If General Electric gets its current move into the industrial internet right, Jeffrey S. Bornstein thinks, the sky will be the limit.

Currently about a year into its transformation from an industrial giant at least partly focused on finance to a full-blown industrial internet company, GE is putting a lot of its eggs into Predix, a cloud-based operating system aimed at helping itself and its customers get the most money out of their assets.

And Bornstein, as the current CFO of the company and former CFO of GE Capital, has been at the center of that change.

Jeffrey S. Bornstein, CFO, General Electric Jeffrey S. Bornstein, CFO, General Electric

In 2013, after seeing the company’s finance arm through the financial crisis and beyond, Bornstein began working on the effort, which the company launched in 2011. With the company betting a fair chunk of its future on Predix — GE is investing $1 billion in it this year — the finance chief is spending about a fifth of his time on overseeing it, assessing it, and allocating capital to it.

Judging by the energy Bornstein seems to be putting into the initiative, he thinks it’s time well spent, bringing potentially big improvements to the company’s own operations and representing almost unlimited potential for revenue growth. If all GE does is to use Predix on its manufacturing systems and customer services and franchises, “we will generate a very, very attractive return on investment for our shareholders,” he told CFO in an in-depth interview in late June.

But f the operating system establishes itself as “one of the two or three operating systems for the world’s industrial internet, including markets and customers and companies that we don’t directly do business with, then the optionality on that is enormous,” he said.

“Then the returns get ridiculous,” Bornstein added. “There are infinite returns, essentially.” Edited excerpts of the interview follow.

What role have you played in the development of Predix?
I’ve spent an enormous amount of time around this. At the highest level, I’ve been involved in evaluating, from a capital allocation perspective, whether the industrial internet broadly, and Predix specifically, is an investment that makes sense for us. I’ve also been involved in evaluating what our expectations ought to be under a number of different scenarios of how that investment is going to pay back for shareholders over the medium and long term.

I’ve also been very involved in the detailed architecture of the technology stack that we’re building around Predix. It’s a very, very different than what most know as the consumer internet. To some degree I’ve even been involved in evaluating the suite of offerings of the micro services embedded in Predix that allow people to develop industrial solutions on it. And I’ve been deeply involved in all the strategy and operating reviews over the past three years as we’ve set this in motion.

I’ve also played a role on the M&A front as we’ve fit this puzzle together. We often find ourselves, in trying to fit in a capability or a technology, having to go through and understand whether an organic build, M&A, or partnership is the best approach, both from investment and speed-to-market perspectives.

The last piece is just the organizational structure. How we organize it, matrix it with our existing vertical businesses, what are the right incentive targets, and the right metrics to measure success. To the extent we’ve got slightly modified compensation schemes around a digital team, I’ve been involved in that. So I would say virtually every aspect of the effort.

What percentage of your time have you been devoting to Predix in recent months?
If you look at Predix both from an internal and an external perspective, I would say I’ve probably been spending upwards of 20% of my time. The two efforts are the predictivity application we’re actually deploying to our customers and what we call the “digital thread,” which is re-architecting our own company.

Do you envision GE becoming totally intertwined with Predix?
We absolutely believe that the industrial internet is going to be the next huge wave of productivity for the world, and we’re uniquely positioned for it. I think it will be very hard to unravel in the future what is Predix and what is GE. We are investing to win in this space. I don’t think this is going to play out that remarkably differently from how it’s played out on the consumer internet, in that in most spaces — mobile telephony, search, retail — you pick your place. The world tends to standardize on maybe three platforms to develop solutions on. I think you’ll have two, maybe three scalable platforms to develop industrial internet solutions on, and we want Predix to be one of those.

What precisely does Predix do?
In its simplest form, what Predix allows is the opportunity, through information and insights, to drive a very different outcome out of an asset or a system of assets. For example, a power plant can make more electricity with the same amount of fuel. Or make the same amount of electricity with less fuel. It’s in our core app, which we call APM, asset performance management.

It’s literally an application like you have on your computer. It’s about connecting assets with sensors, generating the data, and then doing everything to take that data and build insights. Then you figure out how you operate the asset differently, how long do you keep it in revenue [-producing] service longer, how to think about conditional maintenance around it, how to get more economic output from the asset.

What’s been your capital allocation process for Predix?
In most ways, it’s not necessarily materially different than the way you’d approach any other investment opportunity, whether it’s M&A, an organic investment, or a new product. You start by laying out a landscape of what you think you know the opportunities are and doing a lot of work to really try to understand the truths about those assumptions. And then you make an evaluation of what’s it take to win. If your conclusion from that work is that there is a meaningful opportunity, then you determine what the drivers of success are and how to build out the capability to actually deliver against that opportunity. How do you resource against it? What technologies are needed? What partnerships are needed?

It’s an iterative process. We’ve been at this really since 2011, and hundreds of times we’ve taken a step back and re-looked at whether our assumptions are still holding. Is what we thought to be true still true? Is the next dollar of investment moving us closer to realization, or is it going to be just money wasted? So we’ve probably done hundreds of pivots, hundreds of reevaluations since 2011 when we launched this.

Ultimately the way I underwrite it is: If all we ever do with Predix and its capabilities and the applications that get developed on top of it is to deploy it against our own company, our own operations, our own manufacturing systems, and our services and franchises to our customers, we will generate a very, very attractive return on investment for our shareholders.

If Predix actually becomes one of three operating systems for the world’s industrial internet, including markets and customers and companies that we don’t directly do business with, then the optionality on that is enormous. Then the returns get ridiculous —  there are infinite returns, essentially.

What’s been your biggest challenge in terms of capital allocation for Predix?
When we evaluate the next jet engine technology we have a very firm foundation to stand on. We know what we’ve done in the past. We know what our technology is. We understand the thermodynamic and aeronautic physics of jet engines. We deeply understand the market and our ability to forecast demand for aircraft. And so the analysis of evaluating an opportunity and our ability to execute against it is reasonably straightforward. That doesn’t mean we bat one-thousand, but we’re much more often right than wrong.

Predix is unique in that we first had to recognize the opportunity, which nobody really had articulated at any point in 2011, 2012, arguably even in 2013. You have the consumer internet and you’ve got the enterprise space occupied by SAP, Oracle, etc. None of those spaces addressed anything about driving productivity in assets and systems. Foundationally, you’re evaluating an opportunity for which there’s little to no history on. It’s a much bigger leap to envision what the art of the possible is. We had to prove it to ourselves, do some experiments to really understand deeply, starting with our own markets, if we could improve the output of our jet engine fleet by 1%, improve the uptime by 1%, improve the velocity on the railroads by 1%. And would 1% improvements mean anything? What we figured out pretty quickly is yeah, they mean tens of billions of dollars to our customers.

As CFO, how deeply do you go into learning about and assessing the nuts and bolts of the technology?
I go pretty deep. My finance team and the CFOs of our businesses are deeply, deeply operationally involved. In most cases they act not just as a CFO but the COO of their businesses. But our participation in the industrial internet is an opportunity to completely change the face of how industry operates. It completely changed the trajectory of what GE is. I don’t know how you would not be deeply, deeply involved.

What do you look at when you’re making decisions about technology?
What capabilities are perfectly fine, or must either be purchased or obtained through an outside partnership? How do we protect our intellectual property around the technology stack? What are the pieces of the technology stack and with what frequency are we going to update them? How does cybersecurity integrate into each layer of that stack? How do we differentiate what we do versus what others do in terms of the horsepower to compute?

We’re talking about exabytes of data, data that make the data that Facebook or Google would handle look very small. If any one of those things is not thought through as part of a stack or a system, it has the potential to be the weakest link in the chain and keep you from success. Every one of those areas is also a make-versus-buy, capital-allocation decision. What is it that we absolutely must own? What can we source? And what should we at least partially own through partnership?

What metrics do you look at?
Today I’m very much focused on how many people we have developing the platform and writing applications. Today we have 10,000, and we have a goal of 20,000 for the year. It’s no different than it is for the iOS on the Apple phone: If you didn’t have a million people writing apps for iOS on the Apple phone, there would be no Apple phone.

Beyond people, there needs to be outcomes, solutions that come out. We also have a goal of participating in 50 partnerships. We’ve got 31 today, including ones with Cisco, Intel, Oracle, Accenture, and E&Y. Those partnerships are critically important to success.

Beyond the core technology itself, the third question is, do we have the right skills and capabilities peoplewise to execute against it? We just hired the founders and the development team of Siri away from Apple to GE to work on Predix. We have had zero issues hiring the deepest and best technology people from the biggest technology companies of the world to make this vision a reality.