Below is a lightly edited transcript of my conversation with Norm Judah, who is Microsoft’s chief of technology for its services business. Norm has remarkable experience on both the customer side and the vendor side of enterprise technology, and those combined perspectives are uniquely valuable in today’s high paced and turbulent business world.
This interview took place in August 2018.
Norm Judah: Hi, Bob. Thanks for having me. It’s a great opportunity. Great to be here.
Bob Evans: Norm, thanks. And now so Norm, as CTO for Microsoft services, you have an intriguing and, may I say, an unusual role. How does that play out in your work with customers?
NJ: So I have really what I think is the best role in the company, which sits between engineering and customers, really trying to do two things.
One, land early engineering products with customers. But then also use the voice of the customer, coming back to engineering, about what could be done differently. And where these scenarios really make sense. Because most of these business scenarios cross multiple technologies.
So I actually, depending on who I’m talking to, have the voice of the other party that allows me to see both amazing new work that’s being done at a research and engineering – but equivalently, as we’ll talk later, some of these incredibly new scenarios that are coming out of businesses really doing different things.
BE: Yeah. So. You are able to, in some ways, be advocates for both the customer in particular, and, where appropriate, the capabilities that Microsoft can offer.
NJ: The way I put it is that the customer has the right to be right, and for engineering to actually understand that I think is important.
BE: Good. That sounds like a phrase that more people in the tech business ought to try to take a hold of.
NJ: A little bit of empathy helps.
BE: Yeah. And sometimes a lot. Those guys are doing things nobody has done before. So they’re really out beyond the bleeding edge on things.
You can stream this whole conversation on Cloud Wars Live.
Microsoft Services CTO Norm Judah on the “Business Model Revolution”
Norm, when we had lunch recently you mentioned how a lot of your work involves helping customers develop new business models, both to meet their rapidly changing needs, and also take advantage of some very powerful and truly transformative new digital technologies that just were not available until very recently. So what’s this business model revolution all about?
NJ: So what’s happening is there’s both a gateway of new technologies, but also new business thinking. And those two things coming together is almost like a double helix that allow us to see these really interesting changes. So let me give you a couple of examples.
We’ll start off with something relatively easy, which is with Telefonica, a phone company in Europe, but in this case, starting in Italy, where they’ve actually built a bot, whose name is AURA, and the goal of this bot is to really change their customer engagement—not only when you’re calling them up with a potential issue that you have in this service, but really their ability to customize products and services to you. And initially they’ll plan to deploy that in Italy, but then through more of the countries they’ve got in Europe, and really this is the start of artificial intelligence being able to synthesize across a vast amount of data and produce something that’s more valuable, and change more importantly the engagement of the provider of the service, in this case, Telefonica. And the consumer of this service, which is their customers, globally.
But then it gets more interesting. Then we’re starting to see some really interesting dramatic changes to models.
Let me tell you a story about Adecco. Adecco is a staffing company. They’re a global company based around the world that actually has full time employees that they place inside of their customers on a temporary basis but even on a long term basis to really provide detailed and technical staffing into particular roles.
Their business model, yesterday, was made up of their collection of staff they’ve got, full time employees, and the supply side. And then the demand side, which is the requests that they’ve got from their customers to fill roles. And the magic of their system is matching the supply and the demand. But as we spent time with them and looked at this, they also brought up this interesting new environment, which is the gig economy. Which is freelancers who actually are real experts who have been doing what they do, who don’t want to necessarily be connected to one company, but want to be able to freelance everywhere—but also want to be able to freelance, potentially, to provide you some insight for four hours that will accelerate a project, rather than having somebody for a month or a year.
And so Adecco has built a different business. This business is called Yoss. And that business is actually an AI-based system that matches freelancers—hundreds of thousands of freelancers—to the demand that has been generated from their customers. So not only do they actually supply them with their own staff, they can actually find you a world expert who can spend a day with you.
In order to do that, it was again a fundamentally different business model. But what made this work for them is the AI matching model. The matching model that looks for the patterns of consumption and the patterns of available expertise and brought those two things together.
BE: Those are great stories. And in each case I think it comes around to something that you mentioned in our last conversation: that these technologies just weren’t available until recently, but also they kick open the door to a whole new model that can be replicated by other companies in other ways.
We were laughing about that thing of how sometimes you have a tendency to say, well, that idea is obvious. Yeah, after somebody else does it, it becomes obvious.
