Podcast #164 Industry Insight

TRANSCRIPT

Ken: This is Ken Forster. Welcome to our Digital Thread podcast produced by, for, and about digital industry leaders. In this series of conversations, we capture insights from the best and brightest minds in the digital industry- their executives, entrepreneurs, advisors, and other thought leaders. What they have in common is like our team at Momenta; they are deep industry operators. We hope you find these podcasts informative, and as always, we welcome your comments and suggestions. Good day, and welcome to episode 164 of our Momenta Digital Thread podcast series. Today, it's my great pleasure to welcome Andy Bane, CEO of Element. This Silicon Valley company enables industrial organizations to unite their operations data for analytical insights that drive cleaner, safer, healthier, more profitable operations. Previously, Andy served as EVP and Chief Strategy Officer for ABB Enterprise Software, EVP of Product Management and CMO at Ventyx, and VP of Operations and Marketing at Global Energy Decisions. Andy has overseen dozens of new product launches spanning enterprise applications, automation and control systems, and analytical applications for asset-intensive industries. Andy earned his BA at the University of Colorado, Boulder; he and his wife, Lisa, a veterinarian, enjoy biking, skiing, hiking, and camping with their children. Andy, welcome to our Digital Thread podcast.

[00:01:35]

Andy: Thanks for having me, Ken. I've much enjoyed it over the years and appreciate the opportunity to be here.

[00:01:40]

Ken: Yeah. We've done some work with you way back when, and it's high time we finally get together. I appreciate Mike Dolbec's recommendation to bring you in. It looks like it'll be an exciting conversation today.

As you know, I always like to start off talking about one's digital thread. In other words, one or more thematic threads that define their digital industry journey. What would you consider to be your digital thread?

[00:02:05]

Andy: Well, Ken, I have to start way back. I grew up in Boulder, Colorado, and had exposure to digital and computers and a lot of other techs from an early age. We had the University of Colorado. IBM had come to town in the 1960s. And for whatever reason, the US federal government had decided to put many scientific agencies in Boulder. The National Center for Atmospheric Research's headquarters is there; they had one of the first Cray supercomputers to do atmospheric climate modeling. There was NOAA and also at the time National Bureau of Standards, now NIST. And we had many parents who were very tuned into applying digital to scientific problems and put much pressure on the school district, which ended up spending a massive amount on mainframe systems and other gear.

As part of the math curriculum beginning in the seventh grade, we were required to learn how to program, which carried on through high school. Many of us also had PCs in our homes as soon as those were available. Dungeons and Dragons wasn't so much a board game. For us, it was a game on the Apple PC. I grew up in a middle-class family; I didn't realize how privileged we were to access technology and those resources until later in life. For me, digital was always just there. The thread was woven into life patterns growing up and certainly nowhere near as dominant as today with my kids who can't seem to put down their smartphones. Anyway, fast forward to the industry thread. After college, I'd been working in software and tech in the 90s. In particular, one company focused early on commercial applications of the internet, extending legacy systems, the internet, and working in some fairly complex distributed systems around payments, order processing, and some in manufacturing, connecting suppliers with OEM shop floors and things like that. And we were using some technologies that were probably ahead of their time, but we made it work and had much fun. At that time, I had exposure to some very deep thinkers who worked around descriptive psychology and artificial intelligence and classification systems. The idea of digitizing the physical world captivated me; how you might ontologically and semantically represent the real world digitally, using data and software. That passion eventually led to the multi-decade path I've been on as a software provider to asset-intensive industrial firms.

[00:04:21]

Ken: What a great heritage at the center of OT and IT. And indeed, the way you describe analytics, should be a fascinating discussion. I see in your bio that your early leadership at Global Energy Decisions had already put you on to a track record of success in industrial analytics. Maybe tell us a bit about what attracted you to that company, particularly in the space at that time?

[00:04:45]

Andy: Sure. The electric power space has fascinating problems using software and data because electricity is essential for modern life. It's a commodity that, once produced, must be consumed, which gets ever more complex with the growth and nodes on the grid. And while storage is ramping today, and it's essential, it's still insignificant compared to load requirements. Our thesis at Global Energy was that deregulating markets would require better analytical software integrated to ever more comprehensive data to run these advanced workloads.

Use cases where everything from short term decisions like load forecasting and generation portfolio planning and dispatch to trading and risk management, both physical and financial trading- to medium term use cases like fuel budgeting and risk planning to even long-term use cases like price forecasting, CapEx planning, things like that. But back in the day, we called the people using the software and data and doing the work. We call them quants. Today, we call them data scientists. And this was very much a thread through the power industry, these quants, these data scientists. They mainly used traditional approaches to simulation optimization, linear programming, mixed-integer linear programming, and other methods. And at the time, the only AI was using neural nets to do load forecasting, which took a horrendously long time to run, but it worked. There's a lot of data, which I loved in many different techniques to analyze the data.

