Utilizing Tech - Season 7: AI Data Infrastructure Presented by Solidigm - 2x03: Deploying AI in the Business with Per Nyberg of Stradigi AI

Episode Date: January 19, 2021

Per Nyberg of Stradigi AI discusses "blue collar" AI applications with Stephen Foskett.  What problems can businesses solve with AI technology? Machine learning can find anomalies and outliers in manufacturing and finance, look for relationships in data, and cutting through the complexity of multi-disciplinary data. Consider customer churn: Machine learning can discover features in profiles that might not be visible even to an expert. Data scientists and AI experts must learn to present AI technology to average business people in terms they can understand, and this has lead to a "haves/have nots" situation where some companies or business units don't have access to this technology. We also need to reduce the science fiction appeal of AI and express what it can't do. Guests and Hosts Per Nyberg is Chief Commercial Officer for Stradigi AI. Find Per on LinkedIn to continue the conversation and connect on Twitter as @_PerNyberg or @StradigiAI Stephen Foskett, Publisher of Gestalt IT and Organizer of Tech Field Day. Find Stephen’s writing at GestaltIT.com and on Twitter at @SFoskett. Date: 1/19/2021 Tags: @SFoskett, @_PerNyberg, @StradigiAI

Welcome to Utilizing AI, a podcast about enterprise applications for machine learning, deep learning, and other artificial intelligence topics. Each episode brings experts in enterprise technology together to discuss AI in today's data center. Today, we're discussing blue-collar AI, bringing AI to the enterprise from a sort of consultative and operational perspective. Let's meet our guest, Per Nyberg. Hi, my name is Per Nyberg. I'm the Chief Commercial Officer at Strategy AI. Thanks for joining us, Per. I'm Stephen Foskett, Organizer of Tech Field Day and Publisher of Gestalt IT.

You can find me on Twitter at sfoskett. So, Per, I was having a briefing with Strategy AI the other day because that's sort of what we do for Gestalt IT. We learn about new companies and technologies and what they do. And the thing that really stuck in my head was this sort of practical approach

that the company takes toward AI. In other words, you're not looking at it as a science project or a mathematics doctorate. You're looking at how you can bring AI to the enterprise in a really practical way. So I imagine that that's something that's near and dear to your heart personally as well. How do you approach AI when a company comes to you and says, hey, we got to get some of that AI goodness? Yeah, great question. And I think that for just about every company that's looking at AI,

that's, I mean, often they'll come in with some notion of benefit, right? And I think increasingly companies are looking for value out of their investments, right? And so our approach is really start with the business question. So it's not AI for the sake of AI or machine learning for the sake of machine learning. It is really you start with, you know, what what's the business question? What's the use case?

What is you know, what are they expecting to get out of this this project or, you know, their sort of implementation and really start there. And what we found is starting there just drives not only sort of the right metrics and examination of the metrics, but even down to what data is required in order to support that use case. So it's kind of an organizing function, if you will, to really, we feel, look at AI from a more practical, business-oriented perspective. And in terms of, I guess, what you're actually doing, so what does the company get or what does the customer get from the company on these engagements? Yeah, so we're an AI SaaS platform provider, right? And so, you know, our customers use the platform to build models on their own. And those models would be, you know, could be several models, but ultimately is to support this, let's say, for example, demand prediction, right? So where our discussions first start is around exactly that. So the customer will say, you know, we've got challenges in our supply chain. We've got, you know, thousands of SKUs, nine month lead times. You know, we're struggling to make accurate predictions

in, you know, inventory purchases. We would like to use data-driven decision-making to improve that. So that's kind of the starting point. And so in our first engagements is exactly around that. We then look at what data is required to support it. And in our first sort of engagement, the steps of our engagement, we'll actually build those models with the customer and show them the results, the actual results that they can benefit from using their data. And then from that point on, you know, they become users of the platform and they're, you know, they're doing this on their

own, right? But sometimes it's that first case of seeing what data can do for their businesses, right? And it varies greatly depending on, you know, quite frankly, kind of the maturity or experience of the organization with using machine learning and even just using data for these types of decisions. So let's talk about that a little bit then, because I know that, well, there are a lot of different

companies in a lot of different spaces doing a lot of different things. What problems do you look for that are particularly amenable to using AI? And I guess maybe we can also say what kind of AI, because that's an important consideration as well. I mean, I assume that most people are talking about sort of deep learning or machine learning when they're talking about this, but what kind of problems are you looking for in a company? And you say, aha, that's the one that I should apply this machine learning solution to or something like that. Yeah. So maybe a few dimensions there.

