Podcast #17: Understanding Microsoft and the Internet of Things

Good day, this is Ed Maguire, Insights Partner at Momenta Partners, back again with another special guest for Edge Podcast series. Today we have Sam George who’s the director of Azure IoT, he’s responsible for Microsoft IoT Central, Azure IoT Suite, Azure IoT Edge, and many-many other projects. He’s a Microsoft veteran, he’s been there since ’97, he spent time in three engineering disciplines, has worked in DEFT development, and in 2011 he moved to program management leadership. Sam is a fixture and a thought leader in the IoT community, and I’m so thrilled that we’ve had a chance to get you on the podcast. 

Thank you, Ed. It’s great to join you and your listeners this morning, thanks for having me. 

First question is to get a sense of what has shaped you view of the Internet of Things, what are the key forces or experiences you’ve had which have really informed your view? 

I’d say the biggest thing that’s informed my view, and I think really Microsoft’s view, has been all the customers that we’ve met with, that helped shape our understanding of the needs in the space. We benefitted from this very early-on because there was a tremendous amount of interest in IoT very early-on, we had a lot of companies coming to Microsoft, or we were going to them, who were interested in IoT before it was supported broadly in Azure, or any of the other clouds. So, we heard a lot about what they thought about it, what they were expecting from it, what they were looking to benefit from.  

One of the things we did early-on is, we realized IoT can be a complex topic, and part of what our position was going to be in IoT was to simplify it to enable mass adoption. That’s something which really goes to Microsoft’s heritage, if you look back on our past, back in the day building client applications using C++, invented visual basic, sort of revolutionized client programming, and you can think of what we’ve done and what our focus has been on IoT, it’s doing the same thing for IoT. When we step back and think about an IoT solution, and this is from hundreds of hours of customer conversations, and partner conversations; one of the things we realized, and it was important to realize early-on is, what is IoT, what constitutes an IoT solution, why would you do it? Why would you bother?  

In a lot of ways what we found is, what companies were looking for is the ability to sense their products, their assets in real-time, to be able to have that be just a digital process like the rest of their businesses were and completing and marrying the physical world to the digital world. There’s a whole bunch of things you can do once you do that, and when you zoom in and look at an IoT solution we see a very simple three-tiered structure; there’s things you’re monitoring, there’s insights which you gain from bothering to monitor them, and then there’s actions that you drive as a result. Every IoT solution which we’ve seen our customers and partners build sort of adheres to that. 

Then what we’ve done is, we’ve simply gone about it and said, how can we dramatically simplify it for connecting too and managing those things, as the extreme scales you find in IoT, how can you do that securely? How can you do that across a huge array of devices and assets which people are wanting to connect to? How do you make it so that insights are simple to gain? And then how do you make it so that you connect to any business process and drive action? So, the big thing that influenced us was all of those conversations very early-on with customers and partners. 

So, in many respects it was more of a poll conversation than a push, at least as I’ve observed Microsoft’s evolution in the space, you weren’t necessarily focused on the category, but I think what’s interesting of course about Microsoft is there’s a uniquely broad set of capabilities and technologies but connecting to operational technologies and making that bridge is a little bit of a different domain than traditional IT for instance 

Absolutely. 

How did you approach some of the challenges, or the unique aspects of working with industrial companies? 

That’s a great question, we took a very humble approach, we spent a lot of time learning the space and learning it well before we figured out exactly what to do. We have a history for example in industrial automation, we were selling Windows embedded into industrial automation companies for many, many years, and so we had a lot of those relationships that were already built up with companies like Schneider Electric, Siemens, and ABD, and some of the very large players in the industrial space, so we spent a bunch of time with them. We understood what the concerns of the OT teams were, what does it look like to build out a manufacturing plant, what are the concerns of the operators, what are the tensions that you get between IT teams, how we help bridge those. So, we spent a lot of time in a very humble posture asking a lot of questions and understanding. 

I also felt as we were doing this, which was right around when Saatchi came on as our CEO, and as you mentioned earlier in the podcast I’ve been at Microsoft for a long time, there’s such a great revolution that happened when Saatchi took over as CEO. It really helped us especially in the IoT space we really benefited from it, because instead of looking at it at IoT from the lens of what can customers do with what we already have, we looked at it through the lens of, what are the business problems and the technology problems that would support those, which we could do regardless of what we have? Because if we don’t have it, we’ll go build it. As an example, our posture on IoT as far as connecting to devices, from day one, was we’d connect to any device running any operating system, anywhere on the planet. It wasn’t a, ‘Hey, we’ll connect to only Windows devices, as an example. 

