Danny He: Solving Supply Chain Blind Spots with AI-Powered Context

Kyle James (00:01.1) Hey, welcome to the AI Chronicles podcast. I'm your host, Kyle James. Today we'll be talking about how a supply chain SaaS company called Soapbox is using AI inside of their own business. And we'll share the exact steps that you can take in order to implement AI for yourself. Now, before I talk about that, listen closely. Are you looking to implement AI inside of your own company or maybe just struggling to get your AI to stop hallucinating? Speak to GPT Trainer. GPT Trainer. literally builds out and manages your AI agent for you, eliminating hallucinations for good. Go to gpt-trainer.com. I promise you, it'll be the biggest time saving decision that you've made all year. Trying to set up AI on your own is like trying to build a house from scratch. Sure, you could do it, but the time of frustration is going to take you to get it finished. It may not be worth it. It's a thousand times faster and safer to hire professionals. Schedule a consultation today. Once again, that's gpt-trainer.com. Today I have with me Danny He, who is the founder and CEO of Sobox, a leading edge supply chain and logistics technology platform out of California. Hey, Danny, welcome to show. How are doing, my friend? Danny He (01:08.973) Thanks for having me, Kyle. Really excited for this podcast. Kyle James (01:12.898) Yeah, for sure. So tell us a little bit about Soapbox. Like how did it come to be? What's the what's the background, the origin story with it? Obviously the founder of the company. Love to hear your take on the company. Danny He (01:22.491) Sure, yeah. So we started this company six years ago and I was born on the premise that within certain industries, in this case supply chain, the data within this space is more or less distillable to a normalized format, meaning that... it's roughly the same set of data all the way through, right? You're either moving something or you're counting something and that's basically it. That's like all of supply chain data in a nutshell. So if anybody's been in this space or know anything about this space, every single piece of supply chain has its own set of software and therefore their own set of data, micro data, micro sets of data. And that's just really highly inefficient. You can't really do anything right with inaccurate and incomplete data. So we set out to create a platform that really operates as an operating system within supply chain that isn't like an ERP. We didn't want to have an ERP where everything was happening, but we wanted to kind of connect with third party systems within our systems to actually generate the best data in supply chain. Kyle James (02:36.76) Hmm. Yeah, absolutely. So talking, you mentioned the data part of it, like what, cause I know, I know there's a lot of companies out there who like to have some of their data in hand. It's kind of may not be structured or they don't put a priority into it. Like how big of a priority would you say is the data side of things with, within like some of the within soapbox. Danny He (02:55.416) Yeah, that's great question. I think everybody understands the importance of data, but I don't know that too many companies have explored the value of the data that they have. And that's one of the things that, especially again, in this space in supply chain where you're working with a lot of disparate systems, you're working with third party vendors and providers and it's an ecosystem, right? Kyle James (03:09.121) Hmm Danny He (03:25.016) Within any data set, if you're missing a specific piece of that data, you're going to have incomplete sets of contexts. And in supply chain, everything is about context, right? This is a thing that needs to move from point A to point B at this time. That is all contextual. those are things that, as an ERP, doesn't capture. It captures the raw resource of it. Kyle James (03:34.19) Hmm. Danny He (03:52.943) That's the part I think from an operations perspective a lot of companies miss out on and they don't miss out on it not because they don't want it, it's because they don't even know how to consolidate that. They don't know what that looks like and that's where we come in to be able to give them that single pane view of their supply chain. Kyle James (04:12.076) Yeah. So you mentioned like the context, like taking it from data to context, like what, what are some maybe gaps a lot of companies like have within their perspective? Like they have the data, maybe it's structured, but then they're like, miss, they're missing that context. Like what is that actually, what harm or maybe like loss in revenue or opportunities? What is that causing when they don't have that, that context visual that you're talking about? Danny He (04:35.331) Yeah, so it's really interesting. lot of, well, virtually all business intelligence tools will pull in ERP data to be able to generate forecasting and demand planning and a lot of the levers that they use to be able to manage their supply side operations. what's really always interesting for me is that ERP data is inherently stale. everything has already happened and then it goes into the ERP as a data point. when you're always behind the eight balls, so to speak, when you're looking at data that's already done. an example, a perfect example would be if you have a highly perishable good and you've got, let's say 10 warehouses that is has your products and distributing your products, but then you don't have the connectivity to that data, which is your data, right? If it's your products in that warehouse, that's your data. What they're