Ask An Expert with Sr. Data Scientist Juichia Che
In this episode of DataXstream’s Ask An Expert Series, Kathleen Taggart, Jayelyn Ramey, and Juichia Che from DataXstream’s Intelligent Automation team discuss the Fourth Industrial Revolution, the tech trends, and the business outcomes involving automation and machine learning. Reviewing both the traditional difficulties of managing enterprise data and the modern value of Intelligent Automation, the three present an analysis of rising industry trends and how they play a small part to provide the solutions for a strategic approach to riding the revolution.
Hi. Thank you for joining us today on DataXstream’s Ask an Expert series. I’m Kathleen Taggart and I’m here with Jayelyn Ramey. And we’re here to discuss the 4th Industrial Revolution with senior Data Scientist Juichia Che. And Juichia, I know you are very modest and polite, so I’m going to quickly introduce you to our audience before you stop me. So, Juichia has three plus years of experience with DataXstream and originally a Carnegie Mellon Computer Science and Statistics graduate. She entered the ASAP Solutions and consulting world in 2005, when she first consulted for DataXstream, she later cofounded, an SAP Consulting firm and is currently working toward her master’s degree in Analytics from Georgia Institute of Technology. She is currently our senior machine learning engineer and has been leading DataXstream’s, AI and machine learning and data science team while she lives in Georgia with her four-year-old daughter Sophie. So Juichia, yeah. Is there’s anything I missed? Just jump in and feel free to improvise.
UM, instead of leading the whole AI team, it’s more like stumbling and trying to figure our way out in the AI world, right?
Yes, definitely. But you do good job of stumbling, always catching ourselves. We’re pretty good at that.
Uhm, and while we’re kind of discussing your career and how this project came together and how we got started, do you want to briefly tell us why you pursued data science after starting in the SAP space?
Other than having a toddler to take care of and not having as much time for consulting?
Other and doing the career transition to be to contribute more to my work life balance. I’ve always had a big interest in and a knack for analytics and even in SAP I’ve worked a lot with data. In terms of working with master data at different industries and using their databases and mapping out data fields in interfaces as well as doing data migration in upgrades and implementations so data has been a very big theme in my career. And I thought because now data science is such a demanding and hot field and AI is such an exciting new technology that it would be a great field to go into.
I love hearing about somebody’s personal interest in their industry! So, there’s been a lot of discussion recently about the technologies that have been developed to accommodate arising industry challenges. What are a few of those challenges, specifically looking at the wholesale and distribution industry?
Well, I think in the wholesale and distribution industry specifically, they deal with a lot of different types of customers. So, there is a big lack of standardization in their business processes from end to end. So, in terms of how their customers request for information.
And how their customers want to receive information, right. So, all these sorts of lack of standardization have caused a lot of challenges for this industry and this also means that the human resources in these companies in wholesale and distribution, they are often weighted down by menial tasks that take a lot of time, for example data entry for product information and to procreate purchase orders and quotes for their customers.
And these sorts of things. There is also a lot of, I guess, across any industry, right that the data quality and the way data is maintained has been a big challenge in terms of moving from, um, information technology platforms to a more AI driven, and data driven platform in their business.
Yeah. So, as we we’re discussing some of the, like you just said these recent technologies that are coming out to address these industry challenges, and the amount of data that is now available and then also the new technologies that are, and that have been developed that are now driving these solutions, we’ve seen huge advancements in like things like robotic process automation, the Internet of Things has been a very recent development of the last, maybe 20 years or so. We’ve got blockchain, edge computing, robots and cobots, genetic programming, quantum computing, like the list goes on and on and on. And it’s all been this, I guess, like the fruits of the mass amounts of data that are now available, and then also the fact that our data analytics and our things in machine learning are getting much more powerful now.
But let’s kind of go into automation and robotic process automation and how machine learning and AI are powering the automation in today’s economy, but also what does that now mean now that we have all this automation? What is that really going to do to the industry?
Yeah. First, I’ll be honest, I didn’t know half of the things that you mentioned like quantum computing.
It’s really cool!
But I will try my best to talk about. I guess the automation that we have been working on and the needs that we’ve seen in our customers. Right. Number one, there’s. Uhm, human resources at these companies need to be, I guess, allocated to more meaningful tasks. And in order to do that, automation needs to be implemented in business processes where human resources used to be required and dependent upon, especially for their business knowledge. For example, sales representatives who had a deep knowledge of business systems and how data was stored in there, and they spend a lot of time mapping customer requirements to what they know are stored in these business systems. And these sort of parts of the business process can now be automated by machine learning applications that automate the mapping between the input to the business and the output from the business back to the customer.