[00:06:09]

Ken: We operate, in Momenta, across four sectors. Energy, manufacturing- what we generally call smart spaces, so think cities, buildings, farms, and then supply chain. It's interesting because one of our early theses was that the decentralization of energy caused a digital transformation pattern that, in some sense, was descriptive for yet those other three sectors. A lot of the deregulation you talked about and the technology changes in terms of decentralized generation. And, of course, load distribution with electrification was, in some sense, the predecessors for some of the same patterns we see in other industries. I imagine this will be an interesting conversation as we step forward in that regard because you were really at the front end of that. You must have done well because Ventyx acquired global Energy Decisions in 2007, which ABB subsequently acquired in 2010 for about a billion, which seems like a bargain now looking at 2021 pricing. But all in all, you invested over a decade into ABB across several executive leadership roles. If I can, what would you say were your top three insights relative to the digital industry from this long investment of time that you put into ABB?

[00:07:24]

Andy: Three insights? I'd say first that this idea of IT/OT convergence or what started as IT/OT integration is now more convergence will grow in importance. That was a critical insight, and it was going to grow in importance with more and better IT solutions that OT can leverage. But also, decades of automation control were reaching the point of diminishing returns for many owner-operators. And well beyond power, it was in oil and gas, pulp and paper manufacturing; it was regardless of sector. And industry wanted to find new ways to solve hard problems. We spent a lot of effort on IT/OT integration at ABB. I was in the enterprise software group so we could do things like giving the power distribution teams out in the field better information about the state of the grid; from the real-time systems to keep them out of harm's way and get the information from work; asset management system onto their mobile devices so they can be better prepared to do short and mid-cycle work or integrating things like load forecasting, DERMS, SCADA and other data sources. Some of the things that you just mentioned- and apps for improved forecasting operation. That integration of renewables into the grid was another one so that IT/OT convergence was going to be big was one insight. I'd say the second insight was the shift to cloud and connecting cloud and edge, and that cloud was going to be big, really big. Industrials would eventually figure out how to leverage the cloud as they work through concerns about security latency, that it was undoubtedly going to be a long journey.

Shortly after Joe Hogan left the CEO at ABB, I pitched the ABB executive committee to build the ABB industrial cloud. And I use many cartoons in the high-value transformer use case to make it super easy for them. But they couldn't wrap their heads around why anyone would want to use the cloud, let alone buy anything from ABB produced and made available on the cloud, and this was a decade ago. I mean, this has undoubtedly changed more recently. But the insight here is about how hard it is for big organizations to do first principle thinking that can result in innovation. I was a software guy looking to leverage the cloud to make our customers' lives easier, but I talked to many hardware guys. And then we think that the third insight and certainly the one that has played out for me in my life, that Element, is that data was increasingly central to everything. And if you could think about the data first and what it represented and how you could derive insights, this could make a massive difference for our customers, but the technologies just weren't there yet.

And about ten years ago, I had the ABB software product management leads for TND generation mining, oil, and gas put together plans for what I was calling at the time the physical graph. This was an industry view of equipment, key source systems, available data, and associated personas, which we sometimes call the digital twin. And it's a graph-y problem, meaning that you needed a way to efficiently build relationships across a lot of nasty data layers from OT and IT to represent this non-hierarchical industrial world. We wanted to deliver at ABB asset health solutions, so you needed to know how the transformers, batteries, and breakers were related to one another in the substation, how the substation was connected to the grid, all those things. We tried using some new big data technologies like MapReduce, which is powerful for certain unstructured data types, but really couldn't get the job done. You needed higher-level primitives above that. You can get that today from a graph model running on a graph database, which only really becomes possible to run it at scale with Elastic Cloud Compute. So that was the third insight, that data was going to come evermore to the fore.

[00:10:55]

Ken: I suspect that those three great insights all have a nice direct vector or bearing into your leading Element. You joined Element as CEO in 2015. Maybe tell us a bit of the origin story of the company. What problem did you set out to solve, and for whom?

[00:11:13]

Andy: Element's got an exciting genesis story. We were founded in 2015, out of Kleiner Perkins, the venture capital firm. Dave Mount, Element's board chairman, was a partner on Kleiner's Green Growth Fund at the time, an investor in OSI soft, and they founded Element. He recruited Samir Kolani, a data scientist, to run some experiments. And what they're trying to look at was how do you apply machine learning to industrial data? They recognized that OSI soft was going to sort of stay in their lane, and they wanted it; they couldn't find an investment in this area. And that experiment was fruitful enough to get a seed funding commitment from Kleiner, and then Dave recruited me as CEO to launch the company.

