Olivier Bloch, principal program manager, Azure IoT developer ecosystem, Microsoft, is back to talk with Peggy about the complexities with IoT (internet of Things) adoption and the biggest changes companies are facing with many vertical markets. They look into how Azure IoT is helping companies simplify IoT with IoT Plug and Play and the best things to do when adopting an IoT plan and establishing initiatives.

Below is an excerpt from the interview. To hear the entire interview on The Peggy Smedley Show, visit www.peggysmedleyshow.com, and select 1/21/2020 from the archives.

Peggy Smedley:
You and I have had many a conversation about the benefits of IoT (Internet of Things), but during the past few months, there’s been a lot of discussion about how there’s some complexity in adopting IoT. What are we doing? Are we doing it right? Are we doing it wrong? What are the biggest challenges for IoT adoption day?

Olivier Bloch:
Yeah, you’re right, Peggy. And I think we’ve done a good job at convincing people that there’s value in adopting IoT in people’s digital transformation, right? And while the companies, the bigger and smaller companies, are looking at IoT as really something they’re going to extract value from and grow their business and transform the business, they’re realizing when they start working with IoT technology that there are several challenges that they have to face. And some of them are things which have security end-to-end. We have new types of threats, new types of devices, connectivity, and so on. The data, as well, has to be gathered somewhere, stored and you need to comply with regulations. You need to make sure your customer’s privacy is guaranteed and so on.

There’s another aspect of these challenges which is the skills gap. We see the technology evolving super fast. IoT, AI…they gather to drive the digital transformation, but engineers might not be ready or enterprises might not have the right skillsets. They might be developing solutions in the manufacturing world and now they’re told, “You need to be experts in software and data analytics.” But that’s something that they need to catch up with, which is not straight forward.

But one of the biggest challenges that these enterprises or customers are seeing is the complexity of their projects in the implementation of IoT solutions. We recently published a report called, “The IoT Signal” that revealed 30% of companies say that complexity and technical challenges are the biggest barrier to the adoption of IoT and to their digital transformation. 38%, that’s a huge amount of companies that start their IoT project with a clear idea of what they want to achieve and they end up very often failing at proof of concept, not even going to production because they have to face these kinds of complexity that we’re going to talk about today.

Olivier, that’s an interesting point because when I was looking at your “IoT Signal” report, that 38%, that’s the largest number. When you chip away at staff, knowledge, engineers, skills gaps, and security, the highest one was complexity. And you would think that companies would understand that IoT is not easy. Is it because they don’t understand IT/OT? Do they not understand the things that are associated with it? Because there’s a lot of things, a lot of moving parts in the Internet of Things, do companies not understand why IoT is so complex?

I think you’re definitely pointing at something that’s very interesting to look into deeper. So, I think companies have been, in the last few decades, very much focused on adding value in their vertical, in their industries, right? So, the company building medical devices is trying to build the best medical devices that are out there, complying with regulations and so on. People in oil and gas are trying to build the best solutions for extracting oil and be the most efficient as they could. But IoT is the technology that’s going to glue several worlds together. I like to really talk about IoT as the glue, the IT and the OT world. How do you put together the IT world, the traditional IT website, data analytics and so on with the OT world, especially in manufacturing where you have to operate at a different pace, at a different reason with different tools?

There’s also the collision between the cloud development and the embedded devices development. Two worlds that have been living aside or side-by-side, I would say, but has never really overlapped very much. Development cycles are also not the same in the various world or are not colliding. So, this is something that everyone has to adapt. When you have a team building some PCB, some hardware for a specific project, the cycles will be longer than the team working on the software. They will gather the data, extract insights, and eventually send reports to all the teams. The scale of these deployment is also one of the big complexities arising.

The standards are pretty much nonexistent. These worlds are gathering in that new era of digital transformation, but they are not talking the same language. They’re not using the same APIs to interact. The market is super fragmented. Think about all the hardware manufacturers, the ones building sensors, the ones building PCBs and microcontrollers, the ones building the PC-class type of hardware. The connectivity providers, devices have to be connected wireless, wire different protocols. This is also a huge fragmentation issue that we’re facing in the IoT world. Line of business software providers, they are very diverse and when you want to extract value from IoT, you want to take action, you want to indicate to your crew that there’s maintenance required. You want to change the schedules of your workforce.

