The company hoped to remake itself as a software powerhouse. Here's what went wrong.

By Ted Mann and Thomas Gryta 

This article is being republished as part of our daily reproduction of WSJ.com articles that also appeared in the U.S. print edition of The Wall Street Journal (July 18, 2020).

In early 2014, the High-Impact Innovation Team in General Electric Co.'s Connecticut headquarters received an urgent call: It was time to present one of their apps to Jeff Immelt. The chief executive had an expanding interest in tech, having promised an "industrial internet" that GE would help build for the world.

It was the job of the HIIT team to develop apps for use within the company -- such as an iPad app to connect lighting engineers with the lighting sales force, and a web-based application for employee performance reviews. The division's public relations staff pitched the team's innovations to the business press.

Employees jumped at the chance to show Mr. Immelt their hard work, reinforcing his vision of overhauling the conglomerate in the image of a software skunk works. But in 2014 there was a small problem: the app in question wasn't built yet. The team had digital design files -- mock-ups of the way the program might eventually look -- but nothing running on a real machine to show the longtime CEO.

Supervisors waved away these concerns. One employee was told it was time to "fake and bake." Designers in Fairfield went to work animating flat visual designs, inserting enough movement to give the appearance, for the duration of one PowerPoint presentation for the boss at least, of an app that worked.

Word came back the next day. Mr. Immelt loved it.

The CEO wanted GE to be known around the world for more than just jet engines and MRI machines, power plants and nuclear reactors. Deep costs, employee confusion and delusions about what it would take to reposition the company for a digital future would make it more difficult to turn Mr. Immelt's vision into a profitable business, according to former board directors, GE executives and employees.

A representative for Mr. Immelt said funding for the digital business came from cost cutting elsewhere in the company and Mr. Immelt knew building a digital business would be difficult. In response to Mr. Immelt being shown a simulated app, the representative said, "Jeff Immelt also doesn't know how to design a turbine blade, but he helped build one hell of a jet engine business."

Mr. Immelt stepped down as CEO in 2017, as GE's stock began a multiyear decline from which it has yet to recover. The board has since fired Mr. Immelt's replacement in favor of a new CEO, replaced executives and company directors, and jettisoned portions of its industrial portfolio to focus on reviving its core businesses, including power turbines and aircraft engines.

A GE spokeswoman, when asked to comment about this account of GE's digital ambitions, said "we are focused on our future -- strengthening our businesses, serving our customers, and driving long-term value for our employees and shareholders."

A Massive Experiment

To Mr. Immelt, the future of industrial companies was in software and hard-core computing. Even now, that vision is widely considered to be correct and other industrial company leaders, including those who laughed at GE initially, are increasingly following the same path.

As technological innovation shrank the size and cost of digital sensors, they would be embedded in more machinery of every conceivable type. Huge streams of data would be harvested, not just from smartphones and WiFi-enabled thermostats, but from the massive machines at the center of the developed economies around the world. Data could be captured from gas-fired power turbines, jet engines on passenger aircraft, MRI machines and sonograms, the rumbling diesel-electric locomotives hauling freight across the American plains or down the core of the Indian subcontinent from Delhi to Bangalore.

Mr. Immelt said GE would know best which data to gather and how to interpret their meaning because it knew the machines. Its staff designed and built and repaired them for decades. They knew how they worked, and when and why they failed. Engineers empowered with the right digital tools could begin to learn new things from their own machines, the CEO said. "Big data" would help GE better predict when critical parts in engines and turbines were likely to wear out and break. These applications would run on a GE-developed software platform called Predix, which would reside in a new division: GE Digital.

In meetings of the GE board of directors, these grand plans received a blessing of silence, according to people close to the decisions. The board, made up of current and retired business executives and academics, liked Mr. Immelt and didn't like to challenge him. They appreciated his gargantuan work ethic, his optimistic spirit, and his eagerness to articulate far-reaching visions about the company's strategy.

A representative for Mr. Immelt, who also served as GE's chairman, said the board was regularly briefed about the digital effort.

The board tacitly blessed Mr. Immelt's digital dreams and his plans to stand up a software company inside General Electric, but never decided -- much less voted on -- a critical question about such a massive experiment: How much money were they going to spend on this?

The Selling of GE Digital

The marketing of GE Digital was just as important as turning it into a viable business, if not more so.

That job fell to Beth Comstock, who had a taste for black leather biker jackets and a gift for lofty rhetoric that belied her quiet manner.

Ms. Comstock had spent her career in media relations, rising through NBC to the corporate parent, when Mr. Immelt tapped her to be GE's chief marketing officer in 2003. The company hadn't had such a position for two decades. She rose to become a vice chair of the company, molding her own brand as a change maker who was rethinking the American icon. The digital operation was no different.

A series of broadcast television spots introduced a young computer programmer, the titular Owen, who informed increasingly perplexed groups of his friends and family that he had taken a software coding job -- at General Electric. The Owen ads spoke to some of the truth of GE Digital. Notwithstanding some claims that GE was making to investors about its success in luring young programmers away from jobs at Facebook or Apple, in fact, outside recruiters said, the response in Silicon Valley had been not unlike Owen's fictional brunch table.

In time, "digital" suffused the company's vast marketing operation, the adjective having swollen to crowd out all others in everything from Mr. Immelt's public speeches to Ms. Comstock's PR work. The communications office revised the boilerplate attached to every press release, the corporation's description of itself, to pronounce GE the world's first "digital-industrial" company.

Ms. Comstock declined to comment.