I joined in late 2015, and it was just Samir and I and a data scientist. He and I set about to find the market and began building the product. We spent the first year going after analytical use cases, like predicting the failure-critical equipment, the typical things. We did eight POCs that were data science projects with companies, most of whom you'd recognize, and we were building data management software in the background. We realized that the biggest problem was the data and that the biggest value we were bringing to the equation was in data engineering, not in data science. And so that's the problem we focused on the last five years. And today, we support the needs of industrial companies transforming their operations through data-driven insights by building and deploying hundreds, if not thousands, of analytics like predictive maintenance and others. And so today's mostly manual approach using Python code, spreadsheets, and things like that to tap into source systems tends to result in hundreds of point-to-point integrations that can't scale or be governed, which makes it hard to achieve analytical insights. We're going after the problem that disunited IT/OT metadata and the ability to use a graph-based approach to represent the physical world through metadata and then shed light on those relationships that are critical to analytical insights.

[00:13:10]

Ken: Tell us about some of your key use cases and, more importantly, wins.

[00:13:14]

Andy: We built our product Element Unify as it's always been cloud-native and focused, like I said, on uniting IT/OT metadata. And what it enables is industrials to get control of that metadata and give consumers more accessible access to context-rich data. And that is the primary use case- up-to-date views of operations and production, enabling them to build fast using no-code pipelines, the shift from deterministic to declarative programming, and taking advantage of more purpose-built data transformations that can address the OT problem. You can think of Unify as the glue layer that stores and represents relationships across source systems in a knowledge graph to enable that idea that- it's a shopworn term. Still, you can create that idea of a single source of truth to take single slices of truth to deliver whatever use case you need to provide or deliver data models to a data warehouse or a data lake, whatever it may be. We're serving customers in multiple sectors like chemicals, power, paper, packaging, food. I've got motions in pharma and other sectors, but I would say that's generally the approach- specifically wins, use cases. Let's talk about Evonik. Their global specialty is a chemical company, and they certainly stand out for us. Their headquarters is in Essen, Germany. They operate over 400 plants worldwide and have very strong digital aspirations.

Evonik, like other customers, wants to use their IT/OT data to go after multiple use cases. They want to do predictive maintenance on high consequence equipment like pumps, compressors, etc. They want to do root cause analysis, and where appropriate, they want to do OEE, and they want to singular an approach as possible across their fleet of plants. They use Unify to contextualize their time-series data, mainly IP 21, from Aspentech, with data from SAP from engineering design systems and other data. And because we get them off of spreadsheets and out of writing Python code, they're able to speed time to analytics and at a greater scale of data to support all of these use cases, and then some. For them, this has resulted in, by their estimates, $110,000 per year in value per analytic deployed, and they're deploying 1000s of analytics, so the value adds up fast. I'd also say that, yes, they add value because they built data pipelines, and you're storing the data in a graph via Unify. The cost to then govern and maintain their data to keep up with changes in the underlying source systems is a fraction of what they'd have to spend otherwise. It builds a lot more quality and a lot more trust in the data as well.

[00:15:52]

Ken: I certainly hope you contracted that with Evonik on a rev share basis or cost-saving basis, and that alone could fund the company pretty well. But congrats on already showing such value there. Let me ask because obviously, there are many companies that we've looked at and certainly many peers that you have in the industry that claims similar analytics capabilities. How do you know when an organization is ready to adopt your solution? And what best practices have you seen in realizing that potential value?

[00:16:25]

Andy: We are very much a middleware company. We enable the analytics; we don't deliver them. We're more of a brave new world start-up that is pre-category. I expect the category will be data mesh or data fabric. We're not a faster, better, cheaper category that customers already know how to budget for and how to buy. Companies like Seek and TrendMiner and others have done a great job, but they're more familiar from sort of a BI perspective than, say, what we're doing. And for us, typically, a company has experienced some failure trying to scale up these analytical use cases, or they have an urgent pain associated with projects that are stuck because they can't manage their IT/OT data. At companies that are further along in their transformation journey, their people have some skills, experience, and freedom to try software independently without a big procurement process and an enterprise sales motion. As such, during the past year, we put much effort into product-led growth to make it easy for an individual developer to free trial the software and then evangelize the results within their company. And it's early days, of course, but it's beginning to bear fruit. IT is being used to engage in this way; OT is coming up the learning curve on this approach. That's some of the shift in the market for us in terms of best practices. Since we deal with many OT data and industrial subject matter, experts will work with them to scale up processes to contextualize and govern their data. It's many data, and it often changes to keep that digital view in sync with the physical world. We automate much of that. It takes process and discipline because you're typically working across three environments for every plant. Development, staging, testing, and production- you have to get everything right and keep it aligned.

