Leadership Insights: Natural Language Processing

Transcript

Amanda Razani: Hello, I’m Amanda Razani with Digital CxO. I’m excited to be here today with Kaycee Lai. He is the founder and CEO of Prometheum. How are you doing today?

Kaycee Lai: I’m doing great, Amanda. Thank you for having me today.

Amanda Razani: Glad to have you on our show. Can you share a little bit about Prometheum and the services you provide?

Kaycee Lai: Absolutely. Prometheum is a Journey AI company that delivers a data fabric so that way customers can actually much faster and much more easily connect to all their data sources to be able to make fast decisions for analytics. So we pride ourselves in making very complex tasks very, very easy, such as even being able to use natural language to do a lot of those complex tasks such as ETL and writing complex queries. They used to take days, weeks, and sometimes months. To be able to deliver that in near real time for customers so that way they can address their most pressing issues immediately and not have to wait for some drastic change in the market to actually happen because a lot can happen in three or four months. And so we really want to democratize the complexity of using data so that way any employee at any company can actually make data-driven decisions.

Amanda Razani: And harnessing that data more efficiently is certainly a big priority of all companies. So with that being said, you were talking about natural language processing. How does natural language processing technology enable retailers to analyze consumer behavior more effectively?

Kaycee Lai: Yeah, it enables retailers to do that because natural language is, well, the most natural way that human beings think about when we’re trying to figure something out. If you think about it, we use it all the time in our personal lives from search engines to say Google Maps. It’s really always around answering a question. And analytics at the end of the day is really around answering a question. You want to know what was sold, how much of it was sold, why it didn’t sell, where it was sold, et cetera. And so a lot of retail is around answering these things, things around, well, how do we actually offer the best promotion to really move certain products that we want? Why is a certain store not performing as well? Could it be that they’re not getting the products they need in time? Does that have something to do with our distributor? Does that have something to do with some of our suppliers?

And all these questions are racing through people’s minds, people in the finance team, the market team or store managers, but unfortunately not everyone is a seasoned data engineer to be able to go and extract and pull that and assemble that together. And so where natural language comes in is you don’t have to do that. You can simply ask the questions in a way that makes sense to you, in a way that makes business sense, and allowing that technology to go and do the translation for you, all based on what you want to know from a business sense, Amanda, I’m going to figure out for you what data you need and how to put it together and deliver you the result that you need so you can make that decision right there and then.

Amanda Razani: Got it. And so that certainly opens the doors to a lot more employees that maybe don’t have that high-end skillset, but with the natural language processing, are able to do so.

Kaycee Lai: Absolutely.

Amanda Razani: Yeah. So the next question is, how does this help retailers improve the customer experience and enhance their product offerings after they get this information?

Kaycee Lai: Yeah, it does so in two ways. One is we often only think about it from the retailers, the corporate, the company’s perspective. But what about the buyer? What about the customer’s perspective? For example, just going into a store sometimes, if you go into a large store, sometimes you get locked, you don’t even know where to go find that. And if you’re like me, you’re running around and trying to figure out how they label each aisle and wouldn’t it be really easy if there was an app where you can just say, “I’m trying to find where they put the box of Cheerios,” and I can go exactly to know which … or wouldn’t it be great if I know before I start my trek that this store doesn’t have my favorite brand of Cheerios and I should go to another store. So using this type of technology can really also improve the customer experience, but as well as bring together on the other side, what if I know what my customers want by all the questions they’re asking before they get to the store, when they get to the store?

Then that’s probably going to help me adjust my supply chain. It’s going to adjust what I decide to buy, what I decide to stock. It’s going to change how I would price certain things out based upon the demand. And it also helps from an IP perspective, when I start knowing, these seem to be the type of questions that my business folks want to know, what can I do to make sure they get the best performing infrastructure to deliver that? What can I do to make sure that data is always protected? So in a simple term, that ability to use natural language to figure things out really gives you clarity into how you should prioritize based upon the demand of what you’re trying to figure out.

Amanda Razani: And clarity in real time as well, which is helpful.

Kaycee Lai: Very important, yes.

Amanda Razani: So what specific benefits does Promethium’s natural language processing platform offer to retail companies in terms of this real-time data retrieval and analysis?

Kaycee Lai: Well, one of the things that we’ve kind of evolved over the years for retail is that data is never in one place. You cannot expect all your data from suppliers, from logistics company, for manufacturers, from customers, et cetera, to be in one place. And unfortunately, the way that most analytics works is you have to get everything into one place first, into one table or one file before you can do any analysis. And so this involves a lot of time figuring out where things are, moving things. And unfortunately because the data doesn’t reside in one place, one system, they’re in different formats. So someone has to go and wrangle data and to put them into different format before you can do this.

And so what we offer at Promethium is – what if we automated all that for you super quickly? And so the minute you ask, we immediately know what the pick word is, and then we construct that output for you so that your BI tool can instantly take that and give you the visualization, the report to tell you this is what it’s going to look like. And that’s the value we bring is that manual experience would’ve kept someone waiting weeks, which is not a good idea when you have perishable goods if you’re a retailer. So this is where we actually, with our unique technology, really allow you to automate that discovery and assembly of the data in real time.