counting in the warehouse is your information, right? So you should have access to that, should have connectivity to that. But most companies don't. So what do they do? They get a weekly download, a spreadsheet of the inventory and the movements. What's gone in, what's gone out. That spreadsheet then gets uploaded into an ERP and then there's analysis done after that. So in the time that that has already happened, that might be 10, 12, 14 days. How are you supposed to make decisions and gain insight in real time? How are you able to be agile, respond to market dynamics when you don't even have a clear view of what's in front of you? Kyle James (06:13.996) Hmm. Kyle James (06:25.934) Yeah, absolutely. Almost like I would imagine even to having like the that 14 day as a delay like that that could that could hurt different almost not necessarily metrics, but opportunities or like when you have to act fast like the timing issue of it can probably play a huge role. I would imagine even so. So what? Walk me through a little bit here. So boxes you're using AI and love what your team is doing like why did you decide to start? using AI in the first place and specifically what types of challenges were you trying to solve with it, Danny? Danny He (07:00.665) Yeah, so when we first looked at AI, it was really a function of, well, if we're building this platform that can connect to all of these data sources within supply chain, what are you going to do with that data? And it was actually a question that a lot of our customers started asking us is, well, that's great. We have all this. We don't even know what to ask it. Like, what questions should we derive from these things? Or what have other companies looked at this? And we were like, well, you know what? Kyle James (07:22.862) Mm. Danny He (07:30.265) That's a really good question. And why don't we let your data answer that for you? And so that's where having an AI engine to be able to empower business analytics, that's really what we provide within our AI deployment is the ability to be able to ask our AI a question. And it'll be able to respond, whether through text or data, to be able to generate dashboarding. alerts, workflows based off of real-time information that's coming in and out of your system, right, whether it's supply or demand. And when you have something like that, it's almost like you don't know what you have until you've seen it. And then once you've seen it, you'll just never be able to go back to it. And that's what's been exciting about what we've been able to prove out within this ecosystem is uncovering the... the value, the true value of the data that people already have. Kyle James (08:30.604) Yeah, so absolutely. And so when the data comes in, obviously you guys, you're able to get a clear view for a lot of your clients, customers that you're working with. From there though, like what specifically is the AI doing maybe kind of step by step? Like what does that process look like? Once the data is coming in, what is the AI doing? Like you mentioned the questions as an example, what, you know, can, ask the, ask your data questions. Like, is that being done through AI? And like what types of questions and like, I guess like opening, not openings, but. Maybe that's why phrasing like kind of revelation that people are getting from from seeing the data or asking the data through these today. Anything worth mentioning? Danny He (09:09.019) Sure, yeah. So I'll answer the first part, which is how does the AI work? So we partnered with a with graph flow. It's a AI technology that's based off of the graph database. And effectively what's what we really liked about this that fit in perfectly with what we were doing within our ecosystem is for us, we connect into all these data sources within supply chain, right? Whether it's warehousing or inventory, freight, transportation, whatever it is, anything that can generate information and data within supply chain, we are connecting to it, we're pulling into our system and standardizing it, normalizing it and standardizing it into a view that's digestible and makes sense across the board. You might have 10 different vendors that have 10 different views of it, but you just want to see comparatively what they are, right, normalized. So that's what our platform does. Well, if our database is set as a standardized and normalized format, then it makes AI deployment extremely easy using the graph database model and applying AI to it because how the graph works is that rather than gaining context on the questions or the queries that you're looking for directly within the data set, we actually map out the relation between all of the fields within our data. So an address, I'll give you an example, is an address is a location. A location is something that a distribution center or a waypoint can be. A distribution center can house inventory. Inventory can be placed into these things. It can also have an expiration date. And an inventory that's available can be used to be able to fulfill orders and so on, right? So what's interesting about supply chain is all of those data sets Danny He (11:06.551) are static, right? Like those relationships can only grow, they don't contract, but they're all very well mapped out and very well known. And so when you have an AI that then understands, here's your full data set, this is what you can do with it, you can ask it anything you want. there are a couple of holy grail questions that are extremely difficult to be able to identify. So one thing that people have always wanted to have, or at least just like the holy grail, is just-in-time