Orchestrate all the Things - Lexmark wants to lead by example in IoT, unveils Optra Edge platform. Featuring Vishal Gupta, Lexmark SVP of connected technology / CITO

Episode Date: March 23, 2022

Arguably, two of the most iconic examples of digital transformation are Kodak and Amazon. Kodak failed to keep with the times, leading to the demise of a once dominant commercial empire. Amazon, however, had the foresight to not just “stay in its lane,” but step out and shape the world’s digital infrastructure with AWS. Now, Lexmark, is taking note of Kodak’s failure and Amazon’s success and piloting into its own. Moving beyond printers, supplies and accessories — which is how Lexmark first made a name for itself — the company’s current motto seems to be “Print, secure and manage your information.” In 2022, Lexmark lists Print and Capture as just two of its solutions, which also include Cloud, Security and IoT. The Lexmark Optra IoT Platform, unveiled September, is featured prominently on the company’s site. Today, the company is unveiling Optra Edge, the latest addition to its Optra IoT solutions portfolio. Vishal Gupta, Lexmark’s senior VP of connected technology, CTO and CIO, says the company’s new tools and move into IoT are just the beginning. Article published on VentureBeat

Welcome to the Orchestrate All the Things podcast. I'm George Amadiotis and we'll be connecting the dots together. When it comes to digital transformation stories, two of the most iconic examples are Kodak and Amazon. Kodak failed to keep with the times, leading to the demise of a once dominant commercial empire. Amazon, however, had the foresight to not just stay in its lane, but step out of its comfort zone and shape the world's digital infrastructure with AWS. It looks like Lexmark is keen on learning, not only to avoid being left behind, but to turn its vision and lessons learned into products others can use. Founded in 1991, Lexmark is recognized as a global leader in print hardware

service solutions and security, according to the company itself. Moving beyond printers, supplies, and accessories, which is how Lexmark first made the name for itself, the company's current motto seems to be print, secure, and manage your information. In 2022, Lexmark lists print and capture as just two of its solutions, which also include cloud, security, and IoT.

The Lexmark Optra IoT platform, unveiled in September 2021, is featured prominently on the company site. Today, Lexmark is unveiling Optra IEDS, the latest addition to its Optra IoT solutions portfolio. We caught up with Lexmark Senior VP, Connected Technology and Chief Information and Technology Officer Vishal Gupta to discuss the backstory of Lexmark's move into IoT, Optra IoT adoption and Optra Edge. I hope you will enjoy the podcast. If you like my work, you can follow Link Data Orchestration on Twitter, LinkedIn, and Facebook. My name is Vishal Gupta. I joined Lexmark a little over a year ago in February 1st of last year. I had a lot of experience in the technology industry. Prior to this, I was the global CTO for Unisys, and I was also the founder of a lot of the IoT efforts at Cisco. And, you know, I kind of play two or three roles at Lexmark.

So I'm the, you know, I'm responsible for, as a CIO for all the experiences for our employees. We have over 8,000 employees. And then as a CTO, I'm also responsible for building products and experiences for our customers. And so, you know, I kind of crossed both the internal stakeholder and external stakeholder definition. And one of the key things that attracted me to Lexmark was, you know,

our whole focus on innovation in the IoT space. You know, Lexmark has been doing a lot of work in IoT, actually, for over a decade. Each printer has over 120 IoT sensors and they had the ambition to really take that all the learning that they've had over maybe 15 years and take that to the market to enable other manufacturers and other type of enterprise customers to also get value from IoT. And so that was one of the reasons I got attracted. And in addition to these two roles, I'm also in charge of our corporate strategy. So I also look at how do we transform the company overall. And so,

you know, that's a little bit about me, George, in terms of when I joined Lexmark and the areas I'm responsible for. Okay, thank you. Thanks for sharing that. And it kind of makes sense when I hear you talk about your own background and the fact that you've been involved in IoT efforts before because well to be honest with you I did not I wasn't aware of the fact that Lexmark has been active in that area as well so I know that in September you announced the Optra IoT platform and looking and looking at the background of that announcement and the rationale behind it, what you just mentioned all seems to come into play. So reiterating quickly for the benefit of people who may be listening. So as you pointed out, Lexmark has been using IoT sensors to be more specific, in its own hardware for quite a while now and leveraging

