- [Ryan] Welcome everybody to another episode of the AI For All Podcast. I'm Ryan Chacon. My co-host Neil Sahota is out today, but we have our producer Nikolai here. - [Nikolai] Hello. - [Ryan] On today's episode, we're going to be talking about AI in Industry 4.0, along with a lot of other exciting topics.
And to discuss this, we have Sateesh, the CEO of EdgeVerve, a company that is focused on helping enterprises discover and automate processes. Sateesh, great to have you here. Thanks for being on the show. - [Sateesh] Great to be part of this and hi Ryan, hi Nikolai. Thanks for inviting.
- [Ryan] So let's start this off. Let me ask you from your perspective, if someone wants to say, all right, we know about digital transformation at a high level. But where do things currently sit right now? What is the current state of digital transformation from your perspective? - [Sateesh] We're a digital platform company based out of India serving customers across multiple industry segments globally. And basically in the process of really helping them become more connected as an enterprise and unlock multiple values that comes from having a connected enterprise.
And that's the purpose of our existence. Now, coming to your question around digital transformation, this has been in progress for multiple years, if not decades. There have been many phases in this whole transformation.
And we, if you move from the ERP space and you're getting into the internet world and then get into the cloud world and comes the IoT world. So you see, digitization has been an evolution. And every enterprise has embraced all of these technologies in different ways, right? And some are more successful than others. But what, I believe is still missing in this whole transformation is to drive tremendous agility. While there's a lot of investment that's been made, enterprises still don't have the necessary speed and agility to respond either to their customer needs or industry changes that are happening. And obviously, things that happened just a couple of years ago during the COVID time just exposed a lot of this, right? Even today we have supply chain issues and also ability to really move products around the world and serve your customers became really challenging.
So there is some ways to go in this whole digital transformation. And while we're talking about this, as you mentioned earlier, here comes this whole new age of automation along with AI as another, if you will, transformative technology that now has to blend into this digital transformation journey. I think digital transformation is a journey. It's a process enterprises embark on.
And that's where I think this whole thing is right now. - [Ryan] You mentioned something, you talked about agility a second ago and talking about helping companies become more agile in the way they engage with customers, do business, grow, you name it. Also, I'm sure internally, there are agile things that can be done to help the workforce just be more efficient. What do you think leads to or contributes most to making a company more agile? - [Sateesh] I think it is about connecting. I will use the word connection and connected in different ways.
See why if you were to just peel the onion and look at why are enterprises not as agile as they would love to see themselves be. It's because either they don't have data in their systems which are, which have common language. So essentially allowing multiple systems to talk to each other. The interpretation of data is very different. A SKU can mean very different things in the supply chain as you move through the whole value chain, as an example. Or you can have tasks and processes that are disconnected, right? And or you can have customers who really interact with enterprise have a journey.
It's actually they believe is a very smooth homogenous journey, but it's quite broken when it enters enterprise because it goes through multiple organizations. So there are silos in organizations. So there is a lot of, if you will, disconnected activities that happen within an enterprise. So that essentially impedes agility, decision making, the ability to access data at the right time, right place, so you can actually act on it. And these are things, Ryan, that's been discussed for a long time, but it's still an issue, right? So that's what needs to be solved.
And when I say agility, it is about can you really react to your customer's need in a manner in which that's not only satisfies the customer in terms of the kind of response, but also being very timely, as an example. Can you make sure that you can actually ship and serve a product wherever the customer base is irrespective of where you have your inventory place today. There are multiple ways that agility of an enterprise gets tested. - [Ryan] Yeah, I think that access to data and the ability to not just have more data and get data maybe that you weren't getting before, but also the ability to analyze and interpret that data interact with that data is something that a lot of the new age of digital transformation is focusing on, right? With IoT, for instance, it's all about being able to access data from the physical world that you weren't able to access before.