NJ: And that’s that innovation step that I personally am super interested in: who are the companies who are going to take these steps to make these differences? I’ll give you an example of another company that has a strong engineering background. We talked about thyssenkrupp, the elevator company, when we met, which is a global provider, one of the largest in the world, of elevators and escalators. They move people, is the way they think about it. And we did a very early HoloLens application with them. That was around service to a broken elevator.
The idea is that you send a technician in, the technician puts on a HoloLens, they actually can look at what’s happening. In the virtual space they can see engineering diagrams. In fact, even the as built diagrams of this installation. But if they get stuck, they can actually go back, call in an expert remotely who can see what they see through the HoloLens, and aid and assist in the repair. And what they’re finding is that the time to resolve and the time to take a technician, send them out, dispatch them, have them arrive, has actually fallen dramatically with this idea of remote support.
But what’s interesting to them from a service perspective is they quickly realized they were limiting themselves if they were servicing only thyssenkrupp elevators. Why wouldn’t they service somebody else?
And so their service business has actually grown, where a large part of their business is actually third party elevators and not their own, because the differentiation is their ability to use technology to provide on-demand service to the elevator with an expert sitting remotely, a fundamental new business they never thought that they would do that. But it turns out to be an intended or unintended consequence of the work that they were doing.
BE: That’s wild. Because you had started talking about the rapidly changing and quite significantly different business models. And you’ve given us, with Telefonica an example of the engagement model with bots, and now here in the case of thyssenkrupp, a revenue model where before there was this huge pool of other vendors’ elevators that was unavailable to them. But now they’ve tapped into something that they’d never been able to get into before. Right?
NJ: And never really thought about it being a possibility—but they quickly realized that if they’re great at servicing their own elevators, why not somebody else’s?
BE: Yeah. Perfect.
Microsoft Services CTO Norm Judah on the “Reinvention of Manufacturing”
So with those – we move people, and this whole notion of this very sophisticated machinery, and with elevators, I’d like to dig in a little bit more with you on AI manufacturing. For years, we’ve heard how manufacturing sector is in decline, the jobs are disappearing, they’ll never come back. Yet, today, unmistakably, there’s a resurgence in manufacturing. And how is AI helping with this manufacturing renaissance?
NJ: So I think the real awakening of manufacturing, and the reinvention of manufacturing is I think where we are. There’s some wonderful stories here as well. An example would be a company called Bühler. They make food processing equipment, based out of a small village outside of Zurich. They process large quantities of the world’s food – rice and grains go through their machines. And a couple things they’ll end up doing that actually is super valuable, both because it’s actually producing more edible food, and less waste.
So the first thing is just sorting rice for example. How do you sort the rice at speed between a white grain of rice and the others? And so that turns out to be a fairly simple thing for them to do and to actually understand how to do that well. But more interestingly enough, one of the things that we’re seeing today is a toxic fungus growing sometimes on the rice. Not only do you have to sort it, but you have to identify a toxic grain. They do this using ultraviolet light being shone onto the rice, and then a camera using AI models to be able to identify a toxin on the rice or not.
What’s incredible to watch rice moving at speed being sorted is that their productivity is that much higher than their competitors. It’s in real percentage points where they’re higher than competitors. If you think it’s five or ten percent, that’s a lot of rice that they’re actually saving – because the rest would go to waste. And so in this case, they’ve fundamentally changed the notion, the manufacturing notion, of actually sorting the rice and packaging it up in their machines. But in order to do that, they have to run this artificial intelligence in the device. It can’t go to the cloud. It has to run at such a speed that it has to be really close to every single grain.
BE: Yeah. That reminded me of another example I’d seen there with Bayer, the pest control company. But I love that one comment from somebody involved with that story. They said we used to be – because of the limitations of our manual processes, we were focused on catching or identifying rodents. He said with this new capability, we’re able to move beyond that to making food safe, improving people’s lives and health.
So what you mentioned before about these technologies change not only the outcome but how businesses think about what business they’re in, what they can do, what value they offer. These examples you’re sharing really show entirely new approaches to how companies feel that what their value is to the outside world.
NJ: Yeah, agritech seems to be a place where we’re seeing this a lot. We talked about rice and growth. There’s another organization in Australia called The Yield. And what they’re trying to figure out is optimizing the yield of a square meter of pasture land. And because of IoT and the affordability of IoT, they’re starting to measure all sorts of elements in the actual soil. Moisture, PH, weather patterns around it, sunlight, rain, cover, and using various models, they’re actually starting to predict not only weather, but advice to the farmer to actually when they should plant, when they should irrigate, when they should actually feed, when they should actually harvest.