replenishment. So you're selling something and before you run out, just before you run out, you get replenished so that you just don't run out. And that seems like a really kind of easy task. except you have to be able to take into account historic demand data, existing demand velocity, supply chain and replenishment capabilities and timing, any sort of externality like weather delays, traffic delays, et cetera. So there's a lot of things that happen. There's a lot of things that happen that will impact something like that. And so... Kyle James (11:58.499) Hmm. Danny He (12:19.075) For us, being able to then standardize the view of it, right? And it doesn't really matter what that input is on our end. If we're able to kind of have all of this information and be contextualized between each other, we can answer those questions dynamically. Kyle James (12:37.39) Hmm. Yeah, I love that. It's, it's really like putting a different or multiple angles on the data. Cause I think a lot of people like they get the structured data on just one face of it. But in this case, it's like having different angles where it's like, it's the weather it's, you know, the timing, maybe it's during the holiday season or for something like that. Like it, it has to have those multiple angles to make sure that it does work. Cause otherwise if it's just the data on the front, like just strictly data structured, that may not be enough to make those decisions, but, even even on the AI front too, like when the AI has all those handles of the data, then now it gives, guess, like, would you say maybe like a more accurate response or like more accurate suggestions on like, what is the next best step here? Danny He (13:18.447) Yeah, so you pose a very, I think a very big problem within supply chain, which is the ability to be able to just, I think, have ownership of what this data means. And so there's a lot of middleware in this space. They connect one system to another, and they make it very easy, right? Those connections are very simple. And yeah, you got A to B. That's great. But within the context of, okay, well, that's great. You have one connection, that one-to-one connection, but it's not like it's, that's a one-to-one connection, it's like a web, right? The way that we look at it is more like a hub and spoke model, where instead of connecting each data points to each other, they're connecting to our system and we are directing the traffic across the board to different systems. And so we act as a true middleware. Kyle James (14:04.334) Hmm. Danny He (14:09.913) but we act as a middleware as an operating system. So thinking about like the early operating systems, what do you do? You have a platform and then you can install different software on it that would enable you to do different types of tasks on it. So that's what Soapbox. inherently is, is it's an operating system where you can have all these different data sources. We integrate into it. We pull it into our system. now whatever it is that you want to be able to answer from an AI perspective, as long as we have a clear line of sight on the data, we know we can standardize and normalize it, which is the hardest part, right? That's the hardest part of data hygiene within AI is how do you know that this is good data? But we know this is good data because if it is a good data, we wouldn't be able to normalize it. If we're missing any parts of it, then we would know, here's the gaps, because we know all of the points that are required here, not just some of the points and not just the points that you offer. We know all of the points. And if we're able to say, in this case, this is the gap we're missing, that's something where we can identify that, quite frankly, other middleware and other connectivity tools in the system can't do. Kyle James (14:55.726) you Kyle James (15:25.676) Yeah, yeah, absolutely. from you, I feel like you've got like a like a zoom in view and a bird's eye view of data. And that's just like, I think it's so valuable within the marketplace. And for some of your clients have been working with, like what types of results would you say, maybe something that's worth mentioning that you have seen so far either for your clients, something that's worth highlighting or even internally that maybe might be worth mentioning that your team has been internally doing on the results side. Danny He (15:55.235) Yeah, so I think one of the things that has been really illuminating for us is that regardless of how large a company is, and it's actually, this is more pervasive within larger companies is because there is more data sets, there's more connectivity within their supply chain, they just have less visibility in the space. And to the extent that a lot of companies here, they'll use like a brokerage service or like another third party and that third party actually uses a wholesaler and the wholesaler uses a distributor, right? A lot of companies in this space don't get down to the lowest common denominator, which is where the action happens that generates the data, right? So that's the movement and counting of stuff. When we get down to that level, that's where we... dig in where other companies don't and where other people kind don't look into that space. They don't really like to get into the dirt, right? Like who wants to be able to understand how a 25 year old legacy software keeps count of your inventory, right? That's a really difficult task, but that's exactly the task that we thrive on because that probably has so much historical data that you wouldn't even know. And it teaches you. how you should run your business rather than