Digital Transformation

by Thomas M Siebel

  Other industries like banking are informative. With each new wave of innovation, banks invest in capturing that value. Brokerages became services of retail banks; online payments were integrated; and credit card, bill payment, and financial planning offers are now standard.

  Hotels and airlines initially offered excess inventory through sites like Expedia, Hotels.com, Orbitz, Kayak, and Trivago. But they soon captured value—and strengthened customer relationships—with their own apps that offered the best pricing and options.

  The point here is that digital transformation will have profound effects, but not necessarily the effects we can predict or even measure now. Clearly, the building blocks to enable digital transformation are available, robust, and accessible: cloud computing, big data, AI, and IoT.

  Chapter 3

  The Information Age Accelerates

  C EOs and other senior leaders need to understand the technologies driving today’s digital transformation in far more detail than was required of them in the past. Why is that? In contrast to previous waves of technology adoption, we’ve seen that digital transformation goes to the very core of how organizations operate and what they do. If you are an auto manufacturer, for example, is your business fundamentally about making cars—or is it about delivering transportation and mobility to your customers? And what is more valuable: Your IP around powertrain design, or your AI-based self-driving algorithms fed by real-time telemetry and usage data generated by the vehicles you make? Leaders in every industry need to ask such questions and thoroughly examine how these technologies will profoundly change their market and how they do business.

  And the stakes have never been higher—in terms of both the risk of extinction and the potential rewards. In my experience, for any large enterprise, the annual economic value of deploying AI and IoT applications ranges from hundreds of millions to billions of dollars. At Royal Dutch Shell, for example, deploying an AI-based predictive maintenance application for more than 500,000 refinery valves globally is estimated to yield several hundred million dollars a year in reduced maintenance costs and increased operating efficiency. Shell plans to roll out many more AI applications across its upstream, downstream, and midstream operations globally. The expected benefit is on the order of billions of dollars annually.

  Implementing a digital transformation agenda means your organization will build, deploy, and operate dozens, perhaps hundreds or even thousands, of AI and IoT applications across all aspects of your organization—from human resources and customer relationships to financial processes, product design, maintenance, and supply chain operations. No operation will be untouched. It is therefore incumbent on senior leaders to firmly understand these technologies.

  My advice is to learn enough about these technologies to have well-informed discussions with your internal technical staff and to choose the right technology partners who will be critical to your success. The investment will pay valuable dividends, in terms of avoiding unqualified partners and drawn-out internal projects that never deliver value.

  This chapter provides an overview of the four key technologies that both drive and enable digital transformation—elastic cloud computing, big data, AI, and IoT. Readers who want to extend their understanding to a deeper level—particularly executives with direct responsibility for driving digital transformation initiatives—will benefit from reading the subsequent four chapters that delve further into each technology.

  The Challenges Are Daunting—But Proven Solutions Are Available

  The technologies that propel digital transformation are clearly game changing, but we are in the very early stages of this new era. And while the potential is enormous, the challenges in developing and scaling AI and IoT applications across an enterprise can be daunting.

  The baseline capability required is the aggregation and processing of rapidly growing petabyte-scale data sets—that’s 1 million gigabytes—continuously harvested from thousands of disparate legacy IT systems, internet sources, and multimillion-sensor networks. In the case of one Fortune 500 manufacturer, the magnitude of the data aggregation problem is 50 petabytes fragmented across 5,000 systems representing customer, dealer, claims, ordering, pricing, product design, engineering, planning, manufacturing, control systems, accounting, human resources, logistics, and supplier systems, fragmented by mergers and acquisitions, product lines, geographies, and customer engagement channels (i.e., online, stores, call center, field). With hundreds of millions of sensors embedded in products generating high-frequency readings of 1 Hz (1 per second) or greater, these data sets grow by trillions of readings daily.

  The technologies required to aggregate, correlate, and extract the value from these data for business transformation did not exist a decade ago. But today, the availability of inexpensive sensors (under $1) and credit card–sized AI supercomputers interconnected by fast networks provide the infrastructure to dramatically transform organizations into real-time adaptive enterprises. Cloud computing, big data, AI, and IoT converge to unlock business value estimated by McKinsey of up to $23 trillion annually by 2030.1

  A key challenge for organizations is how to bring together and leverage these technologies to create meaningful value and positive return on investment. For now, let me just say there is much good news for organizations embarking on digital transformation—robust tools and expert knowledge are now available to dramatically accelerate digital transformation efforts and ensure successful outcomes. I will return to this topic in more depth in chapter 10, “A New Technology Stack.”