You also have a whole different range of partnerships and have built an ecosystem that is in many respects distinct from working with ISV’s for instance that build on your platform. 

That’s right. What that freed us up to do is to find what are all the problems that companies are going to look to us for? We heard a number of them, we heard problems around security, problems around needing to be processing extreme amounts of data, challenges around managing large, large, large fleets, challenges around wanting these offerings, wanting these solutions to run not just in the cloud but on premises, and in hybrid configurations. Anyway, we took all of that in that form, the genesis of our IoT strategy and the IoT offerings we’d been building out now for four years. So, it’s been an amazing journey, and what’s most exciting to us about the Internet of Things is, we are really getting to the point where companies can benefit from it really quickly today, from a technology point of view. 

The really interesting part I’m looking at is, in a lot of ways when you remove the technology barriers, what you’re left with in IoT is that a company is faced with their own digital transformation, and I know everybody talks about digital transformation, but it’s a real thing which I see every day with companies. So, imagine a company that has an asset which they service, and there’s a project going on to connect those assets, and to use machine learning to drive predictive maintenance, so that you know exactly when to service it. There’s a real human element to that transformation with people who typically would have said, ‘Hey, I know when to service these things’. I don’t know what the world is going to look like for software, it’s helping me know when to serve these things. Or, sales organizations that say, ‘I know how to self-service contracts, I’m not quite sure about uptime guarantees, subscriptions, and things like that’. So, those transformations and the viscosity that they cause are very real. 

You’d spoken about simplicity as a driving mission. IoT over the last several years in many respects was a concept or an idea that had seen some very optimistic expectations, which many vendors expected that we would see up into the right curve, where its tended to be much more of a steady adoption. But could you talk a bit about some of the initial problems you found some of your customers addressing, and how that mapped to your own focus, and how that’s evolved over time? 

First of all to talk about this, I think it’s important to lay out for a moment, when you think about a typical IoT solution, we’ve seen many-many thousands of these now, when you think about a typical IoT solution especially from a cloud point of view where you’re connecting to devices, you’re collecting data, you’re finding insights from that, you’re driving informed actions, there’s a very typical service architecture which we see again and again; we published and recently updated our reference architecture that explains all the different Azure services can work together, how on premises offerings can work together in partnership to form an IoT solution, with the typical benefits being things like, I know where my devices are, I know whether they’re operational, I know if they left a GO sensor or if they entered one. I know what their servicing window is, and what parts to bring when it’s time to service them, things like that. 

When you look at the service architecture, you typically see there’s a cloud gateway for connecting to and managing all of those devices, that’s what our Azure IoT Hub does. There’s data-storage like Azure Data Lake, Azure Storage, for collecting all of those volumes of telemetry, or even things like Azure Time Series Insights for dealing with the time series data that’s coming from devices. Then there’s analytics things, like being able to figure out predictive models for knowing when the optimal service window is; for that there’s offerings like Azure Machine Learning, and then business process integration, when I know something is happening on a device that shouldn’t be, or I need to adjust its servicing window, how do I get into the workflow of my business process integration, be it SAP or sales force, or dynamics, and drive informed action? 

If you think about all of the services I’ve just talked about, in a lot of ways if you’re in enterprise, we lovingly refer to those as our Legos. The reason why we call them our Legos is because from talking to an enterprise point of view you can’t lead with those services, you can’t simply say, ‘Hey look, here’s 50 different services, and if you assemble them according to our reference architecture, things will be awesome’. Instead what we found is businesses want to be able to evaluate that return on investment for IoT, and they want to be able to do it quickly. There’s a lot of companies out there vying for attention in the IoT space, and so if you’re a business you have to consider who are the top ones, and how do I evaluate them. Of course, most businesses they have their day job too, and so its about how can they do that quickly?  

Something we did very early-on is we introduced solution accelerators, and what the solution accelerators are is something where you can add your subscription, tell us which region you want it provisioned into, and we’ll provision all of those services and produce a working end-to-end IoT solution, and it takes about five minutes. We lay down 10 different services, you get an Azure website which includes all the dashboarding capabilities that you need, and all that which you need for an IoT solution. Then it takes about another five minutes, you can connect devices, and so the very same day you can have a proof of concept, that week you can have a pilot, you can get into production much faster. So, that’s helped a lot, it helps because then people could see and touch IoT right away, as opposed to hearing about the value but not being able to get to it. That was one part of our evolution.  

And for several years we’ve had it in market and that’s worked really well. Now what we’ve seen is even with that approach it still leaves an exercise to the reader of, when you decide it’s time to scale this up to a million devices, or 10 million devices, or 100’s of millions of devices, you typically need professional services to come in and do that, either a partner or our own, and things like that. We saw so much of the same solutions, rinse and repeat again, and again, that what that enabled us to do is take IoT solutions to the next level, and so last year is, we introduced another offering called Azure IoT Central, and that’s a SAS offering for IoT, it’s a software as a service offering, it requires no cloud solution development expertise, at all. So, you simply connect devices to it using our software development kits. You can completely customize it to any vertical you want, be it tracking vaccines, or making sure you know where all your trucks are for fleet management. This enables you to customize it, build monitoring logic for all the IoT devices, trigger business, process integration, all without having to write a single line of code in the cloud.  

So that sort of speaks to the evolution we’ve had, we started with really hyper-skill Legos, I don’t mean to insult them by calling them Legos! For example, our messaging infrastructure when we started, it was processing about 10 million messages every month, and today its over two trillion a day. So, they’re pretty significant Legos! But from a business point of view you really don’t care about that, you trust Microsoft is going to have the skill you need, and our job is to take care of that. You want to be able to find out whether you can benefit from IoT, what that looks like, and what the return on investment is. So, today when you go on provision on IoT central application for example, it provisions in 15 seconds, you can connect a device within a minute, and you can get to production within a day. So, that speaks to the simplification drive which we’ve got. 

It’s clear that driving force of Microsoft, that message really resonates with the market, because everyone of these industries which falls under this rubric of Internet of Things, they’ll have their own standards, whether you’re talking about building automation, or manufacturing, or transportation, or energy, these are quite heterogenous businesses. 

I wanted to ask you about the platform strategy, and some of the way you think about partnerships. When I think about Microsoft as a business platform, which is a business where third-parties develop their own IP that leverages really the underlying capabilities, or say the Lego blocks, or building blocks, Microsoft is the epitome, they are the biggest… 

We’re a platform and partner company, yeah. 

Absolutely, and IoT there’s quite a lot of confusion about IoT platforms, I’ve seen estimates of up to 500 platforms which seem quite redundant. Our view is there’s going to be some consolidation, and a lot of those capabilities get rolled up into applications. You’ve had some partnerships, you mentioned Schneider and I think of ABB, and Rockwell, and some others that come to mind, we’d love to get the perspective on your partner philosophy, and  how the emphasis you put on the technology platform works and is helping to accelerate some of the work your partners are doing too. 

That’s a great question. We are very much a platform and partner company, its very much in our DNA, and one of the reasons why IoT is such a natural for Microsoft to put such an emphasis on leading, is IoT is more of a technique than anything! It’s an approach, it’s an approach of being able to sense physical things in real time and take informed action from your insights on them. As a result, because it’s a technique it means its broadly applicable to nearly every market segment there is. As a result of that the amount of vertical specialization you see for IoT in each one of these market segments is extreme. What that really means is, we rely on partners to provide that vertical specialization, be it in industrial automation, be it in smart buildings, be it in healthcare, oil and gas, life sciences. It’s not in our DNA to provide that vertical specialization, it's in our DNA to build the hyperscale capabilities either in the cloud or on premises, which make these partners successful. 

So, that very much drives us, and it’s something I’m glad you asked about, because that’s part of our core philosophy in IoT, and it's an example with IoT Central, it’s a horizontal offering, meaning it doesn’t do any vertical specialization out of the box, partners provide that. And so even though we’re getting to the point where even companies that don’t have technical capabilities can use IoT Central, we still see a strong need for partners to provide that vertical specialization. 

From the standpoint of being able to allocate resources from the partner’s perspective, they’re able to allocate their innovation resources, there are indeed budgets, not on reinventing the wheel, but they’re able to leverage all the work you’re putting into scalability, data management, and all these other abilities to connect devices, and then build that business logic on top. I think that’s pretty remarkable. 

One more thing I might add to that, early-on what you saw is, initially when we were just getting started say four years ago, many of the partners started building their own IoT platform on top of Azure. What they’ve seen over time is that we tend to move pretty fast with all of these horizontal capabilities, and so what we’re seeing more and more is that the partners themselves are just simply using the Azure IoT platform now, as opposed to having to build a large layer on top. That has been a transformation which has been happening over the last several years. Eventually what they realize is, we’re moving pretty darn fast, and we’re able to meet all their needs oftentimes before they could build it up themselves. What we’re seeing is an equilibrium where they’re really starting to focus more and more on just the vertical specialization that they’re doing. There’s no reason for any of these partners to try to support a service that can scale two trillion messages a day! But, providing the machine learning insights tell you exactly when to service, for example an oil and gas pump, that’s hard stuff, and we see them filling in that need now. 

I was interested to follow-up on the AI/machine learning/cognitive offerings. I think as I recently saw, Microsoft has about 8,000 people who are working in research, and a lot of that is focussed on AI and machine learning, and as an accelerant to innovation the potential AI brings to the table to drive a new rate of innovative value creations, it’s pretty unprecedented. I’d love to get your perspective on how the advances in AI and machine learning, and we certainly even over the last couple of years with the development of more powerful and cheaper GPUs and FPGAs, how does it fit into your IoT strategy? Are there any areas where you see the technology having a really significant, or very near-term impact? 

Absolutely, and going back to that structure which I talked about, about the things, the insights, and actions; insight is an incredibly important part of it, in fact I’d say in a lot of ways insight is the most important part, the insights you can gain. No one gets a lot of value from simply connecting to and managing large fleets of devices, you get value out of the insights which that provides, and what we’re seeing increasingly is machine learning and its ease of use is absolutely driving the value on IoT. So that is why we have such a tremendous investment in terms of engineering resources across Microsoft, in building out very sophisticated AI and machine learning platforms. 

Let me give a couple of examples. Broadly speaking if you think about the state of the art right now, if you’re a company or you’re a partner that has data scientists, the offerings we have right now are very-very easy to use; we have a very sophisticated machine learning workbench now which enables you to build very sophisticated machinery models quickly, to be able to deploy those at high scale on the cloud. We’ve even made it now so that we can take those same machine running models and push them right down to IoT devices with IoT Edge, and we will talk about that in a moment. But, if you think of the state of the art and where a lot of innovation is happening, its at that base level of machine learning where you really need to be a data scientist to know what you’re doing.  

Speaking to our heritage and our passion about democratizing technology, and making it available for mass adoption, that’s really where we’ve been coming from when you look at what we’re doing with our cognitive services. Now, our cognitive services are the typical machine learning workflow is collecting data, labelling it, splitting off the data, applying different machine learning algorithms, finding the optimal algorithm which predicts what you’re after, and then testing it on the second-half of the data-set that you didn’t train it on.  

Cognitive services takes a really different approach. As an example, one of our cognitive services is the vision service, where I’m looking for certain objects in images. Instead of having to go through all those steps I’ve just listed, what you do instead is you simply upload a series of images to the servers, and you label them. As an example, we have a fun little example that we showed at our build developer conference in Scott Guthrie’s keynote, our Executive Vice President over cloud in the AI, and the demo we had was called ‘Scott not Scott’, we took a bunch of pictures of Scott and we labelled them, ‘These are Scott’, and then we took a bunch of pictures of the other person who was on stage and we labelled those, ‘Not Scott’. Then we uploaded all those images to the custom vision service which is a cognitive service, and we clicked the train button, and that produced an image classifier in about a minute. 

Then, we took that image classifier, so I could send new images to the cloud, and it would say whether it was Scott or not. But then we also showed taking that same image classifier, that machine learning model, running it right down on a device connected to a camera, and then disconnected that from the cloud, and it was still recognizing Scott, or not Scott. That really speaks to two things, one is our passion of democratizing these AI and machine learning capabilities, and what we’re doing with cognitive services, we’re making it so that anyone, any mere mortal can take advantage of sophisticated AI. The other thing is, we’re making it so that it cannot just run in the cloud, but it can also run right on devices as well. That really speaks to a broader trend which we’re seeing, that as IoT really gets going, and some of the more sophisticated in that scale customers we have, what we’re seeing is, there’s this natural equilibrium which starts to happen between things that are happening in the cloud, and things that are happening right down on the device. We refer to that as Edge computing. 

In fact, I was just about to go there because that’s one of the areas we’ve been focused on a lot. I’d love to get a bit of context on how you’re thinking about IoT Edge, and what the implications are for business logic, and how you design and create applications when you start to incorporate these technologies? 