So, you need to interact with loTs of different light of business applications. Cloud platforms are diverse as well and you need to interact with various of them for implementing an end-to-end solution that includes IoT. All of these fragment the IoT landscape in a way that makes it super complex for companies to really find their way, implement their solution the way they want, and they need.

It’s interesting to hear what you were saying Olivier, because something else that came to mind is you were talking about standards, so I was thinking about inter-operability. I was thinking about things that meant cloud development versus embedded development. Some of these things that the line of business just with software developers and who you work with, and you were talking about sensors and microcontrollers, the same pair of shoes doesn’t fit for every vertical market. So, there’s got to be a lot of challenges then whether you’re in healthcare, whether you’re in manufacturing or transportation, it’s different. And even within those verticals, every IoT application is very unique. I think that’s also making it more complex in some ways when you drill down within the vertical or when you go horizontally amongst those verticals, am I correct?

You’re perfectly right and we’re seeing that these challenges related to the complexity of IoT projects is not specific to a single industry. You would think, “Hey, maybe manufacturing is where there’s the most complexity happening.” Actually, we see the same kind of problems and issues our customers are facing in retail, in transportation, public sector, or healthcare. Pretty much every single vertical that is going through digital transformation sees companies failing at implementing their IoT project because they are facing that complexity, an IoT application an IoT project end-to-end represents.

And so they need to find the right set of tools, the right set of platforms, but then not to have to reinvent the world, but also to be able to implement a solution they need and to basically have the complexity removed for them by a platform provider, by not really having to reinvent the wheel. There’s tons of things that many companies out there already have done, that you can use and reuse. And this is exactly the approach we’re having at Microsoft. We’re trying to simplify that complexity to offer our customers tools, platform, services that will really empower them to transform digitally to realize what their potential is and to really implement the solutions to go where they want to go, transform their business, grow their business. And we really want to be their partner in that journey.

And that’s really important. I think when companies hear that they think it’s a sales pitch, but there is a lot of truth to that, simplifying the complexity and changing the future with IoT. And Microsoft is really trying to do that. Can you walk us through a little bit about what you’re trying to do there? Because I think when people think about edge computing and they think about the cloud, they get overwhelmed. When we think about manufacturing specifically because they’ve been late to the game, so to speak, to really embrace IoT, but now all of a sudden they’re realizing they can actually be more competitive. I think healthcare jumped on a little bit sooner and things like that, but I think what’s happening now is those manufacturers are discovering we have to do this and when we do it they see really positive results, but at the same time they don’t know how to do it and they have to partner with companies like you in order to make IoT successful.

I totally agree and we’ve seen that over and over again. So manufacturing, it’s definitely an area where we can use a lot of samples because people understand that especially in our world, right? Healthcare is a more particular area, retail as well, but manufacturing is a great example that can allow us to really look at the various places where there are complexities, there are issues that we can help solve. A good example is the various types of protocols that exist in manufacturing for devices to communicate. You might have heard about OPC in OPC UA (unified architecture), you might have heard about BACnet and CAN Bus or other protocols that allow devices to exchange information in between each other or with another broader system. The promise of digital transformation is really to harness data that you gather from sensors to optimize your processes to extract the insights that you can extract from that data and then take action.

And one of the aspects that was mentioned is connectivity. These devices don’t all talk the same language. They don’t also all declare their capabilities to a broader system. So you’re really today or in the past, because we’re changing that hopefully, you have to know what a device is doing and how it’s talking before you can interact with it. You have to have custom-made solutions for each of these categories of devices that are out there in your factory floor or other facilities that you do manufacturing in. And one of the technologies we are developing with the industry is called IoT Plug and Play. Plug and Play is based on an open source description language. We called it The Digital Twin Description Language. And we work with the industry and we aim at standardizing that. It allows hardware manufacturers to declare their device’s capabilities and to offer cloud developers and solution developers a very easy way to import that device’s capabilities to then understand what a device is sharing as data as information, but also what a device supports as commands, as settings, as properties.

IoT Plug and Play aims at becoming a way for our partners, hardware vendors, and solution builders to work better together. Having a platform that allows the hardware manufacturers to declare device capabilities on one side and a solution builder to just use that model, and, even better, a catalog of devices available online to enhance the solution, to enrich the solution with more data come from all devices and then being able to really deliver on that promise that IoT offers.

IoT Plug and Play is one of these examples. The work we’re doing to simplify the connectivity and the device capabilities description that is out there. There’s no standard out there for that. Everyone is doing its own way based on the verticals they’ve been evolving in, because verticals were silos in the past and now, because of that digital transformation we’re talking about, they have to work with other industries. They have to start mix and matching.

Think about retail. Retail’s been changing all the time. I would imagine you’re having to do a lot as software as a service on the retail end, correct? I mean I can’t imagine because that’s an industry that I can’t even keep up with its changes so quickly to be competitive.

Totally. Retail is another great example where there’s is a need for advanced platforms that will accelerate developments. So in retail, there’s many things you want to do in order to optimize your stocks or deliver a better service to the consumers. There’s various aspects to it. There’s security as well, there’s privacy for the customers. And if you spend all your time when you’re building a solution, starting from scratch and redeveloping everything: “I’m going to build the piece that connects the device and then I’m going to build the piece that stores the data securely and then the piece is going to do some analytics on that data and then the piece is going to be…” and on and on and on, all the way up to delivering dashboards that allow both the store manager and the person in charge of delivering and eventually the company that is shipping the goods to the various stores.

If you are building the solution end-to-end that can take years if you start from scratch, we have an approach to that which is called Azure IoT Central, which is a platform solution. It allows you to very rapidly get storied with a Wiziwig experience, creating not just dashboard in a website, but also managing access control for different types of users. Also, connecting devices, leveraging the Plug and Play infrastructure I was mentioning before, allowing you to say, “I’m going to add a new type of device. Let me pick from the catalog. Oh, that camera actually has a model.” I can just import a camera in my solution and without even developing a single line of code I’m now able to integrate a new type of device into a solution and really focus on adding value and like in the retail space, it could very much be about, “Let me think about the algorithm that will enhance the flow in my store so that my customers are actually exposed to more goods.” Or, “Let me also optimize on making sure that all my shelves are always full.” And so if I have my cameras that do analytics, they’re able to detect that I’m missing this kind of product or this kind of goods and I can’t actually restock without having to send someone inefficiently through the aisles to see what’s not there anymore. Retail definitely is another great example of a need for platforms that have a lot already developed, already provided for you. So you don’t have to deal with that complexity. You just have to deal with adding value on top for your vertical, for your industry with what you really know and are experts for.

When you talk about IoT Plug and Play, IoT Sensor, Azure IoT is doing a lot in a lot of different markets. Are there specific customers that you say, “Look, here’s some customers who are having great success right now, whether it’s in transportation, manufacturing, or whatever market and we can list those customers and say they’re making really great profits from using our stuff?”

Definitely. We have many customers who are already using our solutions. Azure IoT has been built in the last five years, adding new services. We started with just connecting devices, right? We started simple. I think there’s one philosophy that we want our customers to really embraces is the idea of: start simple. Start with something that is a building block for the rest. We started with connecting devices securely and then we went on to providing more features to control and manage these devices, provision these devices at scale and then added analytics on top, added IoT Central for the UI part of things and so forth. And so a customer I’d like to talk about is called Buhler and they’re doing something that we don’t realize is critical for many industries. It’s called die-casting. And die-casting is a process that is basically about molding metal. They have these huge machines that will inject the very high pressure melted metal into these molds in order to create this metallic object that you need for appliances, that you need for automotive, that you need for electronics devices as well. And the process has been 150-years-old, but it’s not changed a lot in the last 150 years and technology is not been able to really optimize it. There’s lots of waste going on. There’s lots of things that actually fail in the process.

And in order to optimize this process, to make it more efficient, to save a lot of money, Buhler adopted our technologies to gather tons of metrics from various sensors into the molds and extract as much insights as possible in terms of, “Here, is the process going well? Do I need to use that kind of temperature everywhere or adapt the temperature?” And optimize the very old and very important process of die-casting, thanks to little sensors and analytics happening in the cloud. And insights that have been extracted to then take action during the process, adapt the temperature, the pressures and other factors to become one of the leaders in that space, thanks to IoT. This is one of the manufacturing examples.

Another interesting one is Maersk. That’s a company that you might have heard about that works in transportation, electronic equipment, geolocation and others and you see them in lots of public transportation, like metros, buses, and so on. And there’s one thing that IoT offers is the ability to gather the data from millions of devices. However, you want to trust that data, so you want to make sure that all these devices that have been what we’ll call “provisioned” securely, there’s been given and authentication mechanism that allows them to come to a system and say, “Hey, I am so and so and I’m going to share my data.” And for the system to really trust that this device is what it says it is.

So, it’s something that is like fairly simple to do when you have around 10 devices. Now, when you have millions of moving devices from different manufacturers deployed at different customers, it can very rapidly become a nightmare to have to manage all of that. So leveraging a platform like Azure IoT Device Provisioning Service that simplifies the provisioning of these devices is key for a company like Maersk to evolve, to grow and to embrace that digital transformation.

Now isn’t that why you guys have made a $5 billion commitment to IoT, so companies that you’re talking about can have success?

Totally. I think it totally goes in the mantra that we have at Microsoft, which is about empowering people. Empowering people doesn’t mean empowering consumers only. It also says empowering enterprise customers. Empowering them because they know their business and what they need is help in creating more business for themselves in a very sustainable way. And so our investment is really in order to develop these platforms that will help these enterprises thrive by not having to deal with the complexity, by not having to reinvent the wheel, but by having them leverage existing bits and pieces that can aggregate the way they want to need to create solutions. That’s definitely what this $5 billion investment is about. Microsoft heavily invested, these years, in IoT because of that reasons. We want to be with our customers and partners, empowering them realize their potential.

Are there other customers that stand out to you that you say, “Look, this is important. This is why I’m telling this story because as we have these complexity issues as we have maybe the skill gap issue…” These are customers that are finding a way to do what they need to do to be successful, to be competitive, to stand out in their vertical markets.

Totally. There’s another aspect, was not discussed a lot that relates to complexity, which is security. And we’ll get to that at some point. But I wanted to name Starbucks as one of our customers. Starbucks is leveraging Azure Sphere, which is our microcontroller security solution in order to securely connect their existing infrastructure. Starbucks, as you know, is that coffee chain that we like and that has these very specific recipes. And recipes are not just about, “Hey, how much coffee from where…” It’s more about the processes making the coffee itself, so that when you go from one store to another one, it has the same taste and it has the same quality.

Starbucks is very protective of their recipes and they had not connected their coffee machines to the internet because they don’t want these recipes to be stolen. That’s something which is critical to that business. They decided to adopt the Azure Sphere guardian module, which is a way to connect an existing infrastructure to what we’ll call a “Brownfield solution” to the cloud in order to be able to not just gather information, but also gather information from lots of different locations, lots of different devices and being able to extract from that data that insight that is necessary to make that process even better.

The coffee brewing process is very much related to the temperature, the pressure in your location and tons of factors. When you look at data coming from one single coffee machine, you can vaguely optimize that coffee machine. But then if you start mixing and matching that data was data from hundreds of other locations, now you can start visualizing or detecting trends or patterns and make that process of yours even better.

That requirement for connecting these devices really fed into delivering an even better coffee to the customers, but without that risk of losing their IP. And so Starbucks really came onboard with Azure feed that delivered a turnkey solutions for very simply connecting the existing infrastructure without having to reinvent the wheel and very simply make a non-connected store a connected store that would share all its data almost overnight. Starbucks is another great example of a company that wanted to rely on something that was actually developed for them, but it would be totally adapted in working with an existing infrastructure and solution and not having, once again, to reinvent the wheel. They would use something that exists and pair it up with their own solution.