  Cloud Computing

  Cloud computing is the first of the four technologies that drive digital transformation. Without cloud computing, digital transformation would not be possible. Cloud computing is a model of accessing shared pools of configurable hardware and software resources—computer networks, servers, data storage, applications, and other services—that can be rapidly provisioned with minimal management effort, typically via the internet. Those resources may be privately owned by an organization for its exclusive use (“private cloud”) or owned by a third party for use by anyone on a pay-for-what-you-use basis (“public cloud”).

  In its manifestation as the on-demand rental of compute and storage resources from a third-party provider, cloud computing was pioneered by Amazon through its Amazon Web Services unit. What began as an internal service for Amazon developers in 2002 became a public offering in 2006 with the introduction of Elastic Compute Cloud (EC2) and Simple Storage Service (S3). The public cloud computing market is estimated to reach a staggering $162 billion by 2020, just a decade and a half after its inception.2 Amazon Web Services alone is forecasted to grow to $43 billion in annual revenue by 2022.3 Competition from Microsoft and Google is acute, which guarantees rapidly falling compute and storage prices converging toward zero.

  Recognizing that cloud providers can do a better and cheaper job of running a huge number of servers and storage devices across global networks of secure and reliable data centers, organizations are rapidly shifting legacy applications (“workloads”) out of their corporate data centers into public clouds.

  Chief information officers now acknowledge traditional IT data centers will be extinct within a decade.4 Research supports that hypothesis. Cisco forecasts that by 2021, 94 percent of workloads will be processed by cloud data centers and 73 percent of cloud workloads will be in public cloud data centers.5

  Examples of companies shuttering data centers are numerous. Rome-based utility Enel is shuttering 23 data centers with 10,000 servers supporting its operations in 30 countries, and they’re consolidating 1,700 legacy applications to 1,200 applications and moving them to AWS. Netflix, Uber, Deutsche Bank, and countless others now have all or a significant percentage of their information technology operating on public clouds.6

  Virtualization and Containers

  A key enabler of cloud computing’s superior economies of scale is a technology innovation known as “virtualization.” Previously, in traditional data centers, hard
ware was sized and provisioned to handle peak demand. Organizations installed enough servers and storage to support the highest level of computing requirements they anticipated, which typically occurred only for relatively brief periods (e.g., end-of-quarter order processing). This resulted in largely idle data centers with very low hardware utilization rates averaging in single-digit percentages. Virtualization allows the creation of multiple simulated environments and dedicated resources from a single, physical hardware system. “Containers” are another innovation enabling efficient sharing of physical resources. A container is a lightweight, stand-alone, executable software package that includes everything needed to run it—code, runtime, system tools and libraries, and settings. The use of virtualization and containers to share hardware across applications results in significantly higher and vastly more cost-effective utilization rates. This translates into the highly compelling economic value proposition that is driving widespread adoption of public cloud computing platforms like AWS, Microsoft Azure, IBM Cloud, and Google Cloud.

  FIGURE 3.1

  X-as-a-Service: IaaS, PaaS, and SaaS

  Cloud computing was initially driven by both independent and corporate software developers looking to save the upfront time, cost, and effort of acquiring, building, and managing scalable and reliable hardware infrastructures. Developers were attracted to the cloud model because it allowed them to focus on developing software, while the cloud provider handled the infrastructure (Infrastructure-as-a-Service, or IaaS), scalability, and reliability.

  Today’s cloud platforms are “elastic”—that is, they dynamically determine the amount of resources an application requires and then automatically provision and de-provision the computing infrastructure to support the application. This relieves developers and IT teams from many operational tasks, such as hardware and software setup and configuration, software patching, operating a distributed database cluster, and partitioning data over multiple instances as required to scale. The cloud customer pays only for resources actually used.

  Cloud offerings now extend beyond IaaS to include application development platforms (Platform-as-a-Service, or PaaS) and software applications (Software-as-a-Service, or SaaS). PaaS offerings provide software development tools and services specifically for building, deploying, operating, and managing software applications. In addition to managing the underlying infrastructure (servers, storage, networking, virtualization), PaaS offerings manage additional technical components required by the application, including the runtime environment, operating system, and middleware.

  FIGURE 3.2

  SaaS offerings are complete, prebuilt software applications delivered via the internet. The SaaS provider hosts and manages the entire application, including the underlying infrastructure, security, operating environment, and updates. SaaS offerings relieve customers from having to provision hardware or install, maintain, and update software. SaaS offerings generally allow the customer to configure various application settings (for example, customize data fields, workflow, and user access privileges) to fit their needs.

  Multi-Cