Podcast #35: Why Investing in Industrial IoT is Different

Good day everyone and welcome to the Momenta Edge Podcast Series. This is Ed Maguire Insights Partner at Momenta Partners, and today our guest is Mike Dolbec who is Executive Managing director for GE Ventures. We’ve got a lot of ground to cover today, Mike’s got a really interesting background and has his hands in a lot of fascinating areas, so we’re looking forward to diving into that. First of all, Mike, thanks so much for joining us.

Thanks Ed. Thanks for having me, it’s great to be here.

I’d like to start off first with a bit of your background, could you provide a bit of color and context in terms of your experiences, and what has really shaped how you view connected industry, or IoT, whatever we want to call it today.

I’m a beneficiary of the very strong vision that GE’s had for the industrial internet. I joined GE and what became GE Ventures, five years ago, and of course there’s a strong history of a world that always works, to borrow a phrase from Uptake, could be if you used all the data wisely that was available, and attacked one of the most expensive problems for asset operators, which is downtime. They started with that vision, and then imagined all of the solutions and software that would have to be in place to pull that off, and I guess I was one of the people that convinced them that they could move even faster, if they opened themselves up to the innovation economy that’s based so heavily here in Silicon Valley, but also distributed around the world.

Notwithstanding that, there’s also the first 24-years of my investing career where I’ve been connecting stuff; I was involved in the internet wave of investments, and then mobility, and throughout that various waves of AI investment. Then I was fascinated by the potential for IoT, so when I had the chance to join GE, it was the chance to put it all together.

Your background is in finance, what let you to focus on tech and some of the growth of your technologies in Silicon Valley?

My academic background is in computer science, I think I got the first master’s degree in AI from Stanford, there’s two of us in the programme but because my name came first, I was first in line. So, there you go, I win! That was back before the last AI winter when expert systems were all the rage. I was working at Xerox PARC during my graduate degree finishing up at Stanford, and then I entered industry first in what was the very first laptop computer company, Grid Systems, back in 1981. I’ve been an engineer or a computer scientist, and then a business development person, then in sales, marketing, product management, pretty much everything but manufacturing if I look back at it, and I was always fascinated by how these companies in turn started to go…

Sounds like somebody’s having a good time there in the background!

Yeah! Interestingly, today is the day after IBM announced the acquisition of Red Hat, so it might have been a stray Red Hat shareholder!

Definitely, they’re happy, they received a great premium!

It’s quite interesting times, the background and the context though that brings us to where we are today is super-interesting. Along the way what were some of the formative experiences that really shaped your view of the world, having lived through the AI winter as it were, that certainly opened up a lot of people’s eyes, not only to what ultimately would be the potential of artificial intelligence before the winter, but also the phenomenon of technologies that tend to be over-sold, or over-hyped. Along the way are there some lessons or experiences that you would consider formative?

I think that’s a good point. Having been part of what caused the AI winter, I was in sales selling expert systems to the financial services industry, I distinctly remember I think the only time I’d been at the restaurant Windows on the World, on top of the former World Trade Centre, was when I was pitching a room full of insurance executives about expert systems, and how they could semi-automate increase the productivity in underwriting risks. I what you appreciate if you’ve gone through that is, definitely the hype cycle and how you can get ahead of yourself, and how you can create expectations that you can’t deliver on, and you really can’t afford to do that if you’re a Silicon Valley company, you can’t afford to semaphore in advance an entire industry to become suspicious of you. I think we all learned to be humbler after that. That was a great backlash against the promise of expert systems, and it took quite a long time before the tremendous amount of data that became available, and cheap compute power to make what we call deep learning, now possible. It’s been an idea that’s been around in research for quite a long time, but it only became powerful, practical and easy to demonstrate recently. So, I think don’t avoid the next AI winter, don’t overhype this, is something constantly in the back of my mind.

Frankly, journalists are at fault here because they tend to write stories that are sensational headlines, with a lot of maybe’s and could be, and might happen, and usually they confuse the words AI with some value proposition. If they wrote the story as software did this, or, software did that, it wouldn’t be as interesting and salacious, but it would probably be more truth. I worry about that; I worry about the tech press. There’s a community of people who have been through it before, Roger Schank is my favorite debunker, he goes after Elon Musk quite a bit. I think we need to be very careful and pull back to AI techniques, some of them are very powerful, some of them are very brutal, and it’s important to understand how to be effectively a good product manager; how to know when you have situations when you can apply these techniques, when is it worth it to exert that effort, and when will you make a return that exceeds the enterprise effort to put it out there in the first place? I think a lot of people are rushing blindly out to things, and they surprise themselves sometimes, it’s more difficult than they think.