Talk a little bit about security central for IoT, because I think that’s important, but I think it’s critical because when it comes to the IoT. There’s a lot of people concerned about that and I know that’s a top priority for you guys as well.

Definitely. I think security is a top concern and it adds up to the complexity. Security, for the IoT realm, is something that brings new types of challenges. When you have any infrastructure that is totally enclosed in your building with your own servers, it’s all okay. You can put a big lock on the door, you can put a guard at the entrance and your machines would not be touched, right? But suddenly you start connecting things. Now you have the internet. So you’re going to put a big firewall to protect from the internet some threats from outside, right? Now think about IoT devices. IoT devices compliments your solution. They’re going to be gathering data at the edge where the data is generated. And then these devices will have to report through a gateway or directly to the Cloud. But these tiny devices, because they are optimized to run on battery, to be placed in very small spaces, they usually run on what we call a realtime OS (operating system) or sometimes no OS at all. It’s just electronics and a tiny bit of software running on the silicon. And these are not especially equipped for doing the very heavy lifting that encryption and antivirus things and so on are actually requiring. And so these devices, the microcontrollers that are out there, there are roughly billions that are produced every year.

Today, they’re not very much connected because of that security concern, right? If the data generated by the devices start being corrupted, you can get into very big trouble. We have some examples out there, but example that is very simple is that if you cannot trust that a sensor, a temperature sensor… they’re measuring what’s going on into, let’s say, a very expensive wine cave is telling you, then you might ruin millions of dollars of wine because that one sensor’s data is not to trusted.

So you need to be able to make sure that these devices that are physically accessible to people that might want to corrupt, steal or come into your system, that these devices are securely connected and Azure Sphere is a great example of how we thought about, “What are the seven principles of a secure device?” and then implemented them into a solution that is not just about silicon. Azure Sphere is silicon architecture that we license to silicon vendors, a secure OS that is running on this specific silicon, and then a security service Microsoft provides for free for 10 years. Azure Sphere is really a solution that brings all of that together to securely connect devices.

Another thing that we’ve been developing with our decades of experience in managing PCs and deploying services and applications is called Azure Security Service. Azure Security Service is something that allows monitoring and a cloud application end-to-end, so monitoring diverse services that compose an application to see if they’re doing well. We’re talking about the clients and the servers in terms of the physical machines, but also the software services that are running, allowing through active directory people to access these resources and these services.

So Security Center is a tool that has been developed for existing solutions and that has been extended to monitor and track security issues and potential problems for IoT solutions. It comes all the way down to devices with a tiny agent you would add to your configuration on your device and would allow tracking everything that’s going on. Then you’re going to get reports about what’s happening from the device to the computing, to the storage, to the backend, end-to-end for your IoT solution, making you way safer when leveraging this kind of solution. You deploy an application, you’re onboard to Azure Security Center and instantly you have the solution that will monitor end-to-end without having to reinvent the wheel. And benefiting from the decades of experience that we, Microsoft, have in securing IT.

You just described skillsets as it relates to security, you’ve got to go with the people who know it best. Everybody’s been talking about 5G and the hype around it, but now we get to edge computing. This is something that’s really important because we have all these important scenarios now where customers don’t have 24/7 internet connectivity. Walk me through this important thing and how much of a role is it going to play in IoT adoption in 2020 and 2021. Where does Microsoft play IoT edge and how you see that? Because I think that’s important for customers who need to go to the edge, who need to go to the Cloud and where Microsoft sits in all of that.

Totally. I think the edge computing story, something that rose when people realized the complexity of IoT scenarios, right? You first wanted to get the data from that sensor and get something out of it. So the first reflex has been to say, “I’m going to connect that device, put all the data in the cloud from these sensors and do my very challenging analytics with my infinite resources that have in the cloud.” But then suddenly you’re saying, “Wait a minute, this device is connected through a wireless connection. It’s on the moving object and I’m going to lose connection. I need to locally store that data and then eventually delay the upload to a cloud.” And then came a second level, which was, “Okay, this system is critical. Now that I know I have a way to extract insights and take action.” For example, detecting an anomaly is going to happen on my, let’s say, conveyor belt and because I have a vibration pattern that is starting to show up.