One is that, I mean, machine learning, as you know, is a horizontal technology. It's not any specific vertical or use case. It's math, right? So in general, you could say that just about any problem, business challenge that has data, you could apply AI to it. It's obviously a little bit more complex than that, but that's generally the case. If you look at where most companies have a tremendous amount of data that they've been gathering, it could be around their customers, for example. That's actually one of the... What we see as one of the fastest growing, most important areas is just around, you know, customer behavior modeling. But it could

be back office data, supply chain, kind of demand prediction that I mentioned earlier, all the way through to, let's say, IoT data for manufacturing. So you kind of look for areas that one have the data, because that is a necessary prerequisite for AI. So you look for the data, but more importantly, it tends to be around the business challenge. And again, AI is not the solution to everything. So we get customers who approach us with a business problem and we actually respond to them that you could use very conventional statistical methods to solve those things. You do not need machine learning. It seems that many AI applications are really focused, though, on finding, you know, unknown

needle types in unknown haystacks. You know, essentially, you know, you're looking for outliers. You know, other AI problems that are being solved by businesses tend to be of the sort of find me a novel solution to this question kind of, you know, challenges. In other words, you know, cut through all the potential paths that we could take and just find one that seems to work. Does that match your expectations? Are there other sort of general problems that companies are solving with, well, specifically with like neural network, you know, machine learning? Yeah, I would actually frame it that, so certainly outliers is one, right?

Anomalies, right? And there's use cases around fraud detection. It could be even identifying defects in manufacturing, let's say one of the most important ones. And the other is just the kind of complexity of the data, multidisciplinary data, right? So again, going back to the demand prediction example, you have the ability with deep learning to start adding in other variables. It could be weather, for example, right? Things that are unconventional, right? For most businesses. So you can, the power of these methods is that you can look at very complex relationships. And I think that, you know, that notion of complexity

is probably one where it's that tipping point for a lot of companies. An example would be, I don't know, you're a food supply company and, you know, you could live happily with Excel spreadsheets and one person looking at this, let's say once a week or once a month, but suddenly the disruption in your supply chains, the pace of the business, either competitive or just being driven by your customers, knowing where to stock inventory, to optimize logistics and transportation, to reduce inventory spoilage. You see how it starts to get very complex and it's kind of beyond the human. So, you know, we really see a lot of these use cases that I think are most impactful for businesses are often where AI should be viewed as a tool, which is just kind of making your current teams better, right? I mean, these are

very smart individuals. It's just that their Excel spreadsheets are into triple letter columns and whatever number of rows, and it's just, it's too much for into triple letter columns and whatever number of rows. And it's just it's too much for them to be able to make sense of. And so the other aspect is that, you know, when you're when customers look at how they build models and they look at their data often to your point about kind of unknowns or needle in the haystack they actually end up learning about what it is that drives decisions and it could be let's say you're looking at customer churn and you're trying to understand okay so what is it that would you know cause a you know the classic telecom customer to cancel their contracts. And it could be, you know, aspects of their profiles that just were so deeply buried that you didn't know that those were the drivers. So one of the

things that machine learning can do is actually tell you what features in your data are driving the decisions, right? And that kind of, you know, spurs on another investigation by the subject matter expert to further investigate these features. And it could be data that the company is not actively collecting today, but now know or understand that they need to actively collect it. So it's really this kind of journey of better understanding your business drivers through the lens of your data to make better decisions. So another aspect that I'd like to zoom in on there, and one of the things that you mentioned is this idea of multidisciplinary expertise.

In other words, an organization may have expertise in security and in networking and in logistics and in IT or different areas, but not all those people can see all the data, not all those people know all the data. And, you know, an AI model could theoretically cross those disciplinary lines. I think that that's sort of what you were talking about in there as well, right? Yeah, exactly. And, but again, the model is just, it's just math, right? And so it certainly has the capabilities of looking again, for those relationships between, you know, different, different types of data, multidisciplinary data. And, you know, but I think that's where it comes back to this human in the loop. You still need that subject matter expert to make sense of it.

Um, because it could be those, those kind of classic false positives in there as well. Right. And, uh, you know, at least for now, I think most use cases really are, are, are reliant on that human expertise, that subject matter expertise to be able to make sense of these things. AI, in most cases, is just a tool that helps them, you know, understanding their data. So I want to turn the page here a little bit too, because I know another area of focus for you all, and something I think that's pretty interesting to our audience is how do you translate a lot of this sort of egghead stuff into real world business speak? I mean,

how do you go to somebody who's not a, you know, let alone an AI expert, but a, you know, computer systems expert or a data scientist? How do you go to somebody just in the business and say, hey, this is a good solution for you because whatever. I mean, what is the whatever? What is the pitch to an ordinary business person? Yeah, so it's a great question. And maybe I'll back it up a little bit. And I think one of the, I mean, AI has a tremendous amount of promise. I don't think there's any debate about that. I think it's interesting when you look at what the kind of, the challenges have been or the roadblocks or the kind of limiters have been

to not, and I wouldn't even just say the adoption of AI, but the use of AI in day-to-day business decisions. I mean, companies have been exploring and how AI could be used for many years now, but most reports would say that the percentages are still pretty low of those using it in production. So that kind of tells you something. And so for AI to really live up to its promise, it needs to be more accessible,

like any technology, right? It needs to be able to be used on a day-to-day basis, the same way that, I don't know, like an Excel spreadsheet is used today, right? If we as a community can get to that point and really put that power in the hands of literally anybody, that that's that's when it's going to be game changing for companies. And so today you've got this scenario where there's there's kind of the haves and the ability to invest in these large specialized teams of data scientists and, you know, PhD machine learning experts to, to build a handful of use cases and that's okay for them. But what you're finding is that you've got the have nots in the industry where companies simply just can't get started. So even with the large companies, as I mentioned, it might be just a few use cases that they're focusing on. So maybe in a bank, they're focused

on fraud detection, but the marketing department doesn't have the tools to use, to sort of leverage all the data they've been collecting about their customers. So even within these large organizations, there's kind of these pockets. So all this to say that, you know, AI needs to have a degree of automation such that you don't have to build these from scratch. You don't have to be a machine learning expert and really, you know, again, you know, kind of make your current teams better. Now that could be a biologist or it could be a business analyst. And that's really kind of the pitch, if you will, when you go to customers, it's like

you have the expertise today to be able to use these tools because of the automation, because of the user experience, they can, in fact, build and deploy very sophisticated models without having to be, again, a machine learning expert. I guess the flip side of that discussion, though, is that considering the level to which popular culture kind of glorifies artificial intelligence and robotics and so on. We have to think also of how do we express to companies what this technology can't do and isn't. Because I think a lot of people, when they hear AI, they assume we mean Mr. Data from Star Trek or the Terminator or something like that. Or maybe they even just assume sort of pop culture AI that can do marvelous things, you know, the classic, you know, crime show, you know, enhance, enhance, let me see that face, let me see that license plate.

And, you know, I think that maybe there's also a parallel challenge where, you know, we in the industry have to start figuring out how to express the limits of AI? Yeah, absolutely. So a few thoughts there. One is that there's always going to be some degree of, let's say, fundamental research that maybe border on a little bit of science fiction, and that's okay. That's great. So i think it's a little bit of of everyone needs to decide where their focus is uh and certainly you know ours is very much on industry right um and how we can put you know ai in the hands of of business people um and that goes back to you know what i was mentioning early you frame it up as a business problem and that goes back to what I was mentioning early. You frame it up as a business problem. And that tends to kind of reduce the science fiction aspects of it.

And there's often some education that's required. Because there's no question that a lot of people really don't understand what it does. They may not even understand that it requires the data that it does. But I think if you start with a business problem, what is a value to the business, it tends to sort of solve many of those things on itself. All right. So I guess in a nutshell, what is the value that businesses are deriving here? I mean, what's the, you know, the elevator pitch for a CEO or a CFO maybe, you know, I mean, what do you say if you meet somebody and they say, oh, what could AI do for me?

Yeah, I think it really comes back to, you know, being able to improve your day-to-day decision-making through data. That is fundamentally what it is. The ability to put in the hands of your business analysts a tool whereby you can leverage and make sense of very fast-moving, insightful data so that you can move the needle within your organization, right? That's fundamentally what it is at a high level. You know, the data shows that the companies that have adopted AI, have adopted, you know, a digital approach, enjoy faster revenue growth, for example, right? So there are hard numbers behind this as well that do demonstrate those companies that have been leaders in this space, that they are seeing the

benefits. So you're going into an organization and you're bringing in, you know, a pitch for AI. They implement it, they hire, they develop, they apply technology. What does this organization look like in a few years? What kind of people are working there and what are they doing in the future once they've adopted this technology? Yeah, great question. I think that what we see in organizations is that there's kind of this role of the business analyst that is evolving, right? So a business analyst, I mean, they've been around for a long time, obviously, and

their role is really around, you know, again, driving better decisions, informed decisions, data-driven decisions within an organization, right? So they've always had that. I think what we're seeing is that there's a greater emphasis on the data side of things, right? Understanding the power of the technology and to your earlier point, the limitations as well, right? So being able to advise the company on where these, you know, technologies could be applied. And so we're seeing that today already. And you're also seeing this within MBA programs. There's a number of MBA programs out there that now have, you know,

MBA major in data analytics as an example. So you're starting to see the kind of early stages of the workforce and the graduates starting to evolve to really, I guess, have a better ability to understand, again, the power of data and to bridge that gap between the business and IT in these organizations. And I think that's fantastic. I think ultimately, as I mentioned earlier about making your current teams better, I think that's very true. And I think that as the skills evolve

over time, really AI can just start to be used on a day-to-day basis within companies. So I'm interested in this sort of future AI or data MBA that you mentioned. What specific classes are they going to be taking? What skills are they going to be bringing in? You know, what makes them different from a, you know, today's business management graduates? Yeah. So, so, so I think there, there's some, there are some unique things around AI, which do need to be understood by, by, by people making decisions with AI. And it can be around ethics, for example. It can be around bias. Again, AI is just math, right?