And then in sales? There is a big cry for automating the order processing kind of business process where businesses receive requests for quotes as well as purchase orders from customers and often times as we talked about in these wholesale and distribution industries, there’s a lack of standardization in the way they received the quotes and the purchase orders, right. So, there is a way to both standardize how these are received through what channels. So for example, maybe they used receive them in emails as well as maybe text messages to sales representatives tips as well as maybe just handing over a piece of paper to the sales representative. Right? So, these sort of different channels can now be more standardized through different technologies. For example, scanning in paper documents and automated extracting of information from emails and text messages by detecting what sort of topic is included in that message, right? And then there’s also another part that’s more downstream after we standardize the way information is extracted from the customers, then we can automate the way this information is interpreted. It used to be that we depended on the sales representatives to interpret and to understand by reading the documents what the customers wanted. But nowadays with something like optical character recognition and also statistical methods analysis like regression and classification. We can now map exactly what’s read from the document using OCR to what is expected in the target business system and something like SAP where we know that in order requires a customer name, we know that it wants a material number and material description, quantity, and price and so on. So nowadays we can use machine learning models that classify the pieces of text from these sales orders and quotes to know like oh, this piece of text is a customer name and this piece of text is the delivery address that they want their products to go to.
Other than that, there is also automating inventory restocking, integrated material datasets and also creating user accounts. For example, for our customer management and mapping data from external systems to internal business systems. Another big thing that we’ve seen from customers is product recommendations. So, once we can automate the whole order processing and quote processing system we can say, OK, we know that this we can collect data about what customers have ordered before and what they’ve asked for information on. And then based on that be able to recommend other products that may be similar or that may be complementary to the ones that they requested before using machine learning models.
And if I can throw in, also to just, to jump off of that point, it’s very much not about I think I think automation is a scary word because people think when you bring in automation that’s replacing humans and in reality it’s very much and empowerment of the human so that they have the data that they need and they’re able to spend that time on the things that we do best. So, the creativity and the insights and the customer management and the customer interaction and that service because they’re not bogged down trying to manually, like you said, input those orders or convert data from one format to the next format. They’re able to look at the results and say great, I have this data now, what can I do with it to serve the customer.
Absolutely. I think that’s one big fear and also something that I think governments are trying to address is there is a fear of automation replacing people in jobs, right? But I think the solution to that is not too… not do automation because automation has a big benefit to businesses and businesses are critical to the way our economy and our society works. So, I think the best solution is to look at how can we allocate current human resources to higher level tasks that are more meaningful, right?
And if I can also throw in there. I saw a recent study conducted by McKinsey and Company that had determined about 50% of the order management process is highly automatable with what we already have today. So that’s not even what we’re going to see in five years or 10 years. That’s what we can do today. And I think it also shows how inefficient the typical order management system is if 15%, or pardon, if 50% is already something that we can automate. And instead, we have. Humans who are trying to keep up with that work that’s really not suited to them. It’s very much data entry and it’s very menial and the, you know, quality of job satisfaction goes down significantly. Then you got to think about, employee satisfaction and what you’re doing to support your HR team also.
And if we can talk really quickly about, we kind of discussed the need for automation and how the solution isn’t to not automate. And what’s the return for companies that are early adopters and if automation is so great? Why has it been such a struggle for some industries to incorporate and do the digital transformation toward automation? What do they get stuck on? And why do people get stuck in the planning and the pilot phase when they go to bring in one of those new technologies?
Yeah, just like any other sort of technological implementation, there is going to be some challenges in terms of bringing in change and people, I guess people tend toward being comfortable and not wanting change. So, when there is potential for change, even if the benefits are pretty strong and everybody might be considering it. Within an organization there are there are still some difficulties in terms of how to drive that change. One is, preparing, I guess, talking about the risks and the pros and cons of automation and using AI in organizations, right. So, while I think a lot of business leaders and companies, they read a lot about AI and they might go to conferences and read papers about automation and AI and maybe from Gartner they can understand like how these technologies can drive return on investment over time. And but I think there is a gap between what they can read and the reality of implementation, because right now there is a… when we want to implement AI and machine learning, there are costs associated with such technology, even if these applications nowadays are all run in the cloud.