this effort to achieve things such as better visibility into potential hardware failures. And I think you can also share some of the results Lexmark has gotten through those efforts. And the Lexmark Optra IoT platform was basically an effort to productize, let's say, that experience and that know-how for others. So I'm going to let you jump in actually at this point and we'll just share with us a few words on how the platform came along and who is it aimed for, basically. Let's start with that and we'll follow up. Yeah. So, you know, as you mentioned, we launched the platform in September in the IoT World Congress and a number of other forums. And the key thing that we try to do with a platform is, you know, given our own experience as Lexmark with not just the IoT sensors in one printer, which are, as I mentioned, over 120 sensors in one IoT sensor in one printer alone. But we also manage millions of these printers right through what we call as a managed print service. So really, if you think about it, we have experience at scale

with managing hundreds of millions of sensors. And more importantly, if you think about it, the challenge in IoT has not been so much on the concept or the fact that the technology works. It's really been more around the business outcomes, right? If you look at mckenzie said 84 of the iot projects get stuck in pilots because people typically are not able to demonstrate the business outcomes right the so what and so i think what where we were able

to differentiate ourselves was that we were able to not just use and and show in production hundreds of millions of sensors but we were able to create three or four very interesting business outcomes. Let me give you an example. What we've been able to do with these sensors is that in over 95% of the cases, we can predict when a problem will happen. And 70% of those cases, we're able to remotely fix the problem. Now, that's typically thought about predictive maintenance in the industry. And this is across our entire fleet, which is hundreds of types of printers. So this is actually very compelling because this is a win-win for both the customer who gets into much less downtime and for us, because it results into, you know, a more loyal customer, our achievement of the SLA and a lower cost for us. Similarly, we're able to, you know, go from what used to really be a one-time transaction with these customers to now with our management service, really create more of a recurring revenue stream and therefore predictability for both the customers and us. So to give you a sense of that, we have over 800 large enterprise customers in Fortune

3000 working with us, you know, where we manage these thousands of fleet of printers, which for us are one of the first IoT connected devices. And that has helped us create recurring revenue stream and create predictability for the customer. Similarly, we've been able to create other outcomes like we're able to, with all the data that we're getting from the sensors, from the printers, we're able to reduce the time to reduce new products by 50%.

So better systemic innovation, better great automation in terms of automatically sensing the level and automatically shipping the supplies without anybody touching anything. Thereby, again, improving the outcomes for both the customers and us. And there is tens of thousands of those shipped every day with nobody touching anything. anything and so what we try to do is we said okay we've achieved all these pretty amazing business outcomes leveraging an iot platform can we take it to market to enable other makers of connected devices achieve similar outcomes uh because at the end of the day uh this is just one type of connected device and so that's the that was the motivation that we had and you know we've been working with a number of manufacturers since our announcement enabling some very

interesting outcomes and one of the things we've been able to do even for our own business which even in the last maybe six months has been with all the supply chain disruption right it's been very difficult to get materials and create new things for everybody in the industry and so we've been able to also take this data that comes from these uh devices and be able to extend the life of these devices itself by almost 25 percent and so we call this smart refresh uh last year uh almost in our fleet, we were targeting maybe 50% such refreshes, but we were able to do a smart refresh of almost 80% of our devices

by being able to predict which of them after a number of years needed replacement, where we could just replace maybe a component or a subsystem instead of the entire device and do that based upon not just the data coming now, but also looking at the service history, looking at data from a variety of other systems, thereby making the right decision.

And then also that saves capital expense for the customer as well. And so that's a little bit of a genesis on why we introduced the platform, the outcomes that we saw. And initially we focused on connected manufacturers, but later this year, we'll also expand that for retailers as well.

And in this announcement which is what we're announcing now is really around the next step in iot because you know one part is okay how do you collect the data through the cloud and do all this processing and the pipelines and create all those outcomes but everybody doesn't prefer cloud or all the use cases are not relevant to cloud especially if you think about low latency uh high bandwidth use cases especially with real-time audio video and so what we're doing now is really sort of a follow-on from that to really introduce a set of edge capabilities. So if you think about a factory floor, right? We have a lot of factories ourselves. We don't connect factories to cloud

from both a security perspective and also from the perspective of the, we want 24 by seven operation on those factories, we can't afford it to go down. And so that's where I think in the market for IoT, you need a continuum of cloud use cases and edge use cases. And I believe we'll be one of the first companies in the world to actually have both. OK, thanks. Actually, that was a bit more than what I specifically asked for, but that's fine. I mean, in the sense that you also expanded to cover

what you're announcing in a couple of days. So the Lexmark opt-rights, but we'll get to that. Actually, before we do though, I wanted to ask you to clarify something which was not entirely clear to me. So when utilizing IoT and sensors and analytics internally, let's say for your own product line, it's relatively straightforward where the data is coming from. All those sensors that are embedded in each device, as you mentioned. When productizing this

though for third parties to use, where will the data be coming from? And it may sound like a naive question, but, well, I'm presuming that many of the organizations that you're addressing may be Lexmark customers. So some of the data may be coming from your own devices. But what other type of devices is the platform able to work with? And just to be entirely clear about it, it is a platform that entails software, right? It doesn't come with any hardware. Yeah.

So when we announced the Optra IoT platform, that was a platform running in software. There was no hardware with it and where the data comes from is you know we can take both streaming data and batch data the streaming data typically will come from iot sensors to give you an example some of the iot sensors we support on the wireless side will be things with lorawan or bluetooth or or those kinds of them we also support industrial sensors things like modbus or opc we also support things uh for example in building management uh protocols like pacnet and others and so because we're built on top of microsoft azure iot hub, there are a very large list of protocols that the sensors can communicate. And sometimes, you know, the native IoT protocol is called MQTT, but it doesn't just have to be

that. It could be a variety of other protocols. And that's where the edge can come in as well, because the edge can actually convert from one of the proprietary or less known protocols to maybe a standard IoT protocol like MQTT. And so we can collect data from things like cranes, from, say, dental machines, from mining equipment, really from any type of device through the use of those sensors and that data essentially flowing to the cloud with the IoT Hub. And that's sort of the classic platform use case. Okay, thank you. You did mention the Azure IoT platform. That's something I also wanted to ask

you about because just by going through some background material it looked to me like you are using Azure indeed and so that's clear already. What I wanted to ask you in addition to that is whether the platform also runs on other cloud vendors or it's exclusively on Azure. Yeah, currently it's exclusive on Azure. We certainly have, because we use a lot of open industry standards,

we certainly have the capability of running it on other cloud providers as well. But currently, Azure is where we're sort of certified in terms of knowing that, you know, it will work end to end. Okay, thank you. And let's get to what is actually the most interesting part. So thanks for sharing some details on how the infrastructure, let's say, is set up and using those open protocols. applications can users build on Lexmark's platform? Are there some out-of-the-box analytic capabilities or machine learning models or applications that users can customize? Absolutely. And so, you know, when we introduce the platform, and again, I'm talking about the platform because I think that's what you're talking about, not the edge yet, right? So on the platform itself, you know, we thought about for a manufacturer,

what type of machine learning models make a lot of sense. And we wanted to provide two types of capabilities. We wanted to provide a capability for where people could, through just a low-code way, through drag and drop, essentially build both new machine learning models. Because, you know, let's say you have a dental machine and you will need to be able to sort of predict when it will fail and it may have 40 different components. And, you know, there is sort of an iterative journey

that you go along in terms of that, what we call a descriptive, predictive, and prescriptive analytics. So you've got that capability, which we now make much easier by enabling not only a ready-made platform that can process the data in terms of a hot pad, a cold pad, but also enabling people to build these through a drag and drop,. And the idea is that, and you know, we spent a lot of time with a lot of manufacturers and what we found was that people are looking for,

