. So the prediction is absolutely correct, 100%. You still have a huge issue with the decisioning. Because let's, let's, let's take as an example of something which is short shelf, shelf has a short shelf time. Okay. As a fresh milk, not condensed, not packaged, but fresh milk. That means three days, let's say like that. Yeah. Still the fresh milk is delivered in packages. So you, you, you buy a package of fresh milk that could, that has as an example 20 bottles of one liter. And if your, your prediction, which is absolutely correct predicts 27, you have to decide what to, what to do to order in your store one package of 20 bottles, meaning seven bottles shortage, or to order two packages of 20 bottles 40, meaning that you meet the demand. But after that, you have something that might be waste. And this is a tiny balance of the decisioning, which actually has an even bigger impact on the systems when you apply such kind of systems in your, in your supply demand chain compared to the accuracy of the prediction. Based on my entire experience, all the managers which are initially tried at, which for the very first time when they apply and they try to to use such kind of such kind of techniques, they, science techniques, artificial intelligence, et cetera they, they really focus about the precision of the prediction. And my experience is showing that actually the 80% of the quality of the decision it is much bigger compared to the 20 percent how precise you are in the prediction because, as I said into the case with 20 versus 40, this is 50 percent changes. Let's say if your prediction is 5 percent correct, you still operate into decisioning with something that might create a 50 percent change into your decision, which generally means that this 5 percent is irrelevant, small change. Yeah, an accuracy to the, an accuracy you could create into the decisioning world. But for retailers, this this is quite easy that perspective because for the retailers, it's the KPIs that they follow are, could be defined very easily and very straightforward, which means that you might have waste or you might have lost opportunities. And you try to balance between these two. Again, of course, this is simplified. You might have how big is your warehouse and things like that. I'm pretty in the business, but I don't want to get into such more details right now. During that, because we do not have time. But generally, the decisioning is a little bit simpler. Into the difference with logistics, where the demand is much more simpler because not a lot of factors are affecting the demand. As an example, in retail, you might have things like color trends. Let's, let's take a very good example with I have experience with apparel business as well. Okay. Right now, the movie Barbie. It will completely change for a few months, probably even a year, the amount of the pink dresses that will be the people will look, and I'm already seeing that by the way. It's going to be a fact. And this is a change that is going to happen. You don't have such thing into, into the retail, let's say that out. I mean, you have your office where some people are going to drop packages, or you have your contract with some retailer that will give you packages, but you have not to predict what will be the effect of the movie Barbie across your sales of clothes if you're a retailer in the apparel business. But the difference, as I said, I mean, so, the forecasting part is simpler into the retail, into the logistic business, compared to the retail, where the factors could be, you know, hundreds in logistics are much less. But the decisioning is much more complex afterwards, because as I said, the optimization into the retail, after that the decisioning is really to the, two three key key components KPI components. Rather, after that you have a very, very complex structure of decisioning. Generally the decisioning means, if I get a particular amount of packages and I have to distribute these packages across my network and they to reach our client into some form of SOS, could be 24 hours delivery or 48 hours delivery, etc. Actually, Any, even, even small logistic company actually has a very very complex network and very complex processes that are handling just this single package to move across. And a lot of decisions, with which truck this to be loaded, where, what to do with the distribution, etc, etc, etc. And the combination here could be could vary to a million, millions of possible decisions of the, of the, of the task. How to transport just a single package from one place to another, which really complicates the decision in part. And this is the, the, the huge difference. I mean, the huge difference between the logistics part of the supply demand management and purely about what I need, what kind of product, et cetera, et cetera, which is more very on the retail side of the supply demand management.

Tom Raftery: 21:32

Okay. And what is the uptake like of AI solutions in retail versus logistics or retail and logistics?

Hristo Hadjitchonev: 21:43

So, AI serves to, to the need in a similar way, even from, I mean, from a little bit of a different perspective. I would say the following, I mean, if we compare in retail as an example. If we compare a local, small local store, a good example. Generally, this is a meat store. Let's take as an example a meat store. So, probably the, the the seller there, who is probably even the owner operates with few SKUs, 10 SKUs, 20 SKUs, I mean, few types of fresh meat and probably some some sort of assortment of salami and things like that, and that's all And probably he knows his client personally, by the name. Probably they are, they're local neighbors and they know each other perfectly. This is, he's part of the community and generally the prediction that he is still, he has to have a prediction. But let's say he form a business where if somebody is missing his piece of meat today, that's not a big deal or things like that. But if you take the perspective of a huge chain. where you deal with hundreds of SKUs sorry, thousands of SKUs on hundreds of occasion, only multiplying the decisioning between these two. I mean, they operate with 15 to 30,000 SKUs and hundreds to thousand of occasions. So that immediately means that for every single day they're taking millions and millions of decisions to optimize their, what is on the shelves, what is in their warehouses. And I would say that with respect to the people, with respect to the humans, the amount of the decisions is so huge, that AI, I would say currently with the opportunity that AI is giving to the world, is that any retailer that wants to have efficiency, efficient decisions, and this is probably the key, AI helps you to take a really efficient decisions based on the history and to help the people to be much, much more efficient. So that's the, the very rudiment of, for me, at least as working on that field, key difference between pure human decisions and AI. I would say that if we have to compare what is going right now with the people that are more afraid about that technology or more reluctant or have any, some doubts about is it working or not? I would say that it's, it's not a question to be men versus machine. It's a question about men versus men with the machine. I would say AI in a way is going to change our work our world, but in a good way. I mean, it's going to help us to become more efficient, by the way, I'll give you an example. I work for a pastry shop pastry chain, and we've managed to optimize their waste really from 7 percent to amount around 1%. Six percent of waste, you cannot imagine how many tons of cakes thrown are on yearly basis. It's, only from that perspective, you have to, we have to start thinking that might be, it's not the only solution. But in using techniques like that will help to our environment as well. Logistic is a good example. Logistic, when I explained that the transportation, the decisioning part is the harder part. AI actually is helping a lot to the companies to become efficient, which is okay. They became efficient. They will make more profit out of that and et cetera. Or they will have more profit to invest, et cetera, et cetera. Because I mean, profit is not a bad thing. It could be used in various ways. But the important, the important part with that, all the efficiency actually reduce the carbon footprint that any logistic company is producing. And let's say like that, would I prefer to have a very polluted environment, not using AI, if AI is even helping me to have a much more, let's say not much, but more cleaner environment, definitely I would go for a more cleaner environment

Tom Raftery: 26:23

Sure, sure, sure. If organizations are. Interested in starting data science or AI projects internally, you know, to try and become more efficient or to make their local environment cleaner. To your point, what, what are good first steps?

Hristo Hadjitchonev: 26:44

Okay. So, there are two aspects of that. Any, any AI data science project is based on data. So probably the very, very first step is that the companies has to start collecting data and storing data. I, I, I had, I've met such an example. So this is not a rare example. This is quite often example. When, when I ask, okay how long is your history down? And they say, ah, it's wasting a little bit of a more this space here. And it's, we're keeping one year. And I'm saying but see guys, this is one of the more precious things that you have, by the way, your history of your, how the things I mean, how the world impacted you and what's changed with you and et cetera. How did you, I mean, the cost of this storage is it's zero. I mean, it's nothing compared to the cost of the information that you have into that thing. Probably that's the very first thing. I'm seeing less and less things like that, but in my experience, I had some few dramatic changes like that. I mean, they, we decided to start just a little bit about the financials because of things like that and etc. Why? You have to store everything. That's probably the very first thing. So, The data, the data is good, but after that you have to find what to do with that. So, a good formulated need is not often could come only from the business. And in that case, I will say that into the field, there are a lot of experts already to use an expertise as a consultancy to help you forming the idea of the transformation, not necessary to do the transformation itself could be very good. Of course, you could hire a person with experience that I mean, but this really depends on the, this really depends on the size of the company you operate in, etc. And I would say that already, even the medium and the small business has opportunities and of course for them they have to look for solutions that are within the space they operate because probably this case is repeatable and they might find a solution that is already working for their, let's say more in terms of scope, more smaller, a smaller context. And okay, let's assume we have the data and we have already oriented the goal. And here we have three opportunities and they have cost and benefits in general. So the very first opportunity is you to build an expertise in your organization. Especially the big organizations are almost always saying we're going to build an expertise in our organization. And I would say that that's let's say that's the most expensive approach. Sure. Because having a team of experts into the field is one of the most costy things that you might have into your IT organization. I mean, data science and generally the, the, the people that are in AI field are one of the most well paid. They're highly sought after. Yeah, nowadays this is going to be, let's say, a costly exercise. And it's not only about the numbers. So, the good, let's say, the good thing is that you have, you keep the knowledge inside. And you build the knowledge especially around your organization and your needs. And here again, it became the bad thing. And this again, based on my experience. The data science is based on the, based on mathematics. That's it. I mean, from ground zero, everything is mathematics. And the reality is that the answer, there is no answer, is a valid answer in mathematics, which means that you might invest in your personal team, spend a fortune. To find out the answer, there is no answer. At that moment, the business became absolutely frustrated. And they became, I mean, this is, the next the next buzzword, don't bother me, this is wrong, we have not to, to utilise that data set. So, if you're not really ready for investments with errors, with, let's say, big expenses, which, I mean, this is not an error, as I said. It's just, you discover one fact. That's it via mathematics. You might start looking for already existing solutions. So the already existing solution has the benefit that probably that especially solutions which are delivered as a SaaS, they have several benefits. So the very first benefit is that they, they do not meet only you as a customer, but a lot of other customers, which means that you have already quality product. And having a quality product you might have the restriction because the knowledge is not, is not on your site. You might have the restriction that, I mean, you're not exactly building the knowledge about that, but you have the huge benefit that taking that as a ready solution, even it has some constraints, not satisfying 100 percent exactly your needs it's going to, to, to fit the work to what exactly is promised. So, and probably that's one of the best approaches I would say. If you decide to go as a first step, use external expertise, use already proven solution, and experiment in a small scale or medium scale in your organization. And if this experiment is successful, sit with the numbers and start thinking, okay, it turned out that it's working for me and it's, I have a benefit. I have a return, enough in return investment into that. Right now what do I'm going to continue in that road? So. Utilizing more and more services into the area or ready products or I start building my own team. This is really, at that moment, depends on the on the decision.

Tom Raftery: 32:49

Sure, sure, sure, sure. Kristo, we're coming towards the end of the podcast now. Is there any question I haven't asked that you wish I had or any aspect of this we haven't touched on that you think it's important for people to be aware of?

Hristo Hadjitchonev: 33:07

Probably I will, I will, I will say the following. It's not about something you didn't ask. I will just probably try to transmit the message as a one of, let's say, not exactly pioneer, but for quite a while into that industry. I'm seeing right now the transformation, and I'm seeing that this transformation is going to go further more and more. I will say to, to the people, do not afraid to take that change. Let's, I will even use the bigger word to embrace, embrace that change. It's going to help. I mean, that change is for good. If anybody imagines Terminator or things like that don't afraid it's not going to be like that. It's not going to happen like that, or let's say, or at least in very close future, I mean, not very close future, future, future next few centuries. Me back in the, back in the days, the people were ruining machines because they were thinking that they're taking their job. The fact is that we're prospering because of that. Either at the moment when you're, when you're part of the change, that scares you. And any change is very something hurting. Probably that's the only message that I would say. Do not be afraid to take change. By the way, finally I will tell you the following. My, my, my wife is a university professor. And in university right now, they have a huge problem with ChatGPT because ChatGPT is used to, as an example, write thesis and things like that. Sure. The question is quite interesting, is it right or wrong? So generally the question is, you have, you gave an ice cream to your kid. And you're saying to your kid, so there is an ice cream here in front of you, but it's forbidden you to eat it. So I would say we have to find a way because this technology is going to be a part of our society going forward. Yeah. So, yes, it's not very clear how we're going to right now to accommodate in a perfect way, but sitting, the kid sitting in front of the ice cream. is not a good as well. Sure, sure, sure.

Tom Raftery: 35:37

My, I was on a, another podcast with a university professor. Oh. 7, 8 months ago. So just when ChatGPT was taking off and she was talking about similar kind of issues in the university and she was saying her university was thinking of forbidding the use of ChatGPT. And what I said to her was actually, you should require its use because if you require its use, then you've set a level playing field. Everyone will use it and the people who use it best will get the best results. And that's a good thing because they will have learned how properly to use the tool. And so that I think is probably the way universities should consider approaching this. Anyway, we're at the end of the podcast, Hristo. If people would

Hristo Hadjitchonev: 36:27

like... By the way, just one, one final remark, because it's, it's really interesting. It happened yesterday. My son is 16 years old. Okay. And went without me to Oppenheimer movie. Oh, wow. Yeah, so what happened after that? He was really impressed. I mean, he came to me and said, Oh, father, I know that you really like history and you, especially Second World War. You're quite aware about that. You still didn't look the movie. Go look at it after that. I want to discuss with you. And I said, Hey, about what? And he said, Yes, I used ChatGPT after that to make research and I would say that in that case, and he was really impressed about, I mean, we start thinking about the nuclear war, what, what actually nuclear nukes are, things like that. Is it good for the society? But he started developing the idea to use nukes to lift, instead of SpaceX, to lift ships into the into the cosmos, into the space. I explained to him that is. There is an idea about that, but moving ships across the space, not lifting them from the earth, and things like that but I will say he made it up, if ChatGPT was not there, he was not going to be able to crunch, as an example, using Google so fast for half a day, all that information to start this conversation with me. Yeah. Yeah, I hope you get the message. Yeah,

Tom Raftery: 37:56

I know. It's hugely impressive what it can do. It really, really is. Anyhow, as I said, we're at the end of the podcast, Hristo. If people would like to know more about yourself or any of the things we discussed in the podcast today, where would you have me direct them?

Hristo Hadjitchonev: 38:12

So, I'm going to send you the link of my LinkedIn and they could follow. I have in a few of the companies, I'm carrying some blogs and etc, so Please welcome to, to everybody to get a little bit more from what I'm sharing to the public.

Tom Raftery: 38:29

Fantastic. Fantastic. Hristo, that's been great. Thanks a million for coming on the podcast today.

Hristo Hadjitchonev: 38:35

Thank you very much for the invitation and wish you the very best as well, and to all your audience.

Tom Raftery: 38:41

Okay, thank you all for tuning in to this episode of the Digital Supply Chain Podcast with me, Tom Raftery. Each week, over 3, 000 supply chain professionals listen to this show. If you or your organization want to connect with this dedicated audience, consider becoming a sponsor. You can opt for exclusive episode branding where you choose our guests or a personalized 30 second mid roll ad. It's a unique opportunity to reach industry experts and influencers. For more details, hit me up on Twitter or LinkedIn or drop me an email to tomraftery at outlook. com Together, let's shape the future of the digital supply chain. Thanks. Catch you all next time.