[00:17:59]

Ken: We've often heard IT analysts particularly bemoan the slow adoption times in the industrial IoT. Of course, there was a lot of hype at the very front end about how quickly OT would come up to speed as IT has done in the great virtualization we call the cloud. Where do you think we are relative in terms of industrial IoT adoption, close to what you expected in 2015 when you first joined Element?

[00:18:25]

Andy: That's a big question. And I know our time is limited; we could spend hours on that, but let me try to keep it short. Look, the way I think about this is generally, the rate of innovation has been slower than expected or hoped for at the macro level. Innovations are fundamentally social processes that then converge with technological processes. Unfortunately, I think the market has gotten this backward by regarding AIoT as primarily a technical implementation challenge that can be overcome by a few specialists, only to wake up and realize that senior management has to create the conditions for success across the business to get to some sustainable value creation. I think that stretched out the flat part of the S curve we're now in, but the inflection point is in sight. At the macro level, it's certainly easy to forecast the future diffusion of technology like AIoT. Still, it's very difficult to project the exact moment when that shift will occur at the micro-level, and it very much comes down to those human factors. I think industrials are beginning to recognize this. We certainly see it within our customer base where individual customers are starting to approach that inflection point or beginning to work through it, and events are compressing for them. Then the growth curve starts to unfold. We've got one customer who's building 70- what they call squads of between six to eight people on each team, or about 500 people who will be on these agile teams focused on data engineering and analytics development to build that foundation and other customers of ours are doing that as well. You look at those inflection points at individual companies. When combined with the cross impacts coming from things like supply chain fragility and decarbonization to cybersecurity, those slow-building driving forces come together. They will cause that sudden market change, and that inflection point will happen. I think we're close. I think industrials don't like change, but they're beginning to embrace it. And COVID certainly spurred that on.

[00:20:26]

Ken: Yeah. It has created a new normal. We often joke that all of our portfolio companies did well during the COVID downturn, and it's primarily because of the common use cases, remote asset management, right? If you can't get people out to the equipment, you somehow need to activate and monitor and maintain that equipment digitally. Anybody in the industrial IoT space probably did okay during that time, and indeed, coming out of it where the budgets are picking up. That and all of the dry powder sitting in the market from an investment perspective create some interesting valuation. You mentioned inflection point several times, and I'd be remiss if I didn't put you on the spot, maybe putting your prognosticator hat on for a moment- what would you predict for the next five years relative to industrial IoT? And again, as you said, knowing that we only have a few minutes left.

[00:21:21]

Andy: That's a big question. Look, I like LNS's research, industrial transformation chasm model, which builds off the classic crossing the chasm model. We're really in this pre-chasm market that's momentum focused on incubating and kick-starting efforts like industrial orgs. We're going to shift to this value-driven side as companies move through the chasm, but what they've done in the early part of the market will have to change fundamentally. I think that's where we are in that with that as a frame.

I believe that over the next five years, those organizations that move into and through the chasm and make it to the other side are going to achieve real impact because they're the ones who recognize that the technology comes last. They see that people in the process come first, and you need to design the sin. And so when you do go after the technology, the winners will have invested earlier in infrastructure like data management than do their industry peers and avoiding that trap of just going after the whiz-bang tools. And I think the winners will also be more adaptable when it comes to bringing IT and OT together to make sense. An early example of that fusion is OT security, and you see a lot of the ICS security systems vendors who are in there and starting to make real progress. What's happening is that IT and OT cybersecurity is being fused at many industrial companies in the SOC, in the security operation center. And many lessons are being learned around how you bring IT and OT together because the guys in the SOC are sitting there, getting a lot of false positives from the OT cybersecurity systems. They lack the context to understand that. When you have to solve those problems together, you see organizations begin to work more singularly.

[00:22:59]

Ken: That's an interesting one. We've invested in Zage, a hive company that you probably know well in the Bay Area there, and they sit right at that intersection of IT/OT from a cybersecurity perspective. We've heard similar convergence stories as well around, as you say, the SOC and indeed a field activation of it.

[00:23:16]

Andy: I think that cyber-informed engineering will become big in the next five years, and we're certainly getting pulled into this by our customers. There's going to be an expectation that it starts to come within the engineering designs.

