Transforming Business Processes with AI, Predictive Analytics, and AI Interface Designs

Dayle Hall:  

Hi. You’re listening to our podcast, Automating the Enterprise. I’m your host Dayle Hall, the CMO of SnapLogic. This podcast is designed to give every organization out there some insights and best practices on how to integrate, automate and hopefully transform their enterprise. 

Raghu Banda is a senior director of AI product management at SAP Labs. He’s joining us today on the show. He’s responsible for driving the adoption of AI-infused SAP S/4HANA business processes for both cloud and on-prem customers and partners. He focuses on innovating with SAP 4HANA business processes with additional AI technologies. And that’s the key for today, the AI technologies space, to improve the customer experience for all their customers. He’ll be sharing his insights and experience on the podcast today. We’re very happy. We’re delighted to have Raghu join us today. Raghu, welcome to the show.

Raghu Banda:  

Thank you, Dayle. It’s a pleasure to be on your podcast.

Dayle Hall: 

Yeah. And I know you have your own podcast, so I’m going to try and maintain your level of professionalism. I’m not sure I’ll do it because I think you’re more experienced, but we appreciate you joining the show. 

Raghu Banda:

Thank you. 

Dayle Hall:

Okay, so we’re going to kick off. I like to do a little bit of background so people get some sense of who you are, where you’ve worked, how you progressed in your career. Give me a little bit of background on your career, and then I have something specifically to ask you around how you got more involved in AI, but give me the cliff notes of your career.

Raghu Banda:

Maybe I’ll start my career even before SAP days. Back in India, when I graduated from my computer science and engineering, those were the days, mid-’90s. I think those were the days, all this buzz about information technology and everything was happening. And of course, there were traces of AI and robotics going on during those days. Even in my final year, I wanted to do a project on AI and robotics. But my dean, I think he said, no, you can’t do it right now, we don’t have all the resources. So we ended up just writing a white paper. 

But coming back to the career that back in those days in the Silicon Valley of India and Bangalore started in my working for one of these companies, which was a start-up formed by a group of Indian Institute of Science professors in Bangalore. So they used to do some projects, finite element modeling, and all these core software that was used in computer-aided design and all. So this software was used in NASA and some of the other big firms. 

So that’s how I got into my career. Later on, I progressed into working on these latest cutting-edge technologies and got into Wipro, which was one of the big PHI technology consulting firms back in India. And that is when I started learning about all these new technologies like the Internet of late ‘90s, Internet technologies, and how do we build these different software programs. And then I heard a lot about enterprise businesses, enterprise business software. And that is when I started looking into and trying to find what would help me propel my career. And that is when I got an opportunity to work with SAP. SAP started with a start-up firm called SAP markets during those days. It was a joint venture with Commerce One, good old days in early millennial days. And that was the time we were building some procurement-related application, later turned out to be supplier relationship management in the procurement space. 

So that’s how I got into the SAP world. And getting into SAP, it was a huge achievement at that time and there is a lot to learn and a lot to unpack there. So that’s how my journey started, yeah.

Dayle Hall: 

You’re in the customer success organization yourself, so you understand how technology impacts customers. And I think that’s one of the areas where I really do believe if you’re in an organization today and you’re looking at how to get more involved, companies have data, how do I get more involved in some of these new technologies around AI? I think if they’re in the organization, if you think about what you’re trying to solve for customers and the data that you have, it could be in a sales ops, marketing ops, customer success ops, it could be in IT. But if you could think about what problems you’re trying to solve for customers, then start looking at the technology that’s available and what data you have. I think that, in essence, will get you more involved in AI without saying, okay, I want to be an AI engineer now. It feels like it’s more of a, things are coming together with the people and the skillsets, the technologies available. We have all this data. It’s just the right time to potentially move into this area and less so we just make a decision, we’re going to have a change of career and we want to get more involved in AI. And I think your career through SAP is a good example of that.

Raghu Banda: 

Yeah. I very well watch for what you have mentioned. See, the thing is, when all these new technologies came in and we know that there are a lot of business processes out there realized and then we know that there’s a lot of data, the first question obviously when we get in there, we keep asking to answer to solve a particular business challenge or a business problem, can I continue using the regular rule-based mechanism and get it solved, or does it need additional algorithms or additional mechanisms that I have to do? And when the new technologies have arrived, you obviously want to try out some newer or different things. And as we all know, the exponential form of how the technologies increase, you could- see every two years, you see that it is exponentially increased and now you have a lot more power to build better models. So that’s what it is.