Well, it is a bit of the bright shiny object syndrome. Certainly, technologists and entrepreneurs are susceptible to it, but of course investors don’t want to miss the next best thing, and journalists do play the role in hyping up certain technologies. I think what’s so interesting as you go back and look at the initial promise as you live through the first internet bubble, and of course there were all these crazy start-ups, IPOs that got funded just based on some idea to rethink the way that we deliver pet food etc. But many of these ideas were valid ideas that were 10 or 15 years ahead of their time, because of the technology. Just sticking with the AI theme, given you’ve got such an extensive history watching the technology evolve, are there some initial visions that led to over-selling in the beginning, that you now see being realized in full fruition? Conversely were there some expectations that you think may be impractical, and may be impractical forever in terms of thinking of the potential of AI, and whether we want to define more specifically machine-learning, and some of the other techniques?

[Break]

When you look back at some of your experiences with AI or machine learning as you hear it, could you talk about some of the initial visions that may have been disappointing, or difficult to realize early on, that are now being realized effectively, and some of the misperceptions that may persist today about AI. I think it’s certainly appropriate to tie in some of the fear, and certainly doubt that’s getting banded about, about artificial intelligence and what it could mean down the road for society, etc.

That sounds like a long conversation we could have over a drink or two, I’m not sure I’m entirely qualified to opine precisely about that. I can give you some personal opinions but, your mileage may vary as they say. I guess the question I’d like to answer if I interpret your question correctly, is AI works [inaudible 01:18], and as I see it, there are really two schools of thought in term and the formal field AI was coined in ’56, three years before I was born, thank you very much! The two schools grew up to be AI could automate some form of human thinking, and process, so AI could replace people sometimes for certain things was one school of thought. The other school of thought was more humanistic, AI is best when its deployed to augment people, not necessarily replace them, but to make them more productive, or enable them to do either things they had not been able to do before or make them much more productive.

I think in a gross way without having to pin myself down too specifically I think it definitely works best when it augments and enhances the capabilities that enterprise employees are capable of. I’m going to stay away from consumer for a second. We’re nowhere near understanding how intelligence works generally to replace people very well, except in various certain super-nero tasks, and I think there are a lot of problems with that anyway, but I prefer to think that we’ve been increasingly successful in thinking of ways to make the chains of decisions that enterprises have to work to make those more efficient, or more productive from smart forms of robotic process automation that handle basic workflows. The tough part about workflows is when things don’t go perfect, when they go according to plan the workflow that somebody wrote down, all works, step a goes to step b and so forth, it’s the exceptions that need to be handles that are always tricky and you can’t anticipate all possible situations.

I think we’re becoming increasingly adapted at providing wise alternatives when exceptions come up that are not easy to handle. On up to some of my companies help some of the largest companies on earth, some of my investments, make better decisions, more optimal decisions in a fixed amount of time given the enormous amount of data that they have access to. They basically let them sift through a larger amount of options and play the chess game out through and pick the best outcome from that set than they’d normally be able to do under their own power. I think those forms of augment human decision-making, augment human capabilities to achieve an outcome are working and seem to be picking up speed in some fields.

No question that application of AI in almost industry of course is really starting to have at least a lot of dramatic attention being paid to potential. I’d love to get your perspective, going back to your initial work in the technology industry, and what ultimately led you to focusing on investing, and if you could share some of the principles or lessons that you’ve learned along the way, and tell us a bit about the type of investing, the stage that you focused on.

Well I entered investing quite a long time ago, so I’m not sure how relevant this experience is for someone in this decade. So, again its situational, my own experience, I was in graduate school and I was doing research at a very exciting lab called Xerox PARC, I was part of the team that were developing the small talk language and user interface, many of those ideas wound up in the knack in Windows, in fact I remember the team from Apple visiting, and I was recruited by a manager that I worked for, that had left PARC to go to Microsoft. So, I have a lot of connections to the way the personal computer history played out.

Being in that environment was seductive frankly, because it was almost as if there was nothing that you couldn’t achieve as long as you could create a firm enough vision about which direction to go. I worked for Alan Kaye at PARC who was responsible, I think his PhD thesis became the foundation for laptop computers, and later the iPad, this amazing device, amazing to think about in the early eighties when computers weren’t portable, they were big bookshelves in an airconditioned room. But he could play it out and see the progression and see how it could become very useful on the hands of a normal consumer, especially a child; I think child psychology was really fascinating to him.

Being at PARC I watched this tremendous adiaphorous… probably the wrong word, because that has a negative connotation, there was a flowering of innovation, like Athens, and then people would leave and go found other companies, the ethernet world was spun out of PARC, the Mac as I said came out of PARC, all of the ideas about multiple overlapping windows. In some sense our work, the laptop, the clamshell, the way that your computer opens up, the screen from the keyboard was a patent from the first start-up I worked at after PARC, called Grid. All these ideas, people were just making them up on the fly in order to complete a product and bring it to market, and being in that environment watching the stack as it were, we transitioned from a world where IBM did everything, they did the hardware and everything up above it, to a Silicon Valley based industry where people figured out, ‘Okay, I am this layer in the stack, I stand on the shoulders of the people below me, and I build something and then I sell it to the people above me. Together we all create this valuable proposition, but we’re probably not doing it completely by ourselves, that was a fascinating transition.

From PARC I went through a series of start-ups as an engineer, then a product manager, but eventually I felt like I hit a glass ceiling, I was type-cast as an engineer, I apparently had no business judgement even though I had strong ideas. I had worked for Mayfield doing diligence, one of my fraternity brothers was a partner at Mayfield and I guess when I called to meet together with companies together with them, and they asked me what I thought afterwards, I must have said something right because they kept inviting me back.

Anyway, I had this idea that if I went to business school, I could homogenize myself, get my head stamped as a well-rounded person, not just an engineer. That was my strategy, it wasn’t to learn business, it was to be considered a different type of person with a broader perspective. And in that time in the mid-eighties it worked and during the time I was in business school, friends of mine had left their companies and started other companies, received backing from venture firms, and so I decided not to go into investment banking which I easily could have done from business school. I wanted to return desperately to Silicon Valley, and so I networked with my friends, these entrepreneurs, and to some extent their investors, and then I accidentally got a job at Kleiner Perkins as an associate.

I guess my name came up at somebody’s board meeting, then they forgot which business school I was at, and then they couldn’t find me and they gave up, then somebody told me about that and I called them up, they said, ‘Hey, yeah, when can you come to Pallow Alfa to meet us’, and I said, ‘About 15 minutes, I’m just down the street right now’, so that’s how it started.

That’s pretty fortuitous, and of course being right in Silicon Valley you have a lot of exposure to the culture, and a lot of the players, certainly being able to see the decision process of course is hugely valuable. Are there some common themes in your view as we bring this back to connected industry, whether there are unappreciated technological enablers, or a market that may be unaddressed? What to you is a profile for say an appealing investment in infrastructure, a broader application versus a vertical application? I realize this is fairly broad, but I do want to tie it back to your perspective on connected industry, and then tie that into some of the work that you’re doing right now.

That’s an interesting question Ed, because for the most part, my responsibility here at Ventures is much more on I’m definitely in charge of infrastructure investments, and I did and partner on vertical applications as I’ve got that experience. Remember I said I’d been an investor, professional investor during several distinct waves of innovation here in Silicon Valley; so, what I meant by internet was, first it was personal computing which sounds ancient, but it was special back then. Then did networking, so the rise of Cisco and 3Com, and all the other competitors there, so first end points then connect the end points in a large generally private network. Then over in the background the internet started to grow up, and then everybody needed to connect their things and take advantage of services that were there, so service space computing.

I’ve seen layer upon layer of sediment build up into what is now a much more mature bag of services available to anybody who can connect to the internet than before. So, it feels to me connected industry, if you want to call it that, we’ve all seen this movie before, if you go back and look at the way the data-networking world has evolved, you can start to see or anticipate some things that will play out all over again. First of all, I’ll have to qualify this and say I’m very active in the industrial internet world, the internet 40 world, and hardly at all active in the consumer-base side. My purview is skewered in that respect, but I think connecting things is super important, and we take it for granted, but it’s much less standardized in the industrial world, so that makes it awkward, hard, and slow to scale, but I think does also represent opportunities.