they're looking for not an experiment, they're looking for experience, right? And they want it fast, right? That's why a lot of the stuff gets stuck in the pilot stage. And so some of the machine learning models that we really provide out of the box are for things like anomaly detection and you can have many types of anomaly detection right anomaly detection can be specific to a subsystem can be specific to a part can be specific to the

entire device and it'll sort of help you to decide if there is something strange happening that's causing that particular part or device or a subsystem to behave the way it does. We provide machine learning models for fleet optimization. So, for example, if you were trying to decide for a given customer, do I refresh the entire fleet or only a part of it? How old is my fleet? What are the characteristics of it? How does it compare to my other fleets? We have machine learning models for that which we use in the smart refresh concept that i talked about we have machine learning models for predictive services so our ability to predict when something might fail

and and do that with a higher level of accuracy we also have machine learning models for predictive consumption. So all the consumables are perishables. In the case of printers, you know, the consumables, the ability to predict when a cartridge will run out is actually very key because you don't want to under or over predict it. If you, you know, have an error in the prediction, you might have the stuff replaced much earlier than needed, or you might have cases where customers run out and essentially they're not able to do operation and it's critical to optimize that. So those are the types of machine learning models that we have out of the box.

But more importantly, I think we have the overall infrastructure that enables people to build the models that are relevant to them. And we think these out-of-the-box models kind of cut 70 to 80 percent of the time for them to get something that's relevant to them into production. Okay, thanks. You mentioned a no-code environment that people can use to work with the platform. And so I was wondering if basically this is the only way that people can do that. And what I mean by that is that, well, this may actually be a very good choice for people who don't have much of a technical background,

like floor managers or people with that sort of role. However, there may be the need to customize things in a more fine-grained way. So are data scientists and people with more technical background able to um uh to uh to open the hood let's say and uh get uh get uh more hands-on with uh with the code if yeah we provide we support both the personas so we support you know that's the traditional persona that typically the industry has supported which we support as well given to be built on top of open standards where you know you'll have your jupiter notebook you can you know with the python go and make any other changes you want uh but what we found out was that in a number of industries you know it's been very hard for people to retain even if they're able to go

into production now they they have a hard time in terms of retraining those things as new data comes right so that's where we give them the choice To say you could leverage it with both an expert and also with people who are not an expert. So you can not just put them into production initially, but also keep them updated as you go on. Okay, thank you. And just as a curiosity really, are there specific machine learning frameworks that are supported?

And the reason I'm asking you that is that, well, if many data science, for example, must be familiar with the kind of standard industry frameworks such as deep learning frameworks and scikit and python and so on. So are there specific frameworks that are supported that people can reuse their skills in? Yeah, so all of those are supported, the ones you mentioned, because those are just part of common standards. So as part of OpenML framework, they're all supported, the ones you mentioned.

Okay. You know, the challenge is not so much in supporting those standards, the challenge is how do you make them easy, right? So that's the problem we're trying to solve. Yeah, absolutely. And so that brings us to the latest addition to the platform, so the Optra Edge that you're about to announce. And as you briefly mentioned in your introduction this is basically expanding the capabilities of the platform with adding the the option for analytics and applications to be deployed at the edge that means there is no roundtrip no extra latency in sending the data to a cloud platform and also in terms of security

and compliance which is which very often dictate that this is the case so I wanted to ask you again what types of devices does the opto-edge work with and this is a more specialized question than the general case because if you if you move the computer and also much of the data storage I presume to the edge then it means probably that the range of devices that can be supported is limited in the sense that there probably some kind of minimum requirements to be able to do this. Yeah, I know you're exactly right. And so what we are doing in Optra Edge is really providing an end-to-end solution ourselves. This means we're supplying the hardware. So we're essentially providing two types of compute and two types of what we call as Optra Vision devices. The compute devices actually, even though we're announcing now, have been available for about a year. We already are deployed in thousands, especially in retail

stores where they've been running things like, you know, ads or music or things like that, that, you know, we've got a very interesting use case with a partner that they run and they kind of create, you know, an outcome for those retailers in terms of both creating a new revenue stream for them, as well as creating an ability for those retailers to essentially create a more interesting environment in the stores. The Optra Vision devices are, you know, our place where we have, you know, we talk a little bit, as you know, about open standards. So, there what we have done is we have two devices and they they leverage uh you know a very uh one of them device is more targeted towards the recording of all of those camera uh pieces and analyzing things the other is really around uh really uh what we call as the detection of objects.

