The following case studies—which involve solving some of the world’s most complex data science problems—encompass a number of different use cases, including predictive maintenance, inventory optimization, fraud detection, process and yield optimization, and customer insights. Common to all these case studies are the strategic nature of the approach—in particular, the focus on specific, high-priority objectives to drive significant, measurable value—and the CEO-level mandate for change.
ENGIE: Company-Wide Digital Transformation
ENGIE, the integrated French energy company I’ve previously mentioned, is in many ways the archetype of a large enterprise embracing digital transformation. With more than 150,000 employees and operations spanning 70 countries, ENGIE reported €60.6 billion in revenue in 2018. ENGIE generates vast amounts of data from its 22 million IoT devices and hundreds of enterprise and operational systems. In 2016, ENGIE CEO Isabelle Kocher recognized two inextricable forces shaking the core of ENGIE’s industry: digital and energy transformations. In ENGIE’s own words, the revolution in the energy industry is being driven by “decarbonization, decentralization, and digitalization.”11 Kocher recognized that, in order to survive and thrive in this new energy world, ENGIE would have to undergo a fundamental digital transformation.
As we’ve discussed, successful digital transformation has to start at the top. At ENGIE, it starts with Kocher. She has painted the vision for ENGIE’s digital transformation and announced that €1.5 billion will be devoted to the company’s digital transformation from 2016 to 2019. She created ENGIE Digital, a hub for digital transformation efforts across the company. ENGIE Digital includes its Digital Factory—a Center of Excellence (CoE) where the company’s software developers, alongside its partners, incubate and roll out innovative IT tools across the organization. Finally, Kocher has appointed Yves Le Gélard as chief digital officer to oversee these efforts.12
ENGIE’s first step in this transformation was to identify high-value use cases, and to craft a roadmap that prioritizes its trajectory toward full digital transformation. ENGIE’s Digital Factory has created and prioritized a comprehensive project roadmap. Use cases span the company’s lines of businesses. Here are some examples:
• For its gas assets, ENGIE uses predictive analytics and AI algorithms to perform predictive maintenance on its assets and optimize electricity generation—identifying drivers of efficiency loss, reducing asset failures, and improving uptime.
• In customer management, ENGIE is rolling out an entire suite of online services for customers, including self-service applications that allow them to manage their own energy use. For individual residents and building managers, ENGIE has developed an application that analyzes data from smart sensors to pinpoint opportunities to save energy.
• In renewables, ENGIE has developed a digital platform of applications to optimize generation of electricity from renewable sources. These applications use predictive analytics and AI to forecast maintenance requirements, identify underperforming assets, and provide field operators with a real-time view of assets and maintenance needs. The platform already covers more than 2 gigawatts of installed capacity and by 2020 will cover more than 25 gigawatts, making it one of the largest AI deployments in the world for renewable energy management. Today, more than 1,000 machine learning models are continuously trained to adapt to changing operating conditions, providing 140,000 predictions per day at 10-minute intervals for more than 350 wind turbines across the world. By 2020, its digital platform will address more than 20 additional machine learning use cases, unlocking significant economic value.
• In smart cities, ENGIE plans to develop and deploy a number of applications—including efficient district heating and cooling, traffic control, green mobility, waste management, and security—to create sustainable, energy-efficient connected cities, as the percentage of the world population living in cities increases from 50 percent today to 70 percent in 2050.
The list of use cases for ENGIE goes on and on. Across its organization, ENGIE will develop and deploy 28 applications over a three-year period and train over 100 employees. To coordinate and drive this change, ENGIE has established a sophisticated CoE, applying best practices to collaborate with business unit leaders, define requirements, create roadmaps, and develop and deploy applications in a systematic way that achieves measurable results.
One year in, the first four applications are already live. ENGIE is starting to see results, and the potential economic benefits are substantial: for example, the value of just two use cases—predicting and preventing equipment failures and optimizing planned downtime and dispatch operations—is expected to exceed €100 million annually.
Enel: Digital Transformation One Step at a Time
Enel, the Italian utility, is the second largest power producer in the world, with over 95 gigawatts of installed capacity, more than 70 million customers globally, €75.7 billion in 2018 revenue, and 69,000 employees. A smart-grid pioneer, Enel was the first utility in the world to replace traditional electromechanical meters with digital smart meters, a major operation carried out across Enel’s entire Italian customer base. By 2006, Enel had installed 32 million smart meters across Italy; Enel has since deployed a total of more than 40 million smart meters in Europe, representing more than 80 percent of the total smart meters on the continent.
Enel’s digital transformation has been monumental: The company has the largest deployment of AI and IoT applications in the world. Enel’s journey toward digital transformation also started at the top, led by CEO Francesco Starace who appointed Fabio Veronese, Head of Infrastructure & Technological Services, to lead Enel’s digital transformation initiative. Enel projects to spend €5.3 billion to digitize its assets, operations, and processes, and to enhance connectivity.13
Let’s dive into two specific use cases along Enel’s digital transformation journey. First is the predictive maintenance of Enel’s 1.2-million-kilometer distribution network in Italy that comprises numerous assets—substations, distribution lines, transformers, and smart meters—with sensors throughout. To further improve grid reliability and reduce the occurrence of outages and service interruptions due to asset failure, Enel deployed a prebuilt AI-powered predictive maintenance application. The SaaS application applies advanced machine learning to analyze real-time network sensor data, smart meter data, asset maintenance records, and weather data to predict failures along distribution network “feeders” (i.e., distribution lines that carry electricity from substations to transformers and end customers) before they happen. Enel can monitor assets in real time, assign day-by-day risk scores to assets, and immediately catch any anomalies or changed operating conditions that forecast emerging maintenance issues. This AI-powered predictive capability enables Enel to improve reliability, reduce its operations costs, provide greater flexibility in scheduling of maintenance tasks, significantly extend the life cycle of its assets, and improve customer satisfaction.
Key innovations in this project include (a) the ability to construct Enel’s as-operated network state at any point in time using an advanced graph network approach, and (b) the use of an advanced machine learning–based framework that continuously learns to improve the performance of predicting asset failures. Leveraging elastic cloud computing, the predictive maintenance application is able to aggregate petabyte-scale, real-time data from Enel’s grid sensors and smart meters, correlate those data with operational systems data, and, importantly, extend operational insight by subjecting those data to a comprehensive set of power analytics and machine learning algorithms.
The second use case to highlight is revenue protection. Enel transformed its approach to identifying and prioritizing electricity theft (“non-technical loss”) to drive a significant increase in the recovery of unbilled energy, while improving productivity. Enel’s vision for this transformation was an enterprise AI and IoT SaaS application that could be deployed across all of Enel’s operating entities globally within a six-month perio
d. Delivering on this vision required building a machine learning algorithm to match the performance delivered by Enel experts, who used a manual process honed over 30 years of experience. While this was a significant challenge in and of itself, Enel set an ambitious target to double the performance achieved in recent operating years.
A key innovation that enabled this transformation was replacing traditional non-technical loss identification processes. This focused primarily on improving the success of field inspections with advanced AI algorithms to prioritize potential cases of non-technical loss at service points, based on a blend of the magnitude of energy recovery and likelihood of fraud. The AI-powered revenue protection application enabled Enel to reach its target and double the average energy recovered per inspection—a significant achievement, given that Enel’s original process was based on three decades of expert experience.
Digital transformation has delivered enormous value and impact for Enel. Its efforts even earned the company a spot on Fortune’s 2018 “Change the World” list (one of only 57 companies globally)—for the third time in four years. The Fortune list recognizes companies that, through their core business strategy, improve living conditions on both a social and environmental level. Enel has imbued innovation and digital transformation throughout its organization, and it has paid off—with a projected annual economic benefit that exceeds €600 million per year.
Caterpillar: Enterprise Data Hub
Switching gears to the industrial manufacturing sector, we can look to the world’s leading manufacturer of construction and mining equipment—Caterpillar. It is an example of a company that produces extremely complex engineered products and understands the potential value in fundamentally transforming this intensive production process. In 2016, Caterpillar’s then-CEO Doug Oberhelman announced, “Today, we’ve got 400,000 connected assets and growing. By this summer, every one of our machines will come off the line being able to be connected and provide some kind of feedback in operational productivity to the owner, to the dealer and to us.” He pointed to a further vision “where we can show the customer on his iPhone everything going on with his machine, his fleet, its health, its run rate, its productivity and so on.”14
Caterpillar’s digital transformation strategy hinges on the company’s digitally connected equipment—today comprising some 470,000 assets (and expected to grow to over 2 million) in operation across customers worldwide. Caterpillar’s first step was to create an extensible Enterprise Data Hub to act as the source of enterprise data from more than 2,000 of Caterpillar’s applications, systems, and databases globally. These data include business application data, dealer and customer data, supplier data, and machine data. The data will be aggregated, unified, normalized, and federated into a single data image that supports a variety of machine learning, predictive analytics, and IoT applications across Caterpillar’s business units.
Leveraging the Enterprise Data Hub, Caterpillar is building a host of applications to enable its digital transformation. In a first instance, Caterpillar turned to its inventory. How do you manage a supply network that brings together over 28,000 suppliers to ship to 170 dealers, all with fluctuating demand? Having visibility across its supply network, understanding the transit time of parts shipped from overseas, and reducing excess inventory and spare parts inventory are critical business questions that Caterpillar is solving through the use of AI, big data, and predictive analytics.
With an AI-powered application, Caterpillar now has the ability to search and view inventory across its supply chain, receive AI-powered recommendations on optimal stocking levels, and understand the tradeoffs between stockout risks and holding excess inventory. Caterpillar has developed and deployed advanced AI-enabled solutions that provide its dealer network with visibility into finished goods inventory and sophisticated “similarity search” capability. This enables dealers to effectively meet customer demand by surfacing recommendations of in-stock products that very closely match customer requirements. The application also provides Caterpillar’s production planners and product managers with recommendations on configuration options and inventory stocking levels.
