Mike Vizard: Hello, and welcome to the latest edition of Techstrong AI. I’m your host Mike Vizard. Today we’re with an Anar Mammadov, who’s CEO for Senpex. And we’re talking about the impact ChatGPT is going to have on logistics. Anar, welcome to the show.
Anar Mammadov: Yeah. Hi, Mike. Nice to have me here. So nice to see you.
Mike Vizard: You cannot walk down the street these days without somebody jumping out and telling you how they have used ChatGPT to transform something. But what exactly do you think the impact will be on logistics? There’s a lot of paperwork, obviously involved. But what can we do today that maybe we couldn’t do yesterday? Or didn’t even think of?
Anar Mammadov: So yeah, that’s actually a great question. So I will tell you we’re living actually pretty much in challenging times and interesting times, which is like – ChatGPT will change, probably, completely, most of our, you know, the life, the business, the way how the business runs, et cetera. But yes, so in my opinion, we have to use it; we have to use those new technologies and new innovations – that’s 100%. And all those special AI based tools have been invented, for the purpose of efficiency in any type of the business processes, any type of the businesses you want to scale and grow, to enter new locations. But yeah, so my initial, very brief answer to that is 100%, we have to use it. And we as a company are also implementing and using it. And we can talk a bit later, maybe, about that. But that’s very interesting. I think we have to implement those tools and software; its about automated processes and making as much possible efficiency, because right now, specifically during this recession and challenging times, those innovations and improvement and efficiency, and managing your expenses is very important. So because of that, those AI tools are pretty much becoming popular and the people and the most of the businesses, they have to use it. That’s my thoughts overall.
Mike Vizard: So well, let’s jump into that. How are you guys using it? And have you been experimenting with it? Or are you actually embedding it into some sort of process?
Anar Mammadov: Yeah. Well, initially, there’s a lot of talks about that and I didn’t pay too much attention. So, but we are the last mile logistics company. And I do have right now three data analysts, data scientists, basically, who are actually managing all my data just for making the proper decision. So basically, we took charge of building for the new version of the API and fully integrated to our last mile logistics platform, and also integrated to our data analytics dashboards, which we use like Power BI. And the way how we’re using it right now is mainly for one of my challenges. Initially, we’re working with businesses, and the businesses that we are working with, like the health institutions, the food, like restaurants, like the e-commerce marketplaces that were selling the software products, and the logistic as a service. One of the main problems is in our sales department – there’s a lot of followups. So basically, for the sales automation, we implemented ChatGPT. This is a great tool for like arranging, followups, meetings, talking with it, you know, emailing our clients, and what we integrated with ChatGPT was our internal CRM tool, the customer relationship management tool, and then also analytical dashboards. And plus, we took the historical data of our main core dashboard platform, and integrated all those with ChatGPT to get some feed information to ChatGPT to talk with our clients. So basically, we integrate integrator for the purpose of increase our sales, because usually manually our sales team like slowly is falling out with each of our clients, like within one hour, like two, three clients. After implementing ChatGPT, we increased this performance, like, within like one hour, there’s a lot of things happening right now with all of our clients. And if the clients were applying back then our sales was jumping to that, and then just continuing their communications etc. This is the one process where we implemented and second we are pretty much advanced in-route optimization engine. We also implemented a lot of AI optimization overall, lots of planning tools, where we also implemented ChatGPT for the purpose of finding out the efficiency of our optimization. If you have let’s say one or two drivers with let’s say a thousand stops needed to be properly optimized and routed properly. So in that case, we are implementing the ChatGPT for that purpose as well. This is initially the beginning. Still, my initial feedbacks overall about that is like, ChatGPT is great, but still in the learning curve – that’s my thinking, because there’s some mistakes, there was not proper verdicts. But our goal was using ChatGPT to create as much possible personalized sales automation for our clients first – that’s actually the main goal.
Mike Vizard: How is that better than say, the maps that we get from Google or everywhere else these days that people have been using? And usually, if I’m going to someplace well known, they do fine. And if I’m going to someplace that’s a little more obscure, that last mile usually has some weird directions in it anyway. But um, how does generative AI kind of change the way we’ve been using digital mapping systems anyway? And what did you expect to see?