So, you can connect things, barely, but it doesn’t scale very well. One of the issues is management of devices, how do I know what’s connected to me? About 20-30 years ago the network equipment industry was forced by its customers to create a standardized way for devices to respond to the request, ‘Who are you?’ ‘How are you configured?’ and ‘What version or which software is in you?’ so that network management became possible. For reasons I can only guess that really has never happened in the industrial world so there’s no standard S & MP-like command that says, ‘What kind of Plc are you?’ or, ‘What kind of Edge gateway are you?’ and, ‘Where are the blue wires, what version of which firmware, or which car is in there?’ So, I don’t think we’ll get there very soon, but I always remind the customers that they may want to lean on their vendors a little bit.

Device management today is pretty crude, it’s like you’re lucky enough to know what stuff’s connected to your network, at least maybe out to the Edge gateways, and I think it can always get better, and the way that it evolved through evangelism in the data networking world is an important aspect. I’m not saying that’s the only place for innovation, that’s just one of those movies I’ve seen before that hasn’t quite happened yet in this world. There’s layer upon layer of security issues that happen, its worse than the industrial world because the devices, the end points weren’t ever designed with the various attackers in mind, it was always a very trusted environment. And so, they’re more than exposed, they’re hardly protected at all, and I think that’s always been an opportunity there so, a wide variety of companies that are trying to work through those issues of detection of intrusion that’s been the first wave of companies, ‘Okay, what do I do if I know I’ve been compromised? What recipe do I go through to safely control what I think might be a problem until I know what’s going on’, there’s some special issues associated with that because you just can’t isolate certain industrial equipment because it just gives up. That’s another desire if perhaps, in some cases if the equipment doesn’t get contacted very often it goes off track pretty quickly.

So, the standard procedure of isolating part of the network may not always be recommended, it may make things worse in some cases.

How horizontal can technologies like security get? You alluded to this aspect of industrial technologies that there really is no homogeneity of communications, protocols, and certainly if you look at the Internet of Things in many respects it’s really an internet of internets, or an internet of vertical networks, and as you look at certain technologies how do you go about thinking how horizontal or how applicable a certain set of infrastructure technologies which we’ll focus on for now, how horizontal can they get? Do you expect that what we call industrial IoT will continue to remain somewhat distinct in terms of the sets of technologies and protocols that get used across different industries?

That’s a very broad question, could we narrow that down, an example or something?

Sure, if you look at the technologies, and set of technologies for building automation of course, and process automation, or manufacturing, say discreet manufacturing, a facility where you want intelligence around the building, around all the conditions in the building, some level of physical perimeter security versus QA and control certainly within the facilities that will for instance be able to protect against cyber intrusions, but then also tie together for instance physical conditions with the actual performance of machinery, and physical assets and output. Those would seem to be different domains, the management of the physical facility, and the management of say manufacturing equipment, but in many regards, they tend to be viewed still as very much siloed or discreet opportunities. Are there technologies that are able to think more holistically as it were?

Okay, I wish I had the whiteboard, but remember I’m an investor so I look at this through a lens of if you were a company would this be a good space to participate in, could you make money? I’m going to rephrase your question to, if you think about a stack in the vertical sense, where in points and things to be connected are at the bottom, and then there’s multiple layers of protocols and connectivity built up from there, and you could establish a company at any layer, or maybe how much of the stack you could stand, because I agree with it, one of the analogies that you used, that the Industrial IoT or IoT in general is currently a market of verticals, there is no one IoT, its often very use case specific. As an investor I’m qualifying this, I would say it’s really hard for a small company to participate in the lower parts of the stack close to the protocols, networking gear, and so-forth, because it’s tough to scale your business in a situation where there are so many highly fragmented use cases.

It would be great if everybody used the same protocol in one sense, because if you were efficient at creating that you could scale a business. But because they don’t you often get into this tail-chasing exercise of more than half my revenue comes from professional services, and I can’t scale that like I can a product business. So, if I were to answer the question, as you rise up that stack towards… the whole idea usually behind connecting stuff is to get the data where its sensed to someplace else, but as you rise up that stack when is it homogenous enough that you could create a viable business, at least a small one and have a hope of scaling it beyond. I think it’s very challenging as I said, for a start up to be at those lower layers. Most of my investing has started in this sense of where in the stack have I invested? It starts after the data gets transmitted to some place that’s homogenous if you will, so we’re very active in enabling Edge computing, so the software stack that goes into an Edge gateway and enables collection of data and analytic analysis of the data, a local historian sometimes are up to doing some machine learning, and then transmitting the important results but not all the data further upstream, maybe to other Edge gateways or onto the cloud.