Dayle Hall: 

So let’s progress this into one of our first topics, specifically around AI in the enterprise. And obviously, one of the things that is a massive opportunity is to improve, make more efficient within the enterprise, actually, our own business processes. So do you have some examples, based on your experience, your expertise, how AI is actually helping to transform business processes? If you’ve talked to customers, or what you’re seeing within SAP, I think that would be really interesting and would give people that listen to this the opportunity to say, hmm, maybe I should be looking at that kind of thing, how could I do that in my own organization?

Raghu Banda:

There are different examples that I can quote. I can pick up a procurement line of business, or I can pick up sales line of business. Let me pick up a very common scenario. Many people or many enterprises, or even consumer software companies can chip in or can even realize, right, you have a lot of sales inquiries coming in. I would like- as a sales representative or a sales manager, I want to understand which of these sales inquiries will get converted into a sales quotation, and then how I can create a sales order out of it. Once I create a particular sales order out of it, then I have to deliver some item with that sales order. It might be a online service or it might be a particular item, so on and so forth. There is this complete process starting with a sales inquiry to sales quotation, sales order and delivering a particular item, and then monitoring that item, and then obviously, finally getting paid, and then monitoring all this.

Dayle Hall: 

The get paid part, that’s important.

Raghu Banda:

That’s important, right? That’s our bread and butter. So in these various different steps in this process, you understand that there are some things that you could improve to make the process better, provide better insights, increase the efficiency, or even provide better experiences as such and how do you even deal with this complete scenario end to end. So many of the enterprises, whether it is SAP or the other enterprises, have built into or leveraged different AI technologies on how you can improve your business process end to end.

Dayle Hall:

Yeah. And I think one of the things that we hear more when we talk to customers is, if you start with what are we trying to solve could be for the sales process, call to cash, customer experience, whatever, if you start with that and think about, we mentioned this earlier, which is what are you trying to solve for the customer, and the customer could be you if you’re doing something internal, but what are you trying to solve it and then look to see where AI can actually help, will it help the process? And I think one of the things that I’ve heard over these podcasts that we’ve done recently is still, the human is involved, the AI is an assistant to the human so they can go faster, it can be more efficient, but it’s not a replacement for the human. I think that’s your sales, that example. I’m sure you’re seeing that even with your own customers that they want to augment what the humans are doing. It is not a replacement.

Raghu Banda:  

Yeah. I completely agree with you. I think we are far, far away when the AI bots will completely take over human beings. So currently, I think we are still- just like how calculators were replaced with computers, I think we did- there are additional things that how humans could increase their performance and potential. So in the same way, I think I would differentiate between enterprises or consumer software companies or customers using AI-assisted tools or business processes rather than not using them. So obviously, if you use these processes and tools, you will be in a better shape to provide additional benefits to your customers. And coming from the enterprise world, you just don’t look into only your customer, but you also have to look into your customers’ customers, that is where you have two level of complexity,

Dayle Hall:  

For sure. But talk to me a little bit about the challenge of potentially helping with AI business processes or AI-infused business processes when you’ve got a cloud versus on-prem solution. So at SnapLogic, for example, we have many customers. The newer companies are probably just pure cloud. Many customers are still trying to figure out how they move some of the on-prem to cloud. And I think there’s a little bit of a perception that you can only use AI in cloud. If you’re running SAP or AWS or in Snowflake, there is a little bit of that like, well, I can’t do this because I have a lot of on-prem. How are you coping with that with your current customers and is that a stumbling block?

Raghu Banda:

I know that’s a tangential question you have put. I understand where you’re coming in from because nowadays, we see a lot more conversations going on about the cloud, the rise of cloud and the rise of cloud customers and how AI can improve the processes. And then with the cloud, now we are also going into edge computing, right? Like now we are saying, in the cloud, you end up using a lot more data, a lot more space, and then you have to pay more. So I would say that we have different solutions, whether it is cloud or on-premise or edge-based services. When we talk about the cloud solutions, obviously you have- most of these AI services could be running as a service, you could pick up the particular functionality and showcase the results in the context of on-premise customers or on-premise solutions. 