The way we think about this, and again we always like to have a context in which all of these innovations are happening. When you think about it, there’s these waves of computing which have been happening over the last several years, cloud is well understood, being able to have near infinite compute resources anywhere on the plant. IoT is a way that’s been benefitting from cloud, and in fact in our opinion the reason why IoT has taken off is because of cloud. And just to add a little color to that, IoT has been around for a long time, it was called machine-to-machine, and people were doing many of these same techniques 10 to 15 years ago. What’s different about IoT now is I can simply rely on cloud and all the 50 different regions that Azure exists in the world, so if I want to have fleet management in the US, fleet management in China, and fleet management in Germany, I don’t have to worry about building out data centres there, and all the hassle that incurs; I can simply take the exact same application and provision it three different times. 

So, IoT has benefitted from cloud, and then we see the next wave that’s been happening, and in our opinion is it’s been happening for about the last year and a half, that we call Edge computing, and of course Edge computing has also been around for a long time, it was just called something different, it was called embedded development, so, people were already building logic on devices to do things. What’s different about Edge computing is, it takes advantages of those previous ways of computing of IoT in cloud, and in our opinion Edge computing is really enabled by consistency of being able to run something either in the cloud, or out in the physical world, or both, so it enables you to have workload portability. The example I just gave about taking cognitive services and being able to run it in the cloud or on device; the mechanism that we’re using for portability between those two is containers, and so containers for those that are familiar are really revolutionizing modern application development in the cloud, where you’re able to create very fine-grain microservices, and manage fleets of them, to be able to even develop them locally on your laptop and run hundreds of containers, and push those out to the cloud to have constant integration, constant deployment approaches, using containers. 

So, when we looked at our Edge computing approach, containers was a natural choice for how we do portability. What that means is, all these different Azure services that we’re now targeting to the Edge that now run in this product that we call IoT Edge, that can run on devices as small as Raspberry Pi, or as large as you need. It’s a container-based system and what we’ve done is, we’ve gone around and taught all these different Azure services, like Azure Machine Learning, Azure Stream Analytics, Azure Sequel, to take certain bits of functionality like the machine learning scoring algorithms, so the image classifiers, and export them to docker containers so that they can be pushed out remotely to potentially fleets of millions of devices, and yet still managed. 

So, that’s the next challenge that you run into with Edge computing. It’s easy to push workloads out to a million devices, managing, patching them, upgrading them, that’s where the real challenge comes in. So, we did that very intentionally only after we had very sophisticated device management capabilities in our IoT Hub, our Cloud Gateway. So, you can manage millions and millions of devices, and so now we’re able to not just manage those devices, but also these Edge workloads within them. 

That philosophy to me mirrors a bit of when Azure was first designed, this ability to move virtualised workloads from a non-premise to a cloud-based infrastructure, and then be able to have that hybrid model was something that Microsoft really pursued in a unique way, that certainly the pure cloud, or public cloud providers, or pure on-premise providers were not as focussed on, and this extension is really critical. 

Now, what is interesting about this, where you start to get the de-centralisation of application logic, but where you can still maintain some central management, ultimately  it’s the types of applications that you’re going to enable. I’d be interested to know, what type of applications do you think are most promising, or most exciting to you, and blockchain technologies are naturally decentralized; I’d love to get to your thoughts on how blockchain and Edge computing have potential synergies to enable some new types of applications? 

Let’s take an example of smart agriculture; smart agriculture environments typically have low network capabilities, high-latency kind of things, and that really lends itself, it motivates one of the needs for Edge computing. It also highlights in my opinion, one of the most wonderful parts about being a part of this IoT revolution, is that it can really have a profound impact on reducing environmental impact. So, as an example, in the smart agriculture space, with the world’s growing population we’re going to have a 50 percent increase in demand for food over the coming decades. We also have a reduction of land that’s available for farming, and there’s water challenges all over. 

Being able to take this technique of IoT and Edge computing and apply it to smart agriculture can have profound effects. So, let me talk about a real example, one of our partners Schneider Electric, we have a public case study documented on this that you can find online, in New Zealand them and their partners have built a solution which leverages both Azure IoT in the cloud, as well as Edge computing to dramatically reduce the amount of water that’s being applied whilst increasing crop yields, and decreasing the amount of pesticides that are being applied. The way they did that, there’s these large watering booms that are on mile-wide circular plots, they attached little Edge computing devices to that watering boom at different locations, they have down-facing cameras. 