So, at some point in this vertical stack drawing that if you could see it, I’m drawing on a whiteboard, it becomes I think safer for a start-up to establish a business because by the time you get the data, there’s still a lot of problems, it’s not on a protocol and completely incoherent stuff, it’s just a matter of a lot of messy data that you have to organize. I think that’s where I’ve got several investments, data integration is the seminal problem of our world right now. I’m going to say something controversial, but I think the smart start-ups, let somebody else do the heavy lifting of commuting and moving the data to somewhere where you can start monetizing. It’s really dangerous and a bit of a trap if you are the first person that has to handle all the connectivity problems. So far it hasn’t been sustainable for really small businesses, I don’t think.

That’s a really interesting insight, and I think if you look at certainly the rise of the cloud service providers, obviously Amazon, Azure, and Google, they provide this piping and infrastructure which provides a running head start for someone with a really good idea to be able to realize, at least MDP or a prototype of an initial concept. That was a great insight Mike and I’d like to pull it forward to some of the work that you’re doing now, working with GE and being focused a lot on the industrial space, where do you see some of the most attractive opportunities for start-ups working in industrial IoT as it were? Whether its technology or in the verticals, would love to get your insights.

Here’s my speech about that, because I’ve been saying this a couple of times. First here’s a retrospective, I’ve been at GE for five years, there’s obviously a lot going on with GE, but there’s a super strong vision that has driven it all, and I think in hindsight I’ll make three points before I answer your question.

1. When we look back it’s clear that GE has really unchained the industrial world from its electro-mechanical mindset of the past.

Because they framed this very simple relate-to concept, the power of one percent improvement in operational efficiency, what would that be worth to a major asset-operating company, its worth millions of dollars if you kept the correct operational situation. That also I think inspired a lot of people to go, ‘Holy Cow, we could do this. We have these situations at hand, and we have the data’, or at least we could get it after we connect and move it some place. There were all these concepts that they made, in my mind demonstrated the value of them, things like digital twins which these days is another hyped term, it means almost everything you want, the concept of digital twins, remote diagnostics of predictive models detecting early warning signals. My favorite the augmented strategic decision-making decision using AI, these are game-changing concepts I think, about running operations and enterprises, and that’s the sort of wakeup call that digital inspired me to go make that future happen faster.

Back to your question, what companies do I see that excite me, got to be the most active – Edge computing investors, unless you stretch the definition quite a bit. Throughout my very long career at GE all five years of it, what really surprised me was, first they had this fantastic idea for Edge computing which they only later asked customers about, and I think what they found was, it was way more seductive and popular with customers than they expected. Customers almost effectively said, ‘Oh, is that my data and my computing in that box over there? You mean I don’t have to ship it off to the cloud, whatever that is, some place where I can’t control it, and I can keep this box in a locked closet, some place I can control? And if things really get out of hand, I can pull the plug? I like that, I want to start there, I don’t want to start with this… it’s really scary, I don’t know what the cloud is, I’m not sure I want to send my data off my premises to someplace else.

So, that really exhilarated our interest in flushing out what that idea was, and all the problems and challenges associated with it. How do you provision 100,000 things at once? You can’t go to the Apple store with 100,000 Edge gateways, and the genius bar isn’t going to help you. So, what’s the zero-touch provisioning like in a world like that? How do you manage hundreds of thousands, or a million devices? How do you know they’re all running the correct version; how do you update them once they’re out there? So, one of these issues that aren’t a big deal when you have one or two of something, issues of scale become things of interest.

So, I’m very active in the problems and challenges that Edge computing creates that could be solved, the investments we’ve got that come to mind and are relevant are, Iotium, and FogHorn, Resin.io which I think their new name is Balena, each of these are involved in the provisioning and running of analytics, and then the incremental updating of the firmware over time. There’s a steady wave of people realizing pushing more and more computing out of the cloud, out to where the data is so you don’t have to pay the cost to move the data to the cloud, and pay the cost of storing it there, and pay the compute cost to process it. So, there’s dramatic cost efficiencies once you’re talking about gigabytes and terabytes, and petabytes of data. There are quite a few AI silicon companies that are generally experienced in exploiting image interpretation, is something that I know my Edge computing companies and others are exploiting, so things that in order for it to be cost-effective or to happen fast enough really can’t wait until things… you can’t ship all the video to the cloud and ask it how it’s doing, then wait to come back. I guess the classic example people can relate to is, should the car swerve and avoid the person in the crosswalk? Do you have the round-trip time to get to the cloud, and then get back again before you have to make that decision?