When you’re all connected with a very good data connection to the internet, you can send the data, have something analyzing that data, detecting this anomaly that has been detected and then indicate down on the field at the edge that something is going to fail in about a minute, an hour, a day and take action. But this infrastructure might not be that well connected, but the gravity of the problem is still there. And so the idea is to say, “How about we distribute that intelligence across the IoT solution?” It’s no longer about having dumb sensors and a very smart cloud, it becomes about, “How do I make a solution that distributes the intelligence?” Robotics is a good example of that. You want to distribute some of that intelligence so that if something fails, there’s still the rest that still works, right? You cut an arm on robots, all the rest still works, it still walks, and so on.

So you definitely want to have that distribution. That means you need to run analytics at the edge on these devices that you remember are optimized for a specific environment, that are optimized for running on battery sometimes and consume less. You need to be smart in the way you’re going to deploy, manage, update, and train that intelligence you’re putting on that edge. Edge computing is developing as a result of all the road bumps that IoT presents because you are not always connected and you want to distribute for many different reasons beyond connectivity, security, privacy. You want to distribute that intelligence.

Azure IoT Edge has been developed with that in mind. Once again, it’s a platform. It’s something that lets customers build their edge intelligence. IoT Edge allows to monitor and host a set of workloads. They’re going to develop and train in the cloud onto these gateways or, what we call now, edge devices. Bringing for the example of the conveyor belt, that intelligence that allows to detect the anomalies as close as possible to that object, to that sensor, generating that data so that you can stop it immediately or eventually interact with the operator locally without the need to be connected, without the need for your data to be stolen, without the latency you can have for a back and forth to the cloud.

In your report I noted that one in three projects apparently failed during this proof of concept stage and some of the typical reasons I noted that were these failures include implementation costs, a lack of defined expected benefits to the bottom line. When you look at all of these, the skills gap and you say all that, is the best advice you would give to somebody is to really have a good strategy in place for proof of concept? Is it to have the right money involved in here? What do you want to give?

I think people are now getting what they want to extract from IoT. They really want to transform their businesses. They want to harness and leverage that data that can be gathered. The first thing I would recommend, and a good friend of mine, Rob Tiffany, I think he came to your show a couple of times already, he has been in the IT space for a long time; likes to advocate for simplicity, explaining that IoT should be something simple. Get the data, extract the insight, and take action. Start with something that is basic. If you have a sensor, you gather the data, you detect a threshold, and you do something when that’s threshold is reached. That’s the very basic and most common scenario you can see in IoT.

The recommendation is to start simple. You know what you want to achieve and the example of the temperature sensor applies to many different similar examples. You want to detect systematically that something is going wrong or that you need to take an action to change the process or whatever. Make sure that this has value for you. Is automating that process something that you really need? If that’s the case, if the answer is clearly yes because it’s going to reduce my cost, it’s going to enhance the quality of my product, it’s going to allow me to have more customers. These questions that are easily answered by a simple solution to start with are the ones to aim at first. Start simple, start with the basic scenario that leverages the IoT devices and the analytics that are provided in these, out of the box services and solutions and build from there.

Build your skills as you go. Grow and don’t try and solve it all from the get go. Don’t try and develop everything yourself either. You’re not an expert in developing cloud platforms if you’re developing hardware in the consumer space, for example. So rely on existing platforms from proven companies that have been working at delivering that complexity in something that is easy to use. It is simple to implement it that integrates into your existing infrastructure. I think that’s the main advice: start simple, leverage what exists, make sure that it brings value, don’t do technology for the sake of technology. And one more thing that is dear to my heart: use the opportunity to that transformation to inject sustainability and resource suitability into what are you doing. IoT implementation is a great opportunity to do these things right in terms of saving energy, in terms of you’re not making things worse when it comes to the environment. So, I really recommend that you take that opportunity to start simple and have that sustainability in mind.