They built some machine-learning models using Azure Machine Learning, to detect a variety of things right on the ground, as an example whether the crops were doing well, whether there was a pest infestation, whether they were about to water something like a ditch where the water was just going to run off. All of that logic is, they trained it in the cloud but then they pushed it out to the Edge, and then it runs as those sprayers are going along. Its collecting data about what each square meter of the big circular track looks like, what’s in it, whether there’s pest infestations, whether its too dry. They also have moisture meters in the ground to help determine whether to spray or not, and all that logic about whether to spray or not, how much pesticides might be needed in different places, is all being computed out on the Edge. In real time they’re deciding which parts of the field to water.  

After the watering operation happens, then they take a summary of that data that was collected, and then they upload that to the cloud. So, the data is still being sent to the cloud because you want to be able to retrain it, and refine your machine learning models, but it doesn’t have to be sent in real time to make the decision about whether to water that particular square meter or not. So, that’s a good example of one the benefits of IoT for humanity and the planet, but also how Edge computing and cloud computing really start to work together. So, in summary what you typically see is, cloud is useful for being able to train machine learning models over large amounts of data. Cloud is also very well suited for managing large fleets of these Edge devices and pushing workloads out to them. Then the Edge devices are great for processing data locally, especially in situations where you have low or high latency connections, or no connections to the cloud, but you still want to benefit from what’s happened, that training in the cloud. And then Edge sends batch data to the cloud at a lower frequency than you would need to, if it was just simple IoT. 

That’s an amazing case study, because it exhibits the potential of the real time localized processing at the Edge, the ability to incorporate machine learning, or update algorithms on a dynamic basis. Then being able to take advantage of that massive processing power for historical data, to tune algorithms and solve a problem in conjunction with the multiple tiers of processing application logic, which are tuned to the business problem. So that is pretty amazing. 

I wanted to follow up on a second part, about blockchain. What are you guys thinking about in terms of potential there, and I know Microsoft has done a lot of pilots and a lot of work in that area, it would be interesting to hear from your perspective, the relevance of technologies at least in these very early days. 

Blockchain as a secure tamper-proof yet distributed ledger, it really works well in scenarios, its related to IoT in that as your business starts tracking all of these assets and products, as you start benefitting from things like Edge computing, there’s many scenarios where either you need to be able to show as evidence that these devices have been serviced at a certain time like you said they would, in your up-time guarantee. Or, I’m partnering with other companies over a chain of command, where I’ve produced this, I ship it to you, you do something to it, you ship it on, and we want a chain of evidence. So, blockchain winds up being anytime there’s more than one business, you can think of it as a multi-business integration layer, especially where you want a verifiable chain of evidence about a process that’s happened. 

So, we see it as highly relevant, more to businesses that are taking advantage of these connected scenarios, as opposed to being fundamental to IoT itself. 

That aligns with my views as well. From your perspective, are there any industries or specific use cases that you see which are having real successes with unique approaches, or even unconventional uses of technology? 

We see different industries moving at different rates right now, for a variety of reasons, but one of the industries that’s moving pretty quickly right now is the area, smart buildings, there’s a lot of adoption in smart buildings that’s happening organically. We recently announced a set of new capabilities for spatial intelligence that we’re doing as a platform, that some companies like CBRE, and Steelcase, and some others are using to do things in buildings, like help tenants understand how much space are they actually using, ‘I’m leasing a million sq. ft. do I actually need all that? Or do I need more?’ And again, to benefit the planet and humanity, to be able to reduce the amount of energy that they’re using in these spaces. 

What’s fascinating about smart buildings is, that’s just the start, because once you’re able to figure out how much space you actually need, and know how your space is being used, and to reduce the amount of energy, it enables all sorts of new productivity experiences which ties in with our 365 offerings. As an example, a lot of companies have a lot of meeting rooms, and a lot of times those meetings rooms are booked but being able to know whether a meeting room is actually open or not, whether there’s people inside of it, not whose inside of it, but whether there’s anyone in it, can mean the difference between I can actually use my meeting space, and I can’t. 

So, there’s a lot of activity that’s happening in smart buildings that’s been wonderful to watch, and it’s something that we have a lot of investment going into. 