I think it actually becomes more mysterious nowadays to business leaders now that everything is in the cloud. The infrastructure, the data storage, the applications, everything that an organization runs on is now less tangible I guess, because you now don’t have like IT administrators sitting in the office maintaining mainframe systems you have somebody else like Amazon or Google with a centralized locations for servers where you depend on them to maintain these things and you depend on them to be honest and to be competitive in their offerings in terms of how much, these infrastructures and these cloud platforms cost and how dependable they are. So that’s one big area where I think business leaders have a big concern about, right?
And just in terms of the complexity of calculating that cost and understanding how it’s going to impact the return on investment over time. The other big gap in knowledge is how machine learning solutions and automation change overtime, because with something like implementing SAP, you’ve got a plan that you go in with and you implement these business systems to requirements and once they’re in, your users can start using these systems.
And it will map to the exact requirements that you had. So, you could say like this system is working exactly as we planned. But with something like implementing automation and AI, it’s not as clear cut. Especially if the AI or machine learning provider oversells their capabilities and it’s not as clear about how machine learning changes as a solution over time? Like I say, SAP doesn’t really change overtime unless it’s through upgrades and the software changes. But with machine learning it’s changing as your data changes and the solution for automation needs to take into account how the company’s data changes and how that can be leveraged to provide feedback to the machine learning pipeline. So that the machine can learn and become more accurate over time. So, some business leaders might hold off on machine learning as a solution for automating business processes because at the beginning it might seem like, oh, it’s not that accurate in terms of predicting, for example, the text on this document. But what needs to be explained clearly and may be broken down into more of a simpler explanation is that, while it might not be 100% accurate, ever. It can learn from the mistakes that it’s producing by gathering the feedback from the humans that are using these applications. So, I think that’s one another big area where change becomes difficult in organizations, so I think it requires business leaders who are open minded, who are driven about improving efficiency in their organizations, but it also requires these leaders, to seek out trusted advisors in the field of AI as well as machine learning who can show them exactly how these solutions are built and how they work.
Work in in ways that are understandable and not full of jargon, and also how, I guess, somebody who can provide the pros and cons of machine learning versus a traditional software application solution.
Yeah, that’s a very good point that it’s very useful to have a trusted advisor who, you know, it’s not the sales team for their solution obviously. And it’s not someone on their team because sometimes you wonder, you know, obviously they understand their own solution, but someone who understands your industry maybe even has experienced at your company or has experienced with companies very similar to that and has this background with data science and machine learning who can kind of bridge the gap between the technology of the solution and then also the industry that you’re currently operating in?
Yeah. The other challenge is sometimes very difficult to pin down. The exact return on investment for new solutions because. Uh. Even businesses themselves don’t really have metrics for measuring. Manual business processes and how long they’re taking their human, to do a certain kind of work, and what sort of inefficiencies exist in those processes? So sometimes it’s very hard to measure precisely what’s the value of a solution.
Yeah, that’s another good point. Getting and collecting the metrics is another huge challenge and as it is with most transformations and most goals, you need the metrics to prove that you’re working toward the goal and with machine learning, sometimes it is hard to label it out. I saw this really interesting analysis once that described DevOps as you know where the finish line is but with MLOPS you are running, you know into the trail. You know that you’re in the woods but you don’t know where the finish line is because. Machine learning is an evolving solution as you said. So, for there to be a finite completion or finish line is kind of counterintuitive to what machine learning actually is as product. So that’s something that we need and we’re working on getting comfortable with. Both the technology side and the industry side.
Yeah. Oh, that’s what makes it a really challenging field. And a very interesting topic is because you’re not just, like developing some application and deploy and forgetting about it, you’re really creating a whole ecosystem that takes into account how the business is evolving, how its data is changing as you go along and adapting to that.
Right, Yeah. Thank you so much for these excellent insights and your advice on the industry challenges. And then also on the solutions and how companies can prepare for those. Thank you for your time and sharing your wealth of knowledge with us. Thank you also to our audience. We are honored and thrilled that you spent your valuable time with us today. We hope that you learn something interesting and insightful, and we look forward to hearing your feedback on our Ask an Expert forums. They will be located at the bottom of the web page right below the recording. So, if you have any feedback, please let us know. Please also visit us at dataxstream.com to learn more about Juichia Che and hard work and also the machine learning team and our solutions. We hope that you will contact us on the forums. We look forward to hearing back from you, and we hope that you’ll join us again for our next ask the expert interview. So, until then thank you so much and bye for now.