All of that is advice to the farmer. In many cases, you still have the human making the actual judgment. The farmer is still deciding, but they’ve got a system now saying it looks like the PH is a certain level, and it’s going to rain, and the sun’s going to shine, and therefore we recommend you take this action. But the judgment in the end is still left with that individual farmer, and the consequence of this is higher yield. High yield per square meter, which is what every farmer wants to get to. But it does take the IoT measures of being able to measure one square meter of land versus another and look at the differences, and maybe even fertilize them differently, or plant different seed there, based on those measures.
BE: More precision, more localized – an entirely different model there. These all I think roll together. They’re great examples of how we think sometimes well, AI is just around the corner, it’ll be coming soon. You’re very clearly saying no, it’s here right now, today.
Microsoft Services CTO Norm Judah on AI and the C-Suite
So in that context, how do you see AI reshaping the priorities, or the front line responsibilities of the C-suite, the CEO or the CFO, or the CMO?
NJ: So I think the place to start with this is – we used to say, last week, that every company was a software company. But I think that’s actually changing really quickly, where every company in one form or another is going to be an AI company, if they’re not today. Which means we have to have leadership from that C-suite to pick the right opportunities, to actually set a mindset of experimentation to actually form a hypothesis, and then evaluate against the hypothesis, because we’ve seen that you can enhance customer service, you can improve productivity, reduce manufacturing defects, yield on a square meter.
All of those things are unique to your business. And if the CEO is open to the possibility, and has the right culture in the organization, I think there are great scenarios that can come out of that. But it takes a learning company to do that. It takes an approach to the company that actually thinks about data-driven decisions and how artificial intelligence can help them be different. But in many cases, if they don’t do it now, somebody else is going to do it quickly before them.
We do see different kinds of systems being built, though. Systems that are built around systems of intelligence that are actually looking at information and making decisions and providing advice to a doctor on how to behave differently. A lot of work we’re seeing is also around recognition of objects, the movement of people, defect detection on a production line. And then more interestingly enough, some area around AI assisted professionals. AI providing advice to a farmer as a professional, but to a seller, to a doctor. Those are places where we’re seeing early successes that are happening that really are really making some of these cool changes and enabling them to really have businesses that are different.
But it takes leadership from that C-suite. It takes leadership from the CIO finding the right culture, the right organization, the right strategy, to tackle these kinds of problems.
BE: Norm, is this type of thing you’re talking about, in the realm of culture and change management, is that part of the purview of what you and your team do?
NJ: It absolutely is. When we look at AI, we’ve created a maturity model—only one of the categories of maturity is actually around technical competency to actually do you deeply understand the nature of the AI services and machine learning. And the others are really in this notion of cultural change, organization, and strategy. So, for example, things like: does the organization have the ability to do agile development? Does the organization make data-driven decisions? Can we fail fast?
Do you have the right reward environment to allow people to experiment with a hypothesis? Do you have the ability to bring in outsiders to assist, or do you want to do it all yourselves? Most companies, even when you go up to the C-suite, believe they have all the data that they need—and one of the interesting lessons that we learn is that’s not true. There’s more data out there you can actually acquire either commercially or otherwise to actually make a difference. And even if you look at our little agritech example, weather data is one of the most important factors.
Weather data and being able to predict the sort of ultimate output, but weather data also tells me how many barbecues I’m going to sell if I happen to be a hardware store. So there’s this openness to actually bringing in other information that will make your decisions that much better.
Microsoft Services CTO Norm Judah on AI, Human Judgment, and Bias
Norm, what you mentioned there at the end: based on the weather, how many barbecues will I sell? I wanted to ask—for decades it seems, or centuries, the business world has had to operate on something of a backward-looking model. We reconcile our financials after the month or quarter is closed, and based on future production based on past demand. And so on like that. How can AI help businesses in some way reverse that arrow of time, so they can move at the speed of their customers, and anticipate what those customers want and need, instead of this sort of lurching and backward and not always effective catch-up game?
NJ: Well, what’s interesting about AI is it allows you to navigate across vast amounts of data, and in many cases will come up with a recommendation that’s not intuitive, because if it was intuitive, humans could do it. But because there’s this tremendous amount of data, the systems can recognize different patterns that are there. So one of the public case studies we’ve talked about for example is with Microsoft in the treasury, where we’re trying to look at predictions of accounts receivable around the world in all our countries where we operate and all our subsidiaries.
And it turns out that in creating these AI models, in the treasury, the models are about six percent better than people about predicting what the future is going to look like. And that’s based on being able to recognize patterns that humans wouldn’t recognize otherwise. And that’s one of the wonderful things of these models, is being able to do that. However, I do want to come back to a small element, which is the element of bias, and the ability to recognize bias.
I’ll give you just a quick example of bias. There’s some wonderful work that we’ve just talked about in India, which is actually looking at the likelihood of someone in some way in India, a patient in India, getting some kind of cardiac arrest. And the models that we used previously, the survey that was used previously for that, was a survey developed out of a hospital in Framingham in Massachusetts that came up with a standard survey. But it turns out that that survey works for people from a Western environment, but when they applied the survey in India, they were only getting a 31 percent accuracy.
So they went in and actually looked at a different set of criteria, different set of parameters, and came up with a different model for India that actually says certain things. So for example, alcohol contributes less in India than inactiveness and somebody sitting on a sofa. And diabetes doesn’t contribute as much as your blood pressure. So they found a different set of criteria in India that now gives them a 67 percent prediction rate.
So what happens is the original model that came out of Framingham was biased to an American culture, to the American lifestyle. This Indian model is then optimized to India. And it’s a very simple example of bias in the data, and then expecting the North American model to work everywhere around the world. And so the ability of the AI models to navigate over the Indian context, and come up with predictions of Indian behavior, that starts to get to this core essence of the judgment of humans using AI to make predictions.
BE: Yeah. So in this case and I guess in many cases, it’s not an intentional infusion of bias, but the outcome is still the same, it’s much less effective than it could or should be.
NJ: And so the important point here, at least at this point in time where we are with AI, is that data, models, are built by humans, data is cataloged by humans, and therefore it will be inherently biased, but requires the human judgment. And so one of the consequences – going back to your earlier question about the C-suite – is we believe that every company, every organization in one way or another is going to have to write their own AI manifesto. They’re going to have to write what they believe about AI, how they’re going to use data, how they’re going to publish data, and make the consumers of their products and services aware of that.
And that’s going to become a gateway, I believe, to the success of AI, is the companies and individuals developing AI solutions, being very clear about the data and what they’re going to do with it.
BE: Norm, maybe that’s something that they weren’t expecting. You go in, tell them you want to do that, they say well hey, that’s your job, you’re helping us, you come up with – you’ve said no, you need to do this from your own unique perspective and mission inside your company.
NJ: Well, I think it’s actually a three legged stool in this case, which is the technology providers, whether that’s companies like us and others, could be through a partnership like partners in AI. Through the consumers of a technology, so the companies producing these solutions like we’ve talked about. But then there’s a third leg that’s important as well, which we think is government policy, which might vary by geography. And those three legs together is really going to define the way AI is going to be used country by country and area by area.
Some countries are moving fast, some countries are moving a little slower in establishing these policies. But I think it’s going to come with those together, and that each one of these parties has to come forward and do it. And we recently published a book called The Future Computed which is our statement of a principled approach to AI, and we think that other companies should do that. And by the way, it’s probably a board level decision. It’s probably going to be a decision that the board of each one of these companies is going to have to make. Because it may constrain what businesses they will do and how they will treat the data of their current customers.
BE: But without that, they’re going to be flailing, right, they’re just not going to know what’s the right direction to take.
NJ: Establishing that manifesto I think is going to be super important. And there’s some really hard questions that every company is going to have to answer about things like privacy and safety and transparency of the data.
You can stream this whole conversation with Microsoft Services CTO Norm Judah on Cloud Wars Live.
RECOMMENDED READING FROM CLOUD WARS:
The World’s Top 5 Cloud-Computing Suppliers: #1 Microsoft, #2 Amazon, #3 Salesforce, #4 SAP, #5 IBM
Amazon Versus Oracle: The Battle for Cloud Database Leadership
As Amazon Battles with Retailers, Microsoft Leads Them into the Cloud
Why Microsoft Is #1 in the Cloud: 10 Key Insights
SAP’s Stunning Transformation: Qualtrics Already “Crown Jewel of Company”
Watch Out, Microsoft and Amazon: Google Cloud CEO Thomas Kurian Plans To Be #1
The Coming Hybrid Wave: Where Do Microsoft, IBM and Amazon Stand? (Part 1 of 2)
Oracle, SAP and Workday Driving Red-Hot Cloud ERP Growth Into 2019