That’s a great point. I know you guys have been working with Steelcase for a while, and what’s so interesting to me is, to see this evolution from their perspective from being a provider of products, to basically a provider of productivity, software driven solutions that do incorporate some physical goods of course, but that is that realization, the digital transformation that you’re alluding to right at the beginning. 

Your case is a great example of, ‘Hey, we’re going to make this leap between what we’ve always done, and what we’re going to do, going forward’, they’re very much doing it. 

Looking forward into the crystal ball, with so much that’s amassed here, I’d love to get your sense of what you’re looking forward to over the next decade; how do you see the industry evolving in the future? 

We’re obviously going to continue on our mission around simplifying and making it easier and easier, and so, you’re going to see a lot of that from us. That’s going to mean more and more complete parts of the puzzle, as an example we recently launched a product called Azure Sphere, for just a moment of context, of the 20 billion or so devices that everyone’s projecting will be connected by 2020, about 15 billion of those are these tiny microprocessors. A lot of these microprocessors go into heating and air conditioning, they go into appliances, they go into medical devices and things like that. What we were seeing is security for these devices really hadn’t changed over the last 15-years, they’re fundamentally insecure. Whereas, more powerful operating systems, or typical processors, lots and lots of security techniques to deal with threats, so we put together this offering Azure Sphere, which is really a first of its kind for microprocessors, that has some interesting and very unique chip intellectual property that we licence to manufacturers, it has a micro Linux kernel which we maintain, stay behind, and service, and it also has a security service that monitors these devices. 

So, now if I’m doing A-TRACK controls and I want these to be connected directly to the cloud, so that I can provide predictive maintenance, or I can provide up-time guarantees and things like that, I don’t have to worry about is this thing going to get hacked, Microsoft stands behind it. We’re the one partner that will stand behind that. You’ll see more and more of those kinds of things where we’ll have worry-free security across a broad spectrum of offerings, of course always done in partnerships with chip manufacturers, the device value chain and all that. 

The other thing that you’ll see more and more from us is dramatically simplifying the process of finding insights. That applies especially to IoT, and so we’re going to get to the point where you’re able to find insights, like what we’re doing with cognitive services is really just the beginning of making it so that this technique of being able to find these breakthrough opportunities, to be able to find the sort of needle in the haystack as it were of the signal in your data, is democratized, and that any knowledge worker can do that. One of the things in a step towards that which we introduced last year, was a service called Azure time series data, and for context, by 2020 the world is going to be generating somewhere around 44 zettabytes of data, that is a lot of data! Around 4 zettabytes of that will be time series data from devices. What time series data is, repeated measurements that are happening at second or minute increments, like temperature of a device, humidity, engine RPMs and things like that. Finding insights from time series data used to be only available to the realm of data wranglers, like someone would have to run these long-running data wrangling jobs that would happen overnight, or over a week of historical data. 

So, we took some sophisticated capabilities that we had in Azure for monitoring all the Azure services worldwide. We introduced what we call time series insights, and what it lets you do is store petabytes, or exabytes of data into this highly optimized store, and be able to query across it in seconds. Not just query across it in terms of a sequel query, but query across it visually so that you can see visual patterns. What it means is, if I’m an average knowledge worker and I’ve learned how to use Excel, I can find the insights over petabytes of IoT data, simply. That’s a pretty powerful thing, so you’ll see us continuing to make it easier to find these insights quickly, and that anyone can. 

That’s really exciting and clearly coming back to the theme of simplicity around Microsoft, the company has come a long way. When I hear you talk about a micro Linux kernel, I know we’re in the Saatchi era as well. 

Yes, very much. 

It’s been a fantastic conversation Sam, and I’d just like to round up by asking one of my favorite questions, a book or resource recommendation that you can share with our listeners. 

For a little bit of context on this one, when we were starting IoT it was such a broad topic, and there were so many needs to do all at once whilst we were building the team, growing the business, and all of that. There was a certain book recommended to me which is, ‘The Hard Thing About Hard Things’, by Ben Horowitz, that’s a great book about how to deal with situations where there is no easy way to prioritize your way out of it. I found that book to be very insightful and very valuable in my particular journey. 

I appreciate the recommendation. I’ve not read that one so I’m going to have to cover that one soon. It’s been a great conversation. 

This is Ed Maguire, Insights Partner with Moment Partners, and we’ve been speaking with Sam George, director of Azure IoT at Microsoft. 

Sam, thank you very much for your time. 

Thank you so much Ed, take care.