Well there are similar situations in operation of industrial assets where its critical. Something really bad Kinetically could happen if you don’t identify things early and take control and take action. So, the use of AI silicon usually deep learning silicon out at the edge for inference, I think is an increasingly attractive opportunity for us. We have another investment called Ovatos which is an inspection as a service company, they use flying drones and swimming robots and drone with sticky feet, they call them wind turbines, all in the service of collecting imagery that can be later interpreted. Usually they’re looking for corrosion and is going to become an enemy of industrial stuff particularly when its outside. It basically boils down to we’re not looking for cats, we’re not recognizing faces, but we are recognizing corrosion as it progresses over time, and if it starts to increase rapidly in the same place, we can warn a customer and then create a maintenance plan for them. It saves a lot of money but also it saves lives; it turns out that some of these inspections are very dangerous for people to prosecute, and so it’s a safety issue more than anything else.

So, Edge computing, I guess the Venn diagram looks like Edge computing, machine learning interpretation, the other part of the Venn diagram that doesn’t really overlap is this over thing that’s fascinating me, it’s the human augmentation with machine learning and deep learning. There are a lot of global enterprises that have teams of people that make very high-class expensive decisions on a repetitive basis, they don’t have a year, they have a week, they have to figure out what’s the best way to avoid a really expensive problem, or to spend $100 million to achieve some outcome. And helping them be more productive turns out to be very lucrative if you can pull that off, one of my companies, Mona has done a tremendous job, mostly for the oil and gas business, because if you’re an oil and gas company you’re like a small company in terms of your logistics opportunities and decisions you have to make. You’ve got a lot of gear that’s really extensive to operate, it’s either out at sea or its floating, or it’s in the ground, you’re extracting petroleum and moving it someplace else, splitting it up, your monetizing it into different levels and then you have to ship that off to consumers elsewhere. So, it’s a lot of money at stake and a big travelling salesman problem to be optimized. But there’s a lot of efficiencies to be gained on the same accord.

Are there any concerns that you have looking forward? You’re in the business of being a realistic optimist of sorts, you have to gauge your risk profile with a level of conviction and a belief in ideas that will ultimately come to financial fruition, as well as have impact. What are some of the concerns you have with companies with start-ups in the space, and with the evolution of industrial IoT overall?

I think the biggest concern, the biggest thing I keep in mind is that entrepreneurs really need to understand that these businesses that they’re going to try and sell into don’t operate… the playbook that works in enterprise software, the companies make a rational business decision, that’s still true, but the priority that they place on things like safety over novelty is hard sometimes for certain entrepreneurs to get. So, it really stretches the valuation cycle out. I think the entrepreneur would say the sales cycle, but really what it stretches out is, look, I need to make sure this isn’t going to result in something my people will deeply regret and be responsible for. So, it requires a very conservative approach to analyzing what it does, how it works, how could it be deployed, how does it operate, sometimes how complex is it? Sometimes I don’t have a staff, or I don’t want to build an organization that has all the skills to operate something, I’d rather outsource this and buy this as a service.

So, I think the concern is that the enterprise playbook is not the industrial internet playbook all the time, some of the plays are work, and some of them are traps. So, having people who have grown up in that world, or at least have access to a peer group that understands how to succeed in that world is important, so you need to have the right expectations, you can’t run too fast. The other thing is, the investor group needs to be very patient, so if you expect that this is a hot company that’s going to exit in two years as an aqua hire, that might happen, but I don’t think that’s an investment strategy. You really have to build a company that market discovery and channel discovery is super-important, and that’s the other thing I look for is, people who understand that one of their most important early focusses, this would be the CEOs, they have to figure out what’s the laundry list of qualification questions I can ask my customer, so that in the end I don’t waste my time with people that just aren’t going to become a sale for me. So, I think my most successful investments have pretty elaborate ways of qualifying the customers, so they can eventually say, ‘No’, ‘No thank you’, ‘You’re not the right maturity for us yet, you either don’t have budget or you’re not serious about this, or you haven’t tried and failed, there’s just a plethora of different lessons that people have leant arrows in their back.