Of course, AI would still benefit the on-promise customers a lot more. The way you deal with it would be like there are again two different ways, either you can have an enterprise bot, which is implemented on to your on-premise solution, or a machine learning service, which is already implemented into your on-premise solution, or you could run it off as a cloud service though you’re running on a on-premise. So this is where your hybrid frameworks or hybrid solutions will come into picture. And yes, we do work on all these different kinds of solutions there.

Dayle Hall:

Yeah. That’s one of the areas where we don’t talk about that as much publicly. Like you said, we’re talking about AI in the cloud, mainly. A lot of customers we talk to, I’m sure it’s the same from your side, they’re still trying to manage that. So it’s good to know that you can still take advantage of that not to be as nervous, a lot of these enterprises are already nervous about moving to the cloud, let alone how they can leverage AI. 

Let’s move on a little bit and talk about predictive analytics, machine learning, but specifically predictive analytics. And I have a little bit of a prickly reaction to this because I think predictive analytics as a term has been out there for 10 years, and there was a bunch of companies that were trying to create the predictive analytics space. What has happened is they’ve kind of fallen by the wayside, or they’ve had to pivot to be something other than predictive analytics. 

So first off, what I like to get from you is, when you think about predictive analytics in the context of your experience, what do you see as, is it something that is a specific vendor that provides, is it something that is inherent in everything you do now you have the data? If I asked you to define predictive analytics with your background, what would you say?

Raghu Banda:

What I would say is that today’s augmented analytics or predictive analytics that we talked about has come a long way. Yesterday’s predictive analytics was different. I think I would say, we’ve been having these predictions or predictive analytics even the last two, three decades. But the way we have done things at that point of time, we’re providing mere insights with the data that you have. So that is how we used to do in the past 20, 30 years back. You just have- maybe if you go back in time, you still know that you have your database and you have your tables, you have a lot of information there, you were just doing a query, query your database, getting the insights. And once you get the insights, you’re displaying it to the user, putting it in nice dashboards. 

So that’s how it all started. And now, when you got a lot more data available to you and you understand that based on the data that you have and the patterns you could make out, you could say that, hey, I have 10 years’ worth of data and this 10 years’ worth of data has some patterns. Can’t I make some kind of a regression algorithm, can’t I make some kind of a regression analysis or nonlinear analysis? This is how I would term predictive analytics or how we have progressed in that world. And then we still do- based on our information that you have, you do some kind of analysis, you can call it regression analysis or forecasting analysis and then you provide some details and quick predictions, these are close to perfection. But you will never have any 100% perfect predictive solution. You can never say that when you see the curve, I think we always say that above 90% and close to 100%, which is where we say that the predictive power and predictive confidence give you the better results. In a high level, that’s how I would define when we talk about predictive analytics.

Dayle Hall:

So that’s interesting that you say it’s not a perfect solution. So when you’re talking to your customers or you’re talking to internal people, how do you guide them on then measuring effectiveness of that solution? Because in general, when we say that this is what our company does or these are the things we’re going to deliver, we don’t usually say, okay, but it’s not perfect, there’s going to be some errors in there or whatever. So how do you measure the true effectiveness of predictive analytics if we’re saying it’s not perfect?

Raghu Banda:

So when I say it is not perfect, any predictive analytic solution, I think- for example, let us even go back and take what we have with ChatGPT. This is a huge language model. You have a lot of data in there and you get some responses out of it. 95% of the time, these responses are true, but there is this 5% of the time, the responses may not be true, or 5%, or in some scenarios, 0.5% of the time because it’s always that you’re understanding the patterns and providing a result. 

But as a human in the loop and an end user using the tool or the machine or whatever service, a human can always act in a different way. And that will be the first time it happens, there is always a first time in anything. So that is where I wouldn’t believe if somebody says that, hey, I can predict 100% of the time 100% of the results. So that is how we explain. And the efficacy is always provided. We provide the explainability of how you provide a result. When you provide a result, you provide the explainability by providing what are the influencing factors? And that’s how you provide the explanation behind that.

Dayle Hall: 

Do you have some examples within your own organization or experience of how people are using predictive analytics potentially around other parts of integration or automation? It doesn’t necessarily have to be a full RPA-type solution, but how is that predictive modeling then going forward and helping customers be more efficient?