And then you bring in the AI side of it, the models and the different kinds of solutions that are available there to be able to interact, interpret, analyze that data in a better way. And the other side of it is not just for the internal organization to be able to interact with that data and have access to things that matter but also externally as well. The customers having a better way to see and interact with the sensor data that's being collected that they have to deal with for their own jobs as well. So there's a lot of interesting things that we've seen, but I think at the root of it, a lot of it has come down to that access to data, like you said. But if we were to focus on because we've talked many times on our other podcast about IoT and how that's really helping enterprises get access to the data.
But once they have that data, from your perspective, what are you seeing really happen now with the growth of AI solutions and tools to be able to do more with that data for a business that may be listening to this episode, understanding what the power is behind it and why they should care about it so much. - [Sateesh] Ryan, at the end of the day, we used to talk about it in the analytics world, the outcome is as good as the data. And it applies in the AI world as well, similar things. It's as good as the data that you got. And see the harmonization of data, while obviously the sensors are out there collecting different, if you will, data on the physical world, interpretation of that and being able to really triangulate information from disparate set of IoT devices into something that is more contextual to be actionable at a business level, I think it needs a lot of harmonization of data, translation of data and harmonization of data. I think that is one level of, if you will, data, if you will, governance that has to be put in place.
The second is this data essentially becomes, if you will, the fuel or food for AI, right? At the end of the day, when you look at AI or the models behind AI, it learns everything by the data that you feed, right? And that could be digitized data, either it can be actual text format, it can be audio. It just does not matter what it is. It can be video. But all of these is essentially data that AI uses. And in fact, most of the AI companies today, Ryan, probably have access more to money than they have access to data. So what I think enterprises really sit on is their data, right? I think that is, and it is very contextual, right? While there are lots of models and if you were to look at ChatGPT, in fact, It so happens that Verve and Infosys, the company that we, I work for, was the first company to contribute to OpenAI in 2015 because we believed in the power of what AI can do.
And obviously we didn't see through what really has come out now, but it's heartening to see how far the AI journey has moved. But what we really see is that while you have ChatGPT and many of these generative AI models out there, for enterprises to really realize the full potential of this, it comes down to how good of data you have because it's very contextual. - [Ryan] That's a great point. That's actually a question that was coming to mind as you were saying this is what makes enterprise data useful, right? Structured, unstructured data, you know, how are they, we've talked about how they can collect the data potentially from the physical world at least but obviously there's other types of data that matters to them, but when a company thinks about data, like, that's great but what makes that data useful for an AI solution or tool to be able to do more with it so that the organization becomes more efficient, has better understanding of their data, can make better decisions, all the things that this is set out to do. But what is it that really makes that data useful for them? - [Sateesh] So let me maybe pick an example and walk through it.
So there are multiple different industries, as I said, where we have implemented different kinds of solutions and using our platform. And I'll just pick insurance and financial industry. There is this whole process of underwriting where there is an assessment done on the person who's really calling for the insurance, you really, you do a very deep risk analysis, if you will, before the underwriter makes a decision on whether or not should the insurance be approved? If so, what is the risk behind it? It's a pretty, it's very core to the insurance company, the whole process. But if you were to peel the onion, process itself, now there's no IoT in this, but it doesn't matter. If you peel the onion, the individual, the underwriter, who's probably a very highly paid person who has a lot of experience behind to make these somewhat subjective decisions as well sometimes, has an enormous amount of data coming into play, right? It may be history, historical data. It may be historical data of the individual or the company that is actually is in consideration, or it could be data of many such underwriting approvals or disapprovals that you may have made and how they have performed over time, right? There's so many other data points that comes to you, right? So all this data, many of them are hidden in digitized documents, right? They're in documents and you and they don't have really access to that insight in the document, right? So what, if you will, what we do is actually take these documents, uncover some of the most important data elements in those documents, bring them out, and start feeding that to the AI model, so the AI can start understanding and interpreting, okay, if you were to see a request of this kind and this kind of risk, you can actually make a recommendation to the underwriter on how they should really consider this particular case, right? So so what we're doing is actually building that very contextual data layer, right? Whether it is in documents, it seems something very similar applies to IoT.