Mr. Immelt proclaimed in 2015 that GE would be a "top 10 software company" by 2020, with $4 billion in annual revenue from its new Predix software alone. GE Digital reported revenue of $3.6 billion in 2016, largely from operations in its industrial units, and targeted $15 billion by 2020. But the pledge to turn General Electric into a major software company was slightly more challenging than Mr. Immelt had made it sound.

Some executives at the company thought the obvious answer was to form a partnership. There were already massive, global companies that for decades had been building a software infrastructure to host and influence virtually every aspect of human life, communication, and business. Google. Oracle. Microsoft. Amazon. GE's knowledge of the machines it made and the needs of its customers was unmatched, but that didn't make it the most obvious candidate to construct the scaffolding of code on which the "industrial internet" would rest.

GE first launched Predix in 2013, calling it a "first-of-its-kind industrial strength platform that provides a standard and secure way to connect machines, industrial big data and people."

It was perhaps telling that as GE rolled out a product it claimed would change the way industrial machinery and major economic sectors operated, the company's executives couldn't even agree on how to pronounce it. Mr. Immelt extolled the potential of a product he pronounced as "Pree-dix," sometimes then turning over the microphone directly to an executive whose version sounded more like the verb "predicts." GE Digital was on track to spend $5 billion by 2016, according to people close to the operation -- a massive sum even by GE standards, equivalent to about half the R&D budget for a new, clean-sheet jet engine.

Saluting the Flag

GE didn't just pour money into Predix -- it smothered the project with cash. But without a coherent strategy and well-thought-out processes, the product development path was a wasteful one, according to former executives at GE Digital and corporate headquarters. GE's plan to move fast, produce a viable product, and then perfect it in the field got bogged down partly because of the size of the effort. GE hired armies of new employees and gave them all the resources they wanted to build its vision. It was like an auto company building an assembly plant, hiring workers, and leaving them standing on the production line, waiting for the vehicle to be designed.

Instead of charging a small team with developing the best product and then letting the operation grow with the product's evolution, GE set up a huge organization that wasn't quite needed yet. Development was often paused or delayed to start the process over entirely or just to stabilize the systems. Leading executives said the digital operation was designing custom software for individual customers -- inverting the usual industry model of engineering a single piece of software that can then be sold and resold, recouping the cost of its design and reaping profits.

Inside GE, there was pressure from corporate management to use the software and to show results. While some divisions were hesitant to commit to adopting a software program that needed a lot of work, others were quietly going in a different direction to develop their own software tools. Mr. Immelt pushed back on those business leaders, exhorting them to stop complaining. As one executive said, "There was a lot of talk of saluting the Predix flag."

Naturally the marketing and sales teams were concerned that potential customers would see that GE wasn't even using its own software. A representative for Mr. Immelt said he "felt GE could only be successful by reducing duplication and building scale."

Squadrons of salespeople had been deployed to sell the vision, but what customers wanted to see was the proof behind the concept. However, there was little to show. Once again, the marketing was ahead of the product. In fact, the sales teams weren't even entirely confident about what their product could do, according to former executives. Instead of hawking GE's lineup of products and services for a given market they knew, they pitched customers on a deep analytic software platform that was hard to understand and harder to explain.

No Turning Back

As the problems deepened, it appeared to executives that no turning back would be allowed. In a meeting of GE senior executives, one questioned whether it even made sense to move forward with Predix. Mr. Immelt fumed at the suggestion of abandoning the work and quickly nixed the proposition, making it clear that the direction wasn't up for debate and the marching orders were still the same: go build it.

GE wanted the system to work with everything, so customers around the world could just connect to GE equipment and go. Engineers were concerned, however, that even though Mr. Immelt and upper management were sold on the Predix idea, management's lack of understanding of software development would hold back its growth. GE also planned to build its own data centers.

The idea was to have a GE-owned and -operated cloud for its customers' data, but building such an operation from the ground up would be hopelessly slow to achieve and wildly expensive. Besides, companies like Amazon and Microsoft were already pouring billions into providing just such services to other businesses. Why would GE, arriving late to the business, try to duplicate their offerings?

There was also the simple problem of trying to put a lot of things on the same platform. Engineers found that the tiny sensors in GE's machines produced tons of data, but since they used distinct coding on systems spread throughout GE's global businesses, putting everything on the same platform made the functioning of the apps excruciatingly slow.

At the same time, GE's venture capital operation bought stakes in companies that developed various tools that could be used with Predix. The deals added new capabilities to the division, but they also made the conglomeration of Predix's code even messier. The result was software with plenty of bugs, a difficult user interface, and none of the requested features.

Finding itself unable to compete, GE would eventually shift its strategy. After Mr. Immelt left GE in 2017, new CEO John Flannery quickly declared that GE Digital would be more focused and only work in industries where GE was present. Mr. Flannery was fired in 2018 and his successor Larry Culp sold part of the business and named a new CEO to turn it around.

The company doesn't currently disclose financial results for the unit. Earlier this year, Mr. Culp said the digital business was "getting close to break-even."

--

Adapted from "Lights Out: Pride, Delusion, and the Fall of General Electric," written by Wall Street Journal reporters Thomas Gryta and Ted Mann. The book is to be published by Houghton Mifflin Harcourt on July 21. Copyright (c) by Thomas Gryta and Ted Mann

.

 

(END) Dow Jones Newswires

July 18, 2020 02:47 ET (06:47 GMT)

Copyright (c) 2020 Dow Jones & Company, Inc.
General Electric (NYSE:GE)
Historical Stock Chart
From Sep 2020 to Oct 2020 Click Here for more General Electric Charts.
General Electric (NYSE:GE)
Historical Stock Chart
From Oct 2019 to Oct 2020 Click Here for more General Electric Charts.