how you think you should run your business. And I think that those types of insights is kind of unique in this space and unique to having the full view, the full visibility of it. I can give you an example. So aging, product aging. So aging is a really big issue. If you have things that are perishable, it... ranges between five and 25 % of things that just, it just ages out, meaning that it expires, it's out of useful life, whichever. For certain things like bread, as an example, it's up to like 30, 40 % in aging. So it goes back real quickly. That's just the nature of the product. So how do you identify aging dynamically when the processing facilities are not technically connected? Danny He (18:13.583) to the distribution centers that are not technically connected to the retailers, how do you know when how much you need to replenish? So when we connect that supply chain thread all the way through, we don't need somebody to be able to do analysis on it. Those are data points and insights that we can provide immediately through AI. AI knows, right, the technology knows that based on these sales and based on the fact that it takes you, let's say two days to be able to, from the time that a purchase order goes in to when you can actually sell this product to fulfill it, that hey, this is how much inventory we need to be able to replenish it. So that's something, something as simple as that will help reduce shrinkage and just... items that age out by 50 % of where they're at. Kyle James (19:08.014) Yeah. Yeah. Absolutely. was like, when I was like the magnifying glass, like I love how like Sobox is taking it deeper, like almost like it's, it's, it's not just surface level and grass is earthworm in the dirt getting through the fine and the details. Cause that's where like, that's where the change and I guess the most, not most valuable, like even just like the hidden gems that you can find deep within the data. And that can be game-changing for a lot of these companies who, when they're trying to transition back and forth with different A lot of their clients and customers are shipping materials to, like you need to know that fine tune data to make those, especially when you're making those decisions long-term, like having those decisions made long-term, like you have to know the details to it to make sure you're making the right decision. And talk to me a little bit about, know, Sobox's maybe upcoming AI initiatives. Like there's a lot of change happening within the market. Where do you see it maybe kind of visioning of AI playing some of the biggest role in your operations next at Sobox? Danny He (20:06.405) Yeah, yeah. I'm a firm believer that AI won't replace people. People who know how to use AI will replace people, right? And that's really what it comes down to is within the supply chain in this industry, quite frankly, in many industries, a lot of work behind the scenes is manual, it's analog. And you don't really think of it as such, but it happens. So as an example, in freight, there's a huge problem with paper trail. and you have somebody that wants a load to go from point A to point B, they'll reach out to their broker, the broker then send it out to their carrier, their carrier then send it out to the driver, the driver then picks it up from the warehouse, they have to swap physical documents at that point, that physical document that gets then a picture of it gets taken, and then an email gets sent over from the driver to the carrier. then that document gets manually uploaded, typed up by the carrier and then forwarded as a document to the broker that then has to type that up again and pull all that stuff through. That's manual stuff. So those are the problems that we're looking to solve and we're in a really unique position because we have the capability to be able to have a really light touch implementation and connection into all of these technical points so that we can kind of simplify this process as streamlined. Kyle James (21:41.24) Yeah, for sure. Yeah, I definitely see AI playing a big role. I mean, I'm sure there are some things that like still in manual need to be taken care of, like, you maybe taking a photo of a, you know, receipt or, but for the most part, there might be some things. So as we start wrapping this, Danny, I appreciate you, your time today on the podcast and where can people learn a little bit more about you, maybe a little bit more about Soapbox that you'd recommend them checking out? Danny He (21:51.579) Right. Danny He (22:07.065) Yeah, thanks. Thanks Kyle for that. You can check us out at our website at soap-bx.com. We're on LinkedIn as well. It's Soapbox. If you linked in myself and Danny He, Danny He Soapbox, you'll be able to find us. We do release some content from time to time, little gems about supply chain data. We're focusing a lot more of our content around data and AI and analytics. So if you're interested in that, happy to be able to share more of that content. But it's been great. Thank you for having me, Kyle. Kyle James (22:44.174) Yeah, absolutely. It's been a great time. Definitely looking forward to it. And as we wrap today's show up, remember, if those who are looking at implementing AI into your business today, please don't try and do it yourself. The time and stress that the AI could cause, it may not be worth it. Schedule a call with GPT trainer and let them build out and manage your AI for you. Once again, that's gpt-trainer.com. Signing off for now. Have a great rest of your day. Thanks again, Danny. It's been a pleasure, And looking forward to seeing everyone on the next episode of AI Chronicles.