And then this data feeds into the model, which starts learning, and it's not only about, you know, a model that learns with just data, it's also the model also learns with human interaction because so there is this whole thing about human in the loop where the model actually observed how the decision was made by a human and then learned from that. So there is the two, so the whole improvement, if you will, on the decision making using AI is two part process, which is the kind of data that you feed, the contextualization of that, the clarity of the data, removing all the biases and et cetera, all of that stuff. And plus also allowing the model to interact with the human, so it can learn. So this is how slowly but surely these models become richer and better over time.
- [Ryan] When it comes to the adoption and the implementation of these models within an organization, what's the process that a company listening to this would need to go through in order to see that value, see that benefit to not only collect the data that could be used, but also make that data useful and bring these different tools and solutions into their business. How should companies that are listening to this be thinking about that? What's important to note? What advice do you have just to bring this a little bit more directly relatable to them to say, look, if you understand the value of this data, here's how you would go about doing that, at least at a high level. - [Sateesh] There are two parts to this.
There is the data element and there's a human element, right? And from a data perspective, all things to do with cleansing of data, making sure that you have data that you trust. And it's obviously the biggest issue, Ryan, in the AI world is can I really trust the decisions it makes, right? And the first, if you will, input to getting some kind of confidence around the trust level is what you really feed into that model. So data governance has to be a very important discipline that has to be brought into every single enterprise, right? I think that is extremely critical.
Yes, you may have had some level of data governance initially or earlier in the digital transformation journey, but I think we are talking about a very different order of data governance right now. Because you have machines making a lot of decisions in this case. You know while earlier humans were able to see reports to make decisions, which is analytics, which is different than actually machines taking decisions by themselves, right? So data governance really has to go up a few notches, right? And not really double clicking on what that means right now. The other element is also around the human.
We have seen, right, obviously there is a lot of concern, you know, around AI and what it could do to one's job and, you know, the whole world and all of that stuff. So really to embrace successful AI journey as much as one will invest in the data governance and all things to do with data, it is equally important to look at the human aspects of this, right? Because what is really happening in the world is we're creating, if you will, software avatars of humans. This is what this is all about, right? At the end of the day, we've seen physical robots and physical humans, if you will. Now these are software avatars of humans. And obviously humans have to engage with it, humans have to really understand it, and let me call them digital workers, if you will. Really, you need to build that rapport.
And at the end of the day, what we have seen time and time again in every single enterprise, while it's very early days still, we see they're more complementing to humans. They're not necessarily taking away jobs, they're elevating, if you will, human productivity, human ability to make decisions, allowing humans to really do what they're really good at, right? We're good at creativity, we're good at empathy, but there's so many God given attributes to us. Unfortunately, we use less of that because it so happens that enterprises are configured in a certain manner. I think AI enabled these, enabling these digital workers, I think will really make life far more better for humans over time. While there are yes there are concerns that have to be managed and mitigated, but really, it's actually going to be beneficial to humans. It's not going to, it's going to be less of a threat, but more beneficial is the way.
So you have to deal with data and human elements both. - [Ryan] And I think it really comes down to the individual's willingness to adopt these technologies into their workflow as well. Because as any new age of technology comes into the workforce, if you are one that resists it, you can oftentimes be left behind because other people become, level up, become more efficient in their work, are able to push off a lot of the menial tasks that they may be doing to these AI assistants, AI tools, you name it, and be able to focus their time on things that add more value to the organization. And I think once you amplify that out across the entire organization, it's a no brainer on what this can possibly do, but it requires the humans to be willing to adopt and not be stubborn and push back on these different things in fear that their job may be replaced because I think it just is going to make people do their jobs better. And yes, of course it is going to replace some jobs, but it's also going to create other jobs as well. We've talked about this in the past episodes with other guests and like call centers and things like that, where it may become more automated, more AI driven, but it's going to create the need for AI managers and things along those lines.