And so you do need to have, you know, some kind of a careful approach at times, depending on the type of data. But, you know, understanding the limits and, you know, just being able to, you know, guide the business to make the best decisions and which includes also understanding its limitations. So those, for example, would be some of the specific skills. I think they'll evolve to be able to use, just like they use Excel and they do these very powerful visualizations with tools today, it would be exactly the same sort of a thing with AI.

It covers a spectrum everywhere from understanding data, understanding the limits and the potential of AI through to day-to-day working. That's great. And I imagine too that folks listening who want to be more involved in the industry can start looking at this as well. You guys have a lot of AI-focused employees. What resume line items are you looking for that really kind of help separate people from the pack? Maybe not you particularly, but what are companies going to be looking for? Again, I think a lot of it just comes back to that data analytics experience. And today, you might be hard-p pressed to look for an employee who might have,

you know, 25 years of experience and AI, right? So some of this is often, you know, newer grads that might have more specific things around AI. But fundamentally, I think it's those individuals that have a strong competency in data driven decision making. And those people have been around for a long time. It's just that the, you know, the, the, there's more data, there's more complexity to the data and there's that multidisciplinary aspect of, of kind of critical thinking and decision-making that those are typically the things that we hear organizations are looking for

in people. Yeah, absolutely. It is interesting. So just as an aside, my son is actually studying data science and artificial intelligence in college right now. It is amazing the things that he's learning versus the things that folks in industry have not yet learned. So I think that education is adapting rather quickly to this new field. And I imagine as well that we're gonna see a lot more people with this kind of skills on their resume entering the workforce and coming in to do exactly

what we've been talking about this whole time. Another thing as well that I'll just comment on a couple of minutes ago, you mentioned ethics and the fact, we're just dealing with data and we need people to help us to sort of understand the implications of the data. That's another topic as well that we focused on on the podcast here. And I'm glad that you brought it up because that's very important to us. So if listeners are interested, go back to some of few episodes, the season two, as it were, of this podcast, a lightning round where I surprise you with some fun questions. And I just want to hear your gut reaction to some of these things. So I'm putting you on the spot. I would apologize, except we do this for everyone. So I'm not going to. So I'm just going to pick a couple questions.

Give me your one minute thought on each of these things. Okay, ready? All right. Good sport. Okay, here we go. How long will it take for a conventional AI to pass the Turing test and fool an average person in a conversation, like a verbal conversation? 15 years. All right.

So we're not there yet. We're not there yet. No. All right. When will we see a full self-driving car that can drive anywhere at any time? Great question. So I think five to 10 years would be my guess that that would be, that'd be around that. Just, just as an aside, I'll surprise you a bit. My old colleagues and I, this is about five, six years ago, we had a running bet as to when this was going to happen. And I can tell you some had predicted 2020 and that that didn't happen. So I, I think we're, we're still a little bit of the ways away,

but hopefully not that far because it would be great. Well, my young daughter has decided that she doesn't need to get a driver's license because she'll just be able to tell the car where to take her. So, you know, we'll see if that works out. Okay, one more. Are machine learning, deep learning, and artificial intelligence synonymous or are they different? They are subsets of one another. So machine learning is a subset of AI and deep learning is a subset of machine learning. All right. Excellent. Thank you very much. So, and thank you for joining us for the conversation today.

Where can people connect with you and follow your thoughts on enterprise AI and other topics? Yeah. Well, first of all, thank you. This has been a great conversation, but I'd welcome people to follow me on LinkedIn and on Twitter. All right. So just look up Per Nyberg on LinkedIn and Twitter. You'll find them there. And thank you everyone for listening to the Utilizing AI podcast. If you've enjoyed this discussion, remember to subscribe, rate, and review the show on iTunes, since that really does help our visibility. And please share this show with your friends. This podcast was brought to you by gestaltit.com, your home for IT coverage across the enterprise. For show notes and more episodes, go to utilizing-ai.com, or you can find us on Twitter at utilizing underscore AI. Thanks for listening, and we'll see you next week.