Next, Caterpillar is focused on leveraging telemetry from its entire fleet of connected assets along with data related to ambient operating conditions for each asset. Some of this telemetry is being continuously ingested in real time at a rate of over 1,000 messages per second. Such AI-enabled analytics allow Caterpillar to pinpoint anomalies in equipment health, predict asset failure, design competitive warranty offerings, and leverage the full set of operating data to design the next generation of products and features.
Caterpillar makes all these changes to its operations through the establishment of a CoE—a cross-functional team that brings together outside experts and Caterpillar developers for intensive training on how to design, develop, deploy, and maintain applications using AI and predictive analytics. The CoE’s role is to define a prioritized use case roadmap, and then implement a scalable, repeatable program to develop, deploy, and operate a portfolio of high-value applications to transform the enterprise.
John Deere: Transforming Supply Chain and Inventory
John Deere is another industrial manufacturer embarking on a digital transformation strategy to transform its supply chain. Founded in 1837, John Deere is the largest farm equipment manufacturer in the world, with more than $38 billion in annual revenue and over 60,000 employees.
A critical component of John Deere’s digital transformation is managing its inventory. John Deere operates hundreds of factories globally and makes highly complex industrial equipment. The company allows customers to configure hundreds of individual options, leading to products that have thousands of permutations. The customized nature of the product creates significant complexity in managing inventory levels during the manufacturing process. Previously, John Deere had to manage key uncertainties like fluctuations in demand, supplier delivery times, and product line disruptions. Due to these uncertainties, and since the final configuration of a product is often not known until close to the submission of a product’s order, John Deere would often hold excess inventory to fulfill orders on time. This excess inventory is expensive and complicated to manage.
Like other manufacturers in the industry, John Deere deployed material requirements planning (MRP) software solutions to support production planning and inventory management. John Deere had also experimented with different commercial inventory optimization software offerings. However, the existing software solutions were unable to dynamically optimize inventory levels of individual parts at scale while managing uncertainty and learning continually from data. Key sources of uncertainty include variability in demand, supplier risk, quality issues with items delivered by suppliers, and production line disruptions.
To address these issues, John Deere built an AI-powered application to optimize inventory levels, starting with one of its product lines that has over 40,000 unique parts. The company used an algorithm to calculate daily historical inventory levels based on a range of parameters. With the AI-powered application, John Deere was able to simulate and optimize order parameters, quantify the planned use of materials based on production orders, and minimize safety stock levels. The operational impact of these insights is significant—John Deere could potentially reduce parts inventory by 25 to 35 percent, delivering between $100 million and $200 million in annual economic value to the company.
3M: AI-Driven Operational Efficiency
Headquartered in Minnesota, 3M is a multinational conglomerate that produces tens of thousands of product variants derived from 46 core technology platforms. The company is primarily in the business of physical products, although one of its divisions, 3M Health Information Systems, provides software.
3M’s origins go back to 1902, when a group of enterprising young men founded the Minnesota Mining and Manufacturing Company, a mining venture that promptly failed. The founders, in conjunction with a group of investors and employees, refused to fold. Ins
tead they tried commercializing many different products, eventually finding success in manufacturing sandpaper. Along the way they developed a robust culture of innovation that is deeply embedded in the company today. Its numerous industrial and consumer products include many familiar brands such as Scotch Tape, Post-it Notes, and Scotchgard. Today, the 3M Company (renamed in 2002) employs more than 91,000 people and generated more than $32 billion in revenue in 2018.
CEO Mike Roman, a 30-year 3M veteran, and his leadership team are executing against the “3M Playbook”—a set of strategies that are fundamentally about simplification, optimization, innovation, and building on 3M’s strengths. As part of the playbook, 3M’s “business transformation” program is focused on increasing operational productivity while reducing costs. “In fast-changing times, companies have to constantly adapt, change and anticipate,” Roman stated shortly after taking over as CEO in 2018. “This is one reason why our transformation efforts are so critical, as we become more agile, more contemporary, more efficient, and even better equipped to serve the evolving needs of customers.”15
3M sees numerous opportunities to apply AI to drive significant improvements in operational efficiency and productivity across a wide range of business processes with direct bottom-line benefits. The company is developing and deploying multiple AI-enabled applications focused on specific high-value use cases. Let me highlight two use cases that may be of interest to any large manufacturing company.
In one use case, 3M has developed an AI application to dramatically improve its “order-to-promise” process, enabling the company to provide significantly more accurate commitments for when products will be delivered to corporate customers. Given 3M’s extensive supply chain, logistics network, and the tens of thousands of product variants it makes, this is a complex problem to solve.
Digital Transformation Page 19