So I believe that enterprises and individuals within enterprises who are more open to adopting, spending the time to learn how these tools, how these solutions can benefit their business and their individual jobs will be the ones who have the best opportunity for growth and organizations going into the future. When it comes to bringing these solutions in, introducing them to the workforce, adopting them in some capacity with a business, we've talked about the benefits, and we've talked about a lot of the upside. But I imagine at the same time, there are also going to be challenges with doing this. One of them is this human element and willingness to adopt. But what are some other unique challenges you've seen or you predict will occur when companies and enterprises start to more readily and widely adopt AI tools and solutions into their business, whether it's for themselves or for their customers? - [Sateesh] Another great question. This is something that we will discover as time progresses.
But a few things that, it becomes, it's so obvious when I start talking about it is if you start seeing software avatars of human, they will probably be thinking like humans and acting like humans. So some of the human behaviors will manifest within it and that needs to be managed and monitored and checked. As an example, you know, we have seen so many scenarios already out in the world where, you know, these things can hallucinate. How do you really differentiate between real and hallucination? How do you make that happen? They can be biased.
So hence, how do you make sure that they, you need to put mitigations. Today we talk about security and privacy, big things. But I think as AI comes in, there will be many other dimensions to these checks and balances that needs to come, which is around the biases that I mentioned. Maybe is it more invasive, tests around invasive? Is it more greedy? We may not think of it, but at the end of the day, it is software which is thinking like humans.
So it's very likely that it can become very greedy in the way it starts making decisions, right? It may lose some amount of business rationality over time, right? And it can be brittle. It can be opaque. So there are so many other expressions that we have never ever dealt with in the context of software automation till now. I think we'll now become part and parcel of how AI gets implemented and managed. Yes, there is a human element, and we need to mitigate and manage. There's the data element.
But I think enterprises will have to embrace a whole new set of, if you will, controls and structures to manage AI. And it is not about just harnessing AI, right? You have to literally, you know, if you will, manage it on an ongoing basis, right? We've already seen the models that GPTs have put out can actually start underperforming over time. You would imagine that actually the performance will go up, but it may not be necessarily true. It may actually performance may go down. So it's very important that these structures that I mentioned are embedded into the, become an integral part of enterprise, if you will, structures over time. - [Ryan] It's a unique challenge, I think, for a lot of organizations, but the more use cases we see out there in the world of companies adopting these technologies and solutions and bringing them into their business for the benefits of their organization and growing their customers, engaging with their customers, their solutions, you name it, I think that resistance to adoption where those hesitations from companies, I think, will start to decrease, right? Let me ask you.
So we've been talking more generalized throughout most of our conversations so far. I wanted to ask you about an industry we haven't talked a lot about, but I think it seems relevant based on what the work you all do and potentially customers you've dealt with. But when it comes to Industry 4.0, we talked about that and on our IoT
podcast, it's a really popular area. But in AI, it's something that hasn't really gotten a lot of attention in past discussions. But I think it's important to talk about the role AI is going to play. IoT is driving, solutions are being adopted very heavily in the industrial space, but with that data, like we're talking about, now what do you do with it, and that is where AI can come in. What have you seen lead the way when it comes to industrial organizations and their adoption of AI? - [Sateesh] See, there was a time, right, you have the manufacturing floors where humans would work.