Raghu Banda:

This time, let me take up an example from the procurement line of business. And earlier, I talked about the sales line of business. So maybe in the procurement line of business, we know that in the shop floor, you have the purchasing manager. I think we know that a particular spare part or a particular item, you’re running out of stock. And you have to go ahead and make a- you have to either bid for that or you have to go ahead and purchase that item. So you know that you have a bunch of suppliers already available. If you already have some kind of a contractual negotiation with a supplier, automatically an order will be placed. If you do not have a contractual negotiation, obviously, you get into a bidding process and you go through the bidding process. 

Pick up this example where automatically you, the purchasing manager or the inventory manager would know that, hey, I’m running out of something. I want to go ahead and order this particular item. Now, what happens is that the information is already transported, and information is always- already transmitted to the supplier location or the vendor location, and it is now dispatched, I would say, but now you need to understand when you’re getting it to your location, to the particular plant location. So it might take some time. And now you go back to the historical data and see that in the past, when I have done ordering these kind of spare parts from these kinds of locations, from these kind of vendors, there is a typical delay or there is a typical time window it either gets between 5 to 15 days. So based on that, you will do some kind of prediction saying that, hey, now that there are these additional parameters of weather parameters, there is a blizzard coming on or a storm coming in, so the truck routing has to be different. 

In addition to the parameters that you already have in the system, you have these additional parameters in there, you could build an additional custom solution on top of that and provide additional predictions and say that, hey, this time, the item might be delayed by four days, or this time, there is no storm and we already are shipping to some other location. So on the way, I will be able to deliver this as well. So maybe this time, I’ll be able to deliver one or two days ahead of time. So these kind of predictions could already happen based on your historical data information and running through the regression analysis and adding additional seasonal data as well.

Dayle Hall: 

That’s interesting. Basically, if you’re in that line of business, if you’re in procurement, using that modeling, the data you get actually helps your production, customer fulfillment, all those pieces, where you can get more accurate with it, which I think is interesting. Would that help during something completely unforeseen like COVID? So during COVID, we obviously have a bunch of supply chain issues and so on. And I know that will affect the modeling because it says different data, the data that you put into the system is probably different. But is that something that- I don’t anticipate we’re going to have another pandemic or something as major as a disruption. But I just wonder whether is the modeling that we now have because of COVID, does that actually help us to get better with this data, or is it just such an anomaly that it just throws the system off?

Raghu Banda:  

Definitely, it helps you during these kind of pandemics. And I do not exactly remember, but I think during the COVID pandemic, even SAP, I think that specifically the Ariba supplier network, the Ariba-based solutions, I think they have done some kind of additional things which helped their suppliers and the network during the COVID pandemic. I do not know the exact details, but you can go and search it up. 

But to answer your question, just from a personal experience and talking to different people and also the background that I have, yes, it will definitely help during the COVID pandemic. And I will also add an additional parameter here, right? With these IoT technology, with the IoT sensor data, that also getting integrated. Now that we have this additional not only your historical data patterns, data that you have in your system about the suppliers and the vendors and the parts, seasonal data also. 

But in addition to that, I was also adding the weather data, maybe from IBM Watson. And then now you add the IoT sensor data, which is where you have these different- there are quite a lot of these IoT embedded applications that are coming up now wherein you could now say that the truck is on the way, but there is a delay. And this IoT sensor information constantly keep delivering the information saying that, hey, at this point of time, I’m here, those kind of things could also be leveraged in providing better prediction capabilities. So real, real-time applications, yeah.

Dayle Hall:

I like that concept of even when major things happen. Okay, COVID is a little bit more of a major thing than a snowstorm or major crash on the freeway or something like that, but I like the idea of there’s so many different data points that are out- even outside of what you are doing in your own business. But if you can start to pull those things in, everything that you do in terms of how you manage your business, supply chain, fulfilling customer needs and so on, it would just get tighter. Even if there’s- even if you have to tell your customer there’s more of a delay, well, guess what? We’re using so many different sources. You’ll be accurate with telling them that there’s a delay and here’s why. And I think if you’re a customer of anything, you just want to know where things are and just don’t want to be kept in the dark. So I like- I think that would help a lot of major enterprises in their delivery. 

Talk to me a little bit about- so using these kind of tools that are available now, how a human interfaces with them. What I mean by that is more of like the design, how are we using these kind of predictive analytics tool and how important from your mind is the interface and the experience and how you actually interact with these tools? Is it really important, or is the data more important? What would you say is paramount?