Amanda Razani: So for companies looking to digitally transform, what are some of the top concerns and challenges that retailers are facing when implementing this NLP technology into their operations?

Kaycee Lai: Good question. I think it’s related to the first one. I think a lot of retailers still sort of take the traditional approach, which is, well, I’m going to just move everything into one place. And the challenge with that is, well, by the time you’re done moving, you’ve generated data, new data from the sources, so you’re never fully caught up. You can never fully move. And unfortunately, these data migration projects are often very expensive and take a lot of time. And so I see a lot of folks failing at that, and they don’t end up being able to capture the benefit of being able to consume that data. That’s one. The second piece is the data quality is actually super important. If you don’t know what things are called, you can’t find it. If the things that you’re looking for or retrieve are not in a good state, then you’re going to get bad answers. So I think that’s really important on the data quality and the data discovery.
And the third is, I think you guys, everyone, unless you’ve been living under a rock, there’s this thing called Gen AI and large language models.

I think Gen AI and large language models have really, really accelerated. I think the interest for NLP, especially in the enterprise, I think now there’s a legitimate shot of actually doing this and doing it right and doing it quickly, but we’re still in the early innings of adopting large language models in NLP. And one of the things is how do you figure out, number one, where to implement your LLM? Do you implement it locally? Do you use a public one? How do you make sure that LLM is relevant to your domain? Because the challenge is if you use an LLM that’s geared for insurance, you may not like the output you get when you’re a retailer. So that’s complex.
And then third is also if you’re trying to use a staff based LLM like ChatGPT, well, security is kind of important, especially retailers with PII data. How do you make sure you can get the benefits of an LLM without compromising on the importance of security data access? So these are some of the things I think retailers need to really think about before they jump in. It’s not simply, let’s grab a chatbot and here we go. All those problems are still going to persist. And so that’s where I really encourage folks to really take the time and think about implementing the technologies in all these different components and all these different areas.

Amanda Razani: Okay. Can you provide some examples of successful retail companies that have leveraged this technology to transform their business and achieve growth?

Kaycee Lai: Yeah, absolutely. So we have a customer who’s in the consumer packaged food industry. And just for them, they have to deal with all different types of suppliers, all different types of distributors, and they have different bakeries that actually go and take the raw materials and actually create them. Well, it turns out that not every bakery can produce the same amount of goods with the same quality at the same time. So that’s really interesting to kind of figure out what’s the optimal materials, raw goods, raw ingredients I give to a bakery. Some are much better at baking a certain type of pastry, some are much better at baking at a certain time. And so all this kind of factors into not only the ingredients, the type of ingredients, but also the logistics of how do you get that there. Some of them you may struggle in letting them bake a certain type of good because the closest distributor for that particular raw ingredient might be too far away.

And then you’re also dealing with inclement weather conditions sometimes. So you might be on the way to deliver a whole batch to Walmart and all of a sudden a snow storm hits. And at that point you have to now figure out what do you do? Do you tell Walmart to wait? Do you figure out what’s the closest bakery that you can now reroute those shipments to? Do you do something with the pricing? So all these things need to happen quickly and in real time. And with Promethium, we allow users to be able to simply ask these questions and being able to see where are my phase of bread? Where are my products? How long ago did they leave the truck? Did all the pallets make it onto the truck? And then being able to then understand the profit margin. Now I can understand what’s the optimal yield I should be giving to each factory, each bakery, to each distributor.

Do I have a discrepancy between what my distributors are telling me that they sent to the bakeries and what the bakeries are actually reporting? And is there discrepancy in time? Because if it takes too long to get there, this is perishable goods, it can really impact the output. So all these are questions that you can now simply ask the system in Promethium to be able to provide this level of insight and this level of granularity so you can make these decisions that significantly increase the top line, as well as improve the bottom line and avoid wastage. Nobody wants to have five cartons worth of trucks shipping to a place when they know it’s going to be five days late, and it’s going to impact the shelf life of those products. So these are things that we really help these customers avoid.

Amanda Razani: So – very helpful for the business, and also when we talk about sustainability and being earth friendly, less waste as well.

Kaycee Lai: Absolutely. And that’s something that we try [inaudible] because I think there’s already too much of it, and unfortunately, a lot of it, it’s not on purpose; but it’s just if you don’t have visibility into what’s going on, the waste does happen. And so we can actually really help them being able to give them as fast a timely insight. Because let’s face it, a lot of people can understand a natural language question and natural language answer, and they can do something with that. Not everyone is skilled at analyzing complex charts and dashboards and SQL queries to figure out what they should do. And so we feel like this is really game changing and empowering in terms of just now liberating a lot of folks who have the desire and the intent to do something good, but were always held back at a technical skillset.

Amanda Razani: Most definitely. Well, I want to thank you so much, Kaycee, for coming online today and sharing your insights with us, and about Promethium and the services that you provide, and I look forward to speaking with you in the future.

Kaycee Lai: Thank you so much for having me. I had a great time and I really appreciate it.

Amanda Razani: Thank you.