So it's really more on visual detection capability. They're built on top of NVIDIA's chips that essentially provide very high-end processing in terms of being able to do this type of analysis. And this is actually, you know, we're providing not just the boxes themselves, which are, you know, which are also certified by Microsoft. So all these four boxes are Azure level one certified,

which means that they can instantly connect to the cloud. Not all the time, but they're typically for provisioning. They're kind of managed from a central place. So that way the customer can easily manage them on an ongoing basis if they have to update them, update the firmware, do anything. And then what they have is we have taken Docker containers as a standard protocol in terms of we're able to provide both

the hardware the management and the actual end-to-end application and so to give you example of that we're introducing when we introduce optra edge we have about 20 use cases that are built out of the box in the manufacturing world, in the retail world, in the parking or transportation world, where people will be, customers will be able to essentially run that AI on the edge for very specific outcomes. And I will give you an example of an outcome which we have leveraged ourselves. So as I mentioned to you earlier, in the manufacturing floor, you don't really want, you know, you don't want to be connected all the time to cloud because if the cloud goes down,

you don't want manufacturing to go down. But intermittent connection is typically okay. And so we, you know, when you think about a manufacturing line, through the cameras, you can do visual inspection as part of quality, right? So the visual inspection can check for, is the right packaging done? Is the right screws put in place? Is the right screws put in place?

Is the right wires put in place? Is the right, all the other capabilities in place to essentially make sure that this thing meets the standards. In the past, humans have done it. Now through this Optra Edge device with the AI running in it for visual inspection, we're able to do it and we create amazing outcomes. For example, we're able to do it and we create amazing outcomes. For example, we're able to do things like 40% improvement in the inspection speed.

We saw that in three months, we can actually break even in terms of investment. We are able to achieve almost a 99% reduction in the errors that were happening manually. And so what this will let the customers do is whether you are in manufacturing for all these visual inspection use cases, or you're in retail for a number of use cases that we have in retail, or for example, you're dealing with logistics in terms of parking or the movement of goods, we can create some very interesting outcomes. And so that's what the Optra Edge announcement is about. It's to really create that open ecosystem where both we, because we're providing these, we're also, by the way, working with about a dozen startups. So think about it like an app store. We're providing

a number of AI use cases ourselves but a number of startups you know who are working with us who don't have a big go-to market are creating the capability to essentially run that ai in our uh opcred devices which are sort of can be managed from the cloud and thereby creating a very easy to use solution to create very specific outcomes for the manufacturers, for the retailers, for even, you know, folks who have, say, vehicles and logistics and things like that. And later on, towards the end of the year, we'll introduce even additional verticals as well. Okay, thanks. And that sort of addresses another question I had to you, which was going to be about vertical integration and what it meant exactly. So I think it makes sense now having

having listened to what you just said. So basically, as opposed to the the initial, let's say, Lexmark Optra IoT platform, in that case, you, you don't just provide access to the initial, let's say, Lexmark Optra IoT platform. In that case, you don't just provide access to the platform, but you also provide the hardware and devices and the sensors and everything that goes with it. And you also mentioned that you have some partnerships going on with third parties, startups, and so on, that they basically leverage this infrastructure

to create domain specific applications if i got it right you're exactly right and so you know because we think the amount of innovation you know um uh george uh if you look at the predictions right they talk about edge becoming a 284 billion dollar market i think what will be needed is not so much how big that edge market is, but how do you make it relevant to solve the problems that people are having, right? How do you create those business outcomes? And so what we're trying to do with both the platform and the edge is to really give our enterprise customers choices of being able to really become ai and data rich and leverage those capabilities to create very specific outcomes for them and give them choices as well okay so uh let's wrap up