It’s some combination of that realizing, plus that experience. And I guess the final thing I’d say is somebody with domain expertise is pretty important, so not always present in Silicon Valley companies.

Particularly in the industrial world where a lot of the companies just aren’t geographically located nearby.

Yes, if you go over to the bar at the Rosewood across the street from me, on Thursday night when it’s absolutely packed with VCs and people that want to meet VCs, and you yell, ‘How many people have ever been in a factory before?’ If they can hear you, they’re not going to raise their hands, because hardly anybody has. How many people have ever seen somebody make something physical, rather than at a keyboard? In this part of the world we haven’t got the old fart who leaves the company and starts a small company to address a problem they were familiar with, that cycle hasn’t completed very often. It is starting to where you get some serial entrepreneurs, but until it does you get people with some intuition of what worked last time when they were successful on an adjacent domain, and often those lessons are pretty good rules of thumb, and you just have to learn when not to follow.

Those are great insights, it’s an important distinction and really fundamental to the ability for investors to look at the opportunities in the appropriate lens, and I really do appreciate your perspective. I think you’ve got a unique combination of insights and operational perspective too. It’s really refreshing to hear.

My partner Marianne and I like to joke that over the last five years we’ve received our PhD, or maybe our Post-Doc in the industrial software world, because it’s been quite an eye-opener.

And it’s one of the reasons why Momenta Partners is very focused on the space, we see a lot of opportunities. I do like to wind up every podcast with a question about a resource or a recommendation for a resource, are there any books or resources that you could share with the listeners which would be something you would recommend to a friend?

Yes, I would. I must confess I didn’t scan your podcast lists, so somebody else may have recommended this book, I really like a business book about AI for enterprise called, ‘Prediction Machines’, you’ve probably read about it.

I think I’ve heard of the book, absolutely.

The authors are AJ Agrawal and Avi Goldfarb, it’s written as if you have no expertise in AI, and I’ve introduced it to people who have a product management responsibility, I know how to be a product manager for the current set of software that I’m responsible for, but, ‘How do I be a data entrepreneur?’ I’ve got access to data, systems and machines, when should I use AI to make something better? There’s a system in this book, and it eventually gets to it near the end, they even make an analogy, they have something they call the AI canvas which is an illusion to the start-up canvas, helping people decompose the series of decision it make in enterprises, and I think they make the analogy that every decision is effectively a prediction. Some predictions are commodities, and some are valuable because they’re difficult to do, and they kind of guide you to prioritize, pick a prediction that’s doable but hard, it’s kind of the watch word there, because if you could be more productive there that would pay off a lot. If it was a commodity maybe you’re not going to get much leverage out of it.

I’ve seen a couple of my companies use this analogy to think through the way that they explain the value of their products to their customers. So, basically, they’re saying there’s a series of important predictions, either about failure, or in Mona’s case some big monetary decision that has to be made, it’s really a series of predictions. But here’s how you take a goal, like I have an economic outcome I want to achieve, how do you back-up from that to understand the decisions that need to be made, in order to achieve that goal. Then how do you break down some of those decisions and use machine learning, or other techniques, usually semi-automate, in some cases fully-automate the string of some of the predictions that lead up to this economic outcome. How do I make the way I currently achieve this outcome, better at a group level? I really enjoyed the book.

That sounds super-relevant, not just business and not just people in technology, but life overall, they’re able to apply that filter, and perspective.

It’s one of a few books I’ve seen that instead of just saying, ‘AI is wonderful you ought to learn about it, and figure out how to use it’, instead it actually gives you some practical ways to think about the problems that you might apply AI to, and help you sought which ones are probably better to prioritize than others, that’s what I found.

Mike, I appreciate all of your insights, it’s been super-helpful. I know it took us a little while to get both of us on the line together, with a couple of technical hiccups along the way, but it’s been absolutely fascinating.

Again, we’ve been speaking with Mike Dolbec who is Executive Managing Director at GE Ventures. This is Ed Maguire, the Insights Partner here at Momenta with another episode of our Edge Podcast. Mike, I want to thank you once again for taking the time.

Well thank you Ed, I really enjoy speaking with you, it was great.

[End]