[00:23:30]

Ken: Speaking of new normal, you've probably been a Silicon Valley company. I'm curious, how have the working and living patterns changed for Element and you since the pandemic?

[00:23:41]

Andy: Ken, I'd say that Silicon Valley is today more of a mindset, more of a concept. Element Analytics is headquartered in Silicon Valley, but we've always been a distributed workforce with a core team- mostly software engineering and DevOps in the Bay Area. But we've also got a team in Houston where most of our four deployed engineers are located. They're typically electrical, chemical, mechanical engineers, often with a Ph.D. in a similar discipline or control science. And they're excellent hackers too. And we also have a team of software engineers in Medellín, Colombia. We have folks scattered around and also in other parts of the US. 

We're a mash-up of IT and OT people. But our pride comes from the work we do helping people who work for industrial companies, the ones who make things work in our world and make modern life possible. We like to help them solve hard problems. And so, regarding the pandemic and work for us, COVID didn't accelerate digital adoption as it did for our customers. But it certainly had an impact on a lot of other things, both good and bad. I'd say it drove our team to think more deeply about meaning in our work and purpose in our lives, and it surfaced a lot of conversations within Element about how to achieve better balance. Many people have struggled to step away from their work with the line between work life and home life blurring. We've gone hard after that and looked much more deeply at employee engagement and building trust across our team. We talked more about the importance of mental health at our All Hands meetings, and we make sure people know what resources are available to them. We've instituted a relief day, the first Friday of every quarter, which is a mandatory day off because people just weren't taking vacations. We've got this incredible team at Element. We consistently exceed our 95% quarterly employee retention goal because we've worked hard on building our culture and still do, but now building the culture remotely. I'd say we're a work in progress and have a lot more listening and learning to do on this front.

[00:25:31]

Ken: A work in progress, perhaps, but wow. You guys have jumped quickly to reframe your culture and company again, as you mentioned, remote first. I'm impressed, and I like the retention rate. In closing, I'm curious, where do you find your inspiration?

[00:25:46]

Andy: Ken, I read a lot of history, biography, business, technical, a lot of science fiction. I like authors who stretch my thinking about the future but who themselves are curious and humble. I've been reading a lot lately on economic history, AI, and automation, thinking about where we are with technology and where we're going, and the importance of delivering labor-enabling technologies and not labor-destroying technology. I believe an important book that will stand the test of time is Carl Frey's "The Technology Trap." If you want to understand this moment in time, I think Brian Cantwell Smith's book "The Promise of Artificial Intelligence" has become one of my go-to's. I've read it three or four times, which is rare for me. Smith gets it right that we're far from incorporating human judgment into machine decision-making.

We risk closing up what progress we've made with AI for reckoning tasks if we don't factor in judgment, ethics, human consciousness, things like that. We've got to make sure that these technologies address problems that help people live safer and healthier lives. And so, I've been reading a lot about that. With COVID, I've found myself revisiting books that I haven't read for decades- books that formed my early views on the human condition, like CS Lewis' books on theology, Viktor Frankl and Carl Jung on psychology, and many others that are sort of timeless and provide anchors when things get turbulent. But mostly though, I'd say I gain inspiration from people who make that hero's journey every day who want to make things work, and mostly you don't hear about them because they don't have PR agents, and they don't spend their free time on LinkedIn and YouTube. These especially include Element's customers who are willing to take a risk to work with a start-up like us to go after hard problems and do things they've never done before, where there's no playbook. I'm also deeply inspired by our team at Element. I can't tell you how many flashes of inspiration I get every day working with our team. They're dedicated, brilliant, creative people; I'm lucky I get to work with them. And then last, I'm most inspired by my family. My four children, my wife, Lisa- they challenge me to think differently. They're people who actively seek justice for those in need, and they often whisper in my ear that all glory is fleeting.

[00:27:52]

Ken: Well, you have many, many different reasons to be inspired and to be the inspiration that you are and have been in this podcast. Andy, thank you for sharing this time and these inspirations with us today.

[00:28:04]

Andy: It's been a pleasure, Ken. I truly appreciate the opportunity and the work that you and the team at Momenta are doing.

[00:28:09]

Ken: Thank you so much. I appreciate that. This has been Andy Bane, CEO of Element and a deep practitioner in industrial insights. Thank you for listening, and please join us next week for the next episode of our Digital Thread podcast series. Thank you, and have a great day. You've been listening to the Momenta Digital Thread podcast series. We hope you've enjoyed the discussion, and as always, we welcome your comments and suggestions. Please check our website at momenta.one for archived versions of podcasts, as well as resources to help with your digital industry journey. Thank you for listening.

[The End]