with a little bit on the um the business strategy basically for this so i'm presuming that addressing existing customers, brownfield case, let's say, was probably your starting point. And so I wanted to ask you on that, how has it been working and if there are any adoption metrics or specific use cases that you can share. And for the second part, I was wondering if it is within Lexmark's strategy to also use this as an opportunity to address potentially non-customers and not existing customers following in the footsteps of, well, the birth of AWS, if you could say that. Yeah. No, it's a great question. So I think we certainly look, when we are looking at this market, we're looking at a combination of both existing. And, you know, if you look at it, we have, for example, 100 retailers as existing customers today, right, in our managed print

services business, which is where we're managing printers, which happen to be an ID device. So that makes a lot of sense. But what's interesting is what we're finding is, George, that there's a lot of customers who are not existing customers, especially in the manufacturing side, because we don't have a very big manufacturing base, even though we're manufacturing our kind of company ourselves, who are very interested in partnering with us. And so I would say what our experience is so far, that it's been 50-50, even on Opter Edge that we're reducing, you know, we're engaged in about maybe a dozen opportunities and it's half and half.

Half of them are existing, half of them are new. In fact, as you know, this week HIMSS is going on. And, you know, we were just engaging with a number of healthcare providers, out of which I think they were about, yesterday, about a dozen of them engaging with us, out of which probably four of them were only existing, eight were new. But they saw so many interesting use cases for themselves that they could see, you know, why they would use it. business perspective is obviously start first we start with ourselves because we want to prove that this thing actually works and creates real outcomes because you know we're very large enterprise customer ourselves so our teams are very relentless in terms of they want very easy

to manage very high security very easy to change things right and so first we try to perfect the technology ourselves then we try to typically take it to existing customers because they uh you know there's a trusted relationship and they can give us very open feedback. We create kind of an advisory board. And then simultaneously, we end up also engaging with new customers to, when we have the confidence that this thing will work. And so that's the stage where we're at, where it's sort of been half and half. And what we're also doing is we're also having some of the system integrators uh likes the likes of cognizance of the world and even microsoft itself be a very strong partner with us on the go to market side where they're not only certifying

these but jointly talking to customers uh as these things kind of run on their platform as well as i mentioned on Azure. And similarly, the system integrators like Harkness Center are building a practice where they can help in rolling these out for large customers. Okay, well, I would say that all things considered, I mean, the fact that it's relatively early in the lifecycle of this offering, it sounds encouraging if you say that you have about 50-50 split

in terms of existing and new customers. So let's wrap up by asking you, well, to share a little bit on your roadmap, let's say. So what's next in the coming year after releasing Lexmark Optra Edge? Sure. next in the in the coming year after releasing lexmark optra heads sure so as you know in september we released the uh the optra platform which was specifically for manufacturing we call it bishop now in march uh end of march as you know we're releasing the optra Edge, which initially is focused on manufacturing, retail, and some of the the logistics use cases, and we'll have about maybe 20 use cases out of the box. What we want

to get by the end of the year is both to get into retail for the platform, so we want to extend the platform use cases to retail as well, and we want to scale up our number of edge use cases to retail as well and we want to scale up our number of edge use cases to about uh 50 or more and that's where the work with the startups becomes very important in terms of uh you know adding a number of use cases like right now we may have about five retail use cases i want to actually expand to say 10 retail use cases so it doesn't cover just the front office of retail but also covers the warehouses and how do you manage that and all of those pieces as well and so that's kind of what we see uh for uh the year uh 22 and then you know uh we're also looking at how do we not just think about platform and the edge separately but even in the same use case think of them together and how do we enable that

seamless integration where you know the data the the ai gets trained in the cloud and then the on the edge it essentially gets executed but then all the metadata of it gets sent to the cloud so that way it can get retrained better and better and so there's sort of very interesting use cases from an integration perspective that we're working on as well. And the next year will expand into more, both more verticals and also more areas of integration. I hope you enjoyed the podcast. If you like my work, you can follow Link Data Orchestration on Twitter, LinkedIn and Facebook.