Raghu Banda:

Let me now tie up to my earlier thought process when I started with people processes and technologies with data being the underlying element. So you have now we are in 2023, we are at a stage, we got humongous amount of data. Whichever enterprise firm or whatever it is, you have matured enough technologies. Obviously, the technologies will keep improving. We have realized business processes. 

So people is the factor that is very important now to understand how better you can make your business processes or your enterprise processes, leveraging the data that you have, implementing the tools that you have at your disposal, so it all ties up to how it can help the end user. So the point that you brought up is great. Even when I’m designing the system, if the system is adaptable for the end user, whether it is a novice end user or an expert end user, because you will have different kinds of people using the system, when you are doing the design, it should be simplistically available for people to understand on how I can use the system so that it can deliver the results. So obviously, user-centric design, persona-based designs play a very big role when we talk about AI-infused business processes. That’s the reason the people factor is very key there. 

Dayle Hall:

Yeah. And a lot of the discussions that we’ve had on this podcast series about AI in general, we haven’t really focused on that. We’ve talked about responsible AI and AI ethics and starting with a business case to make sure AI is what you need. But the more and more, I start to think about the user interface has to be key. Do you think it’s better if there’s a separate interface as you’re using some of these, or should you be using existing tools and AI is underlying? Does it make any difference?

Raghu Banda:

This is where digital assistants will play a very big role. There are situations where you cannot redesign the system, you cannot change the user design in such a way that it can make intuitive for the end user to use the AI-centric processes, but you could always have an additional integration layer using the digital assistant. So this is where the digital assistant or your conversational AI in some aspects can help you fasten this process. Eventually, you can migrate to a better system or a better user-centric design.

Dayle Hall:

In the future, if you look at AI outside of the enterprise, what do you want it to do for you and your life in the future? What’s the most exciting thing for Raghu in your personal life? If AI could solve something for you, what would it be?

Raghu Banda:

That’s a very great question. I think one thing which I am very fond of, apart from the enterprise side of the things, I’m very big into biking and I’m very big into traveling and very big into reading or listening. I would want- when I go on biking or when I travel, I want to listen to some of these conversations. Earlier I used to read, but now I do a half and half of this. There are a lot of conversations happening out there, but I could only get recommendations about these bigger players out there. But there are a lot of these smaller conversations that are happening which you do not even know that these are happening. So you have to do a lot more search on that. 

So if AI can provide me a snippet of what is happening out there, maybe this is one of those conversations. So there are a lot more conversations like that happening. But I can only get the big conversations happening out there like the Elon Musks of the world or Lex Fridmans of the world. But there, my biggest challenge is that there is so much information out there, but the visibility is not there. If AI can provide the visibility out of a lot of the conversations happening, that will be helpful in a different way. I don’t even know how it will come out, but- this thing.

Dayle Hall:

I like that. I have one which is with 10,000 martech tools, I get 300 emails a day from martech. And we do it, too. Like I get it. I’m a marketer. I know how these things work. But if there was a way that AI could help understand what I have in my business, what our challenges are and then actually could better filter or could actually go out and find those tools, the things I should be focused on, that will help me in my daily life. And then the rest is mainly if I could have an AI interface where I can understand where my kids need to be at what time of day and who’s driving over there and their friends could take them and all that kind of thing just to make me not the parent Uber for my kids every single evening, that would be a massive win. 

Raghu Banda:

Of course, definitely, yeah, that personal side of the thing, I didn’t go. Yeah, there are a lot of things I would definitely want there.

Dayle Hall:

Yeah. I appreciate the time today. We’ve covered a lot of very different topics. We’ve delved into some of the AI topics that we haven’t dealt with before. I love the concept that you talk about, which is the people process and technology are the key things around AI, but the underlying data always has to be there. I love thinking about customers’ customers. So when we have these technologies that we’re putting out into the market, don’t just think about how we’re helping customers. Is it really helping them with their own customers? I think that is definitely something that people should be out there as they’re listening to this thinking about, hmm, have I thought about that, or am I just focused on what I’m giving to my own customer? And I think that, if we think about it that way, that will actually, hopefully promote even more of this responsible and ethical AI and the outcome. 

So Raghu, thank you so much for being part of the podcast today.

Raghu Banda:

Thank you, Dayle, for getting me on your podcast. I really loved the conversation.

Dayle Hall:

Great. Well, thank you, everyone, for listening. This is the end of the episode, and we’ll see you on our next episode of Automating the Enterprise.