And then slowly but surely we've automated, right? There are lots of machines and IODs and robots and all of that stuff that actually does the job far more better. Now, if you were to look at software enterprises, you look at SAPs of the world or Salesforce, it just doesn't matter, any Oracles and so on. They are like, now, in the software world, I would equate them to be the manufacturing flows, right? There are lots of humans on it doing lots of different things. While implementing those ERP systems brought certain advantages in terms of standard operating procedures, best practices from other industries, et cetera, it also created a whole new generation of inefficiencies. And I think so this when you look at Industry 4.0 and everything to
do with it, to me, it is about how do we really realize transformation about delivering real time decision making on top of this new platform, enhance productivity and flexibility or agility, that's what it is all about. So it's about eliminating more and more human activity, connecting all these disjointed, but what we also see in large manufacturing organizations, Ryan, is these enterprises, there are multiple. So you have lots of islands created and this is all happens either because organically it so happen or you've gone through mergers and acquisitions and so on so forth, so you've created lots of islands. While at a very aggregated level you see a lot of synergies, but when you get to the systems and processes, there's actually no synergies. You have duplications and so on so forth.
So when you look at industry 4.0, to me, you know, in the context of AI led automation, it is about how do you really homogenize all of this. See, you cannot take all these systems away and put new systems in. It is not as simple as that because you've invested billions of dollars into this.
So it's about how do you create another layer on top of this, which drives agility and near real time decision making, where data harmonization becomes a fundamental thing, a layer on top which harmonizes all the data. And feeds in such a way that human decision making becomes easier and then you can start moving a lot of the decision making to AI led. - [Ryan] One of the questions I wanted to ask you before we wrap up here soon is where is this all going in your mind? What does digital transformation mean, look like in the future with AI in organizations? Obviously, our guess is, at a high level, probably that adoption will increase. More solutions will be out there.
People see more benefits and efficiency within organizations, better access to data, better understanding of their data, things like that. But just if you were to talk to somebody about where this is going, why they should care, why they should be excited about this from an enterprise perspective, what would you say? - [Sateesh] There is a certain cost to intelligence in the world, and I firmly believe that cost of intelligence is dropping and dropping dramatically. If you were to really zoom this out, question is what if the cost of intelligence really becomes zero? I'm maybe playing it out very dramatically here, but I think that's where this is all going. And the role of humans in the context of that, the role of enterprises in the context of that, and the role of how this technology really transforms society. I think there are so many implications and ramifications of this technology, not only at a human level, at an enterprise business level, but the environment and society in which these business enterprises operate will all be dramatically different if we were to take this technology and its full potential, right? So I think it's going to be very interesting, Ryan, if we look and observe and shape this technology into the future.
This is undoubtedly going to be something that is not just transformative, but I think it's going to really, it's going to shake a lot of fundamental assumptions that we've made as a society and social structure. How knowledge gets disseminated, all kinds of stuff is going to change. So, it's going to be really exciting.
- [Ryan] Very exciting times for sure, especially in the enterprise space. A lot of the content out there is very much focused generally speaking about AI or consumer based AI conversations, but we've really tried to hone in our conversations to be more focused on enterprises and how these technologies, solutions, tools can be adopted, brought into a company, and make them better in different ways, whether it's internal operations and efficiency, whether it's their solutions and their offerings better, interactions with customers, the value they provide to their customers or just the outside world better. Nikolai, do you have any questions before we wrap up here? - [Nikolai] Yeah, no, I've just been listening, and it's all been very interesting. If you think about like an enterprise or a corporation, it is already an AI, like it's its own entity that has its own agency in the world, made up of all these people that don't independently have the same vision or agency, they're supporting this larger system.
We could potentially see, especially in the long-term, this become very, much more literal with enterprises being more and more automated. - [Ryan] Let me ask you, the last thing before I let you go here is for our audience who wants to learn more about what you all are doing or follow up on this discussion, reach out in any way, what's the best way they can do that? - [Sateesh] edgeverve.com is the company address, and we got information on what we do. And we have a LinkedIn handle where we're active pretty much on all the things that happen to the company and how we service our customers, right? So that would be the best way to reach out, Ryan. - [Ryan] I appreciate your time.
The audience is going to get a lot of value out of this conversation. So thank you so much for taking the time and hopefully we'll find other ways to talk again soon and do more content together. - [Sateesh] Thanks Ryan. Thanks Nikolai.
It was great having this conversation. Thank you very much.
2023-10-06