Anar Mammadov: Well, I will tell you, like I’ve worked for one mapping system for almost 18 years. So I’ve seen, historically, how every single change – like I remember, we started initially with systems – like manually mapping. So the different areas in Europe and also here in the US, I had some experience. So basically, the mapping is actually – there’s some kind of efficiency behind the mappings and then there’s a regression analysis; there’s a lot of things that make for sure as short a period of time to reach from point A to point B. That’s actually the main goal of the all those mapping systems. And still, I respect the Google Maps, I respect like, like open source street maps – there’s actual open source platforms that exist as well, besides Google. But still, there’s problems there. Like even at Google Maps, like whatever ATA, they’re just telling you, it’s just an approximate number so that you cannot get the exact, like the estimates about the timing that you need from point A to point B. So in my opinion, like this mapping system is still growing and improving based on using different AI tools based on the different sources and feeding more information of the location to the maps. So they became advanced and advanced, but with ChatGPT, that will be completely different on the way of like automation of that process. Because right now, the ChatGPT is all implemented for the text messaging. But if you’re not imagining, that will be implemented for the images purposes, to take the pictures of the world, or the location, whatever. And there’s some kind of device we can actually manage through ChatGPT. But that will be completed a different level of the things. But yeah, overall answer to your question, in my opinion – there is like a paid version of like Google Map systems, and also open source mapping systems that keep improving, but ChatGPT will completely change this direction as well for the mapping purposes and routing purposes.
Mike Vizard: So essentially, rather than just having a map that says I can get from point A to point B, I might get a map that says, here’s the way from point A to point B. And then by the way, here are all the best Italian restaurants along that route, when I or whatever else I may decide that’s of interest, or, in the case of logistics, maybe here are all the issues with historic traffic issues that might influence your decision.
Anar Mammadov: Yeah, exactly. I mean, right now, there’s a lot of AI’s that the Google is building behind their mapping system. But in my opinion, there’s still like improvements that need to be done. Like it’s not 100%. And like calculations, and over estimation, whatever timing they’re giving us are not like, exact, accurate as it’s expected, right? I mean, I don’t know if that answers your question about that part. But that’s my initial thoughts. And I’m just talking that as not like, like as a manager, but I’m talking about, you know, the person and scientists in a data analytics direction. And I see a lot of gaps. And I’ve met a lot of like Google guys who works for Google Maps. I know a lot of friends actually, who – I’m actually right now in the valley, Silicon Valley. So I met with them. And the technical team member feedbacks are like, still, there are some issues. The gaps need to be improved on the Google Map.
Mike Vizard: What is your sense of what is the expectation that your customers have in terms of your capabilities with AI and data science in general? It almost feels like to me we’re reaching a point where, rather than thinking of this as some great new capability, a lot of folks are coming to the point where they expect it.
Anar Mammadov: Well, I mean, the industry that we’re working with – we specialize in last mile logistics where we are. We’re in more than 15 states right now; we’re growing with more than a thousand overall niche vehicle drivers, which is, which is who we’re managing fully, but basically answering your question, we are – our clients expectation’s are first – providing the best customer service, because we’re in the service industry. So to provide the best customer service, you have to make sure like, you get a reliable service, reliable drivers, reliable dispatch support teams who can actually support you in case there’s any issues. And then the technology behind that with the proof of the deliveries, you know, apps where you can track all that experience. And the last pieces are, like our client’s expectations are analytical dashboards, so they just want to see their support for transparency purposes, what’s going on with overall, with my delivery experience, right? Like we’re providing special analytics for that, and which is a very important piece, like, from the clients perspectives. But from our perspective, we are -so definitely efficiency is important. Like with the minimum resources of my team members, I have to scale and grow the business, that’s actually my expectation overall. And I prefer these type of businesses – like, if the business can be easily duplicated any locations, without your presence or without, let’s say, like, that’s actually how you can grow. That’s how you can grow, like scale the business. So for that purpose, we implemented the AI for like dispatch, controlling management, for drivers controlling management piece, like proper routing and planning, the routing piece of that. We need to implement those AI tools to get as less human involvement to the process and make sure the process is automated properly, on the way right? And this third part is like, setup – the organization structure is also very important, like you need to have as a company, like, the businesses and the tools that we’re using to make as much possible efficiency on the way of how you’re executing the tasks. So like, you know, I’m talking about the proper CRM tool, the proper internal dashboard management tool, then if you have it, an electrical dashboard tool – you need to get it. So you have to put all those needs, to be integrated with each other through API. This is also a very important piece, in my opinion, for any businesses to get transparency on the way of insights, of the data first, and a second is like to grow and scale their businesses.
Mike Vizard: You think people will fully appreciate the challenges of route optimization. It’s always been something of a gnarly problem. And a lot of folks in the age of digital transformation are making some assumptions about logistics that perhaps they’ll find surprising once they get in. So, what advances can we expect in the area of routing algorithms as we go along? What have you seen your team working on? I mean, are there things coming down the pike that you’re looking forward to?
Anar Mammadov: Yeah, so I will tell you that these things like route optimization – there’s a lot of algorithms behind what the route optimization is. There’s probably like a thousand algorithms, and you have to work on those algorithms to make sure it’s properly optimized. So like, our engine, we have started on building like, within, like, let’s say six months, right? We build this for almost like six years, based on trial and errors. So you have to upload the route to see if it’s properly optimized. So there’s a lot of like algorithms, like, for example, the optimization divide – let’s say, for instance, like sales, whatever, right? There’s also like, maybe we have another options, for example, optimized routes from a warehouse – start from the warehouse, go to the front, cause by location to go back to the long distances, or you can optimize routes on the way of like, start from the long distances and make sure it’s returning back to the warehouses. But there’s a lot of a lot of variables when you’re building the optimization engine that we keep considering those, and there’s a lot of mistakes happening. But overall, the main goal of those optimization engines are like build the efficiency, like efficiency on the way of like how to make sure within like, a given time period. So let’s imagine the use case – I had the challenge before, like in 2018. I got one project, actually in Oakland, California. It was a special client. They called me saying I need like a thousand stops within two hours with 25 drivers. Can you make that happen? That was actually – I think it was like a Mexican restaurant, they’re special burritos. They’re just preparing and they’re saying okay, it’s a perishable product. It needs to be delivered within two hours, but because it’s a special event, it goes to a thousand people, within a two hours time windows, like, please find me as much possible drivers to make that happen. So that’s actually pretty much a challenging time where we came in on 2018, and we haven’t failed. But out of that we built actually our product, which actually created huge efficiency of that. So our logics right now on the optimization, divide as much possible drivers with limited number of stops with a limited number of the time that we have as a time window, to make those deliveries happen properly. So that’s how we implemented this efficiency, so route optimization. And based on that improvement, we started building a new algorithm and optimization. customizations, probably how it’s actually we call it route customizations – like it’s clustering within the proper logics. Yeah, and divided between multiple drivers. And those deliveries need to be done properly. And at that time, the challenges were the limitation, even the Google Maps API. So we also reached out to them. Hey, guys, please, because we haven’t had this type of solution before, can we get their API, and they actually had some kind of limitations, like 25 stops, no more than 25 stops, that’s actually kind of created a bit like, like even we wanted to pay Google at that time. Please help us with that, because we needed even more. But now our optimization engine can optimize, within one driver, even more than like 100 stops, like within one route, like the person can, can execute the task, even with more stops. So this is actually one of the challenges I had, but answering to your point – this route optimization tool keep improving, and, you know, growing; and I think like AI and especially ChatGPT will help a lot in that industry as well, to make much more efficiency.
Mike Vizard: All right, folks, you heard it here. Not only are we gonna get where we’re going faster, we might enjoy the ride more thanks to AI. Anar, thanks for being on the show.
Anar Mammadov: Thank you, Mike, for having me here.
Mike Vizard: And thank you all for watching the latest episode of Techstrong AI. We invite you to check out this episode and others on the techstrong.ai website. Once again, thanks for watching and we’ll see you all next time.