Jun 20, 2017 | 8 min read

Insight Vector: The Emerging IoT Architectural Paradigm, an interview with Randy Krotowski

MP_blogpost_header_Insight_vector_LinkedIn_v4.png

Innovation and market perspectives from leading IOT innovators

Our conversation with Randy Krotowski explored the growing importance of machine learning to drive business value in Connected Industry. 

Randy sees a new architectural paradigm developing where there is a new layer of cognitive technologies emerging above the application stack, where machine learning can drive incremental value creation across industry.  As part of our commitment to C3IoT, a Momenta client, Randy will be joining them full-time, and we are grateful for his contributions to Momenta Partners.

Randy, could you provide a bit of context on your perspective?

I am a chemical engineer by training; I spent 15 years in engineering, operations and marketing at Chevron, then I came into technology leading an SAP project before I became CIO for the Exploration and Production business at Chevron.  The experience in operational and information technologies was what shaped my views of the market.  After Chevron, I moved to Caterpillar as the Corporate CIO to help drive a digital transformation.  There I began to see IoT and machine learning technologies start coming together. In 2014 I retired to spend more time focused on IoT and machine learning.  I spent 18 months working with GE on the Oil and Gas aspects of Predix and most recently I’ve been working as an advisor at C3IoT for Tom Siebel.  In my experience, they have the best platform for looking at machine learning. I teamed up with Momenta Partners because of their heavy focus on the IoT space including connectivity and devices.  The insight I’ve gained is an architectural perspective on what is coming next. 

What do you see as the new architectural paradigm for IoT?

Think of IoT and machine learning as a new architectural wrapper for existing enterprise operations and IT.  If you think about the technology stack for data to support analytics in any industry, you have to deal with a heavy legacy IT footprint.  This leads to three challenges – a diverse set of operations, a diverse set of technologies and a diverse set of data.  IT has been trying to knit this stuff together for a long time.  15 years ago, we had a vision of closed loop analytics, but we are still a long way from realizing it.  The complex “spaghetti code” data diagrams that we saw back then have not changed a lot.

What has changed is connectivity and cloud.  It’s like a sandwich. At the bottom is connectivity - how to extract data from thousands of different devices and interact with them. On top of the application layer, I see a new layer of machine learning.  What I see with platforms like C3 is a true platform approach that takes data from any source and allows the application of machine technology to solve a broad range of business problems.

You still need the legacy systems that run your core business and they are not going to go away, but there is a huge opportunity to layer on new technologies. The ability to put in a new layer at the base of the stack (connectivity) and machine learning at the top of the stack will enable the optimization of existing value chains and creation of new value chains.  You can collect data from anywhere that has data from the entire value chain, then optimize and predict at a larger scale. 

What are the drivers and constraints of adoption?

I see the drivers of digital transformation as pretty straightforward: growing business value, reducing inventories, increasing efficiencies. As I see it, the constraints are twofold - one is making the right technology choices.  Most people are thinking about complex solutions for machine learning rather than finding the simplest ways to add value. The second challenge is finding the right kind of data scientists, which are in short supply.  The technology has evolved but there will still be a shortage of the right kind of people.

Back to machine learning – it really works when there is lots of data around what happens and what you want to happen.  Machine learning is about creating a behavioral-based model that predicts the actual behavior of a system rather than a physics-based model which is, at best, an approximation and unworkable in any sufficiently complex system. You need lots of data, and data has historically been trapped across the organization. This is why platforms are becoming more and more relevant.

What have you learned from the convergence of Operational and Information Technology?

Chevron had been doing digital oilfields for 5 years before I became CIO - they could describe what a digital oilfield was, but the technology was so fragmented it was analogous to building a new city.  People were thinking about a comprehensive solution but the economics were not practical.  When I became the Upstream CIO we had to change the philosophy, and go after discrete problems with a lot of value.  How could we solve a real problem in a way that solutions could be replicated despite operational, technology and data diversity. 

In 2010 the technology we have today wasn’t there.  At Chevron, we adoption agile development and service oriented architecture.  Ultimately, we spent several hundred million and got a 3X return.  It was a great investment.  But, scalability was a huge challenge.  The philosophy was find a better way to deal with diversity and get something of value for the enterprise. Then it was service orientation.  Today, a platform allows you to separate yourself from the complexity of the environment and scale. 

Who gets it and who doesn’t?

Around 95% of companies across all industries don’t have the right philosophy and mindset. Financial services companies understand the value of machine learning because these are data-oriented businesses.  Fraud detection and other predictive analytics are pretty successful.  Outside of financial services, it’s pretty spotty. Almost nobody really gets what’s possible yet and I can count on one hand the companies that are driving machine learning at scale. People are doing data science projects for now.  It’s like cloud computing where it took several years for people to find the right use cases.

What do you see as the most relevant platforms for IOT?

There are two types of platforms I see as relevant– artificial intelligence/machine learning and connectivity/device platforms that allow you to interact with the devices.  These types of platforms hide the complexity. Engie is a company that recently bought technology from C3IoT, and they are rolling out predictive fraud, predictive maintenance for transmission facilities and a suite of other applications.  I’m impressed with both their vision and with their ability to execute their vision.  There are health care companies that sees the value of a platform for patient data – for instance, can you predict the likelihood that an individual is likely to get a particular disease so that they can be treated with preventative measures?  Can you predict the efficacy of a particular medical practice and apply these predictions across different practices?

How do you see the role of partnerships in IOT?

The nature of partnerships will change. One of the reasons I joined Momenta was because I like the model.  How do you respond and react to the business opportunities while finding the right people to fill key roles?  It’s valuable to find the link between technology and business applications. 

Many traditional systems integrators are struggling because they have been “body shops” built around large ERP and system integration projects.  They are struggling with what their value proposition is in a rapid development, agile platform based world.  A large SI firm can build anything, but the vast amount of the IOT opportunity is building smaller apps.  C3 believes that if an application can’t be developed in 16 weeks with 6 people you’re thinking about it wrong. With IoT and machine learning you can start small and build to a grander vision.

What got you excited about platforms?

Platforms have advantages in speed to market and scalability.  Platforms simplify data by abstracting it from existing enterprise systems.  Then you can connect everything and figuring out how to make best use of the data.  Everything is orders of magnitude easier, simpler, more scalable

Are there any books you would recommend?

Data science is going to be significant so it’s critical to be educated.  There is a book called The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World by Pedro Dominges that can help get your mind around the potential for data to solve problems better than throwing 100 engineers at a problem.

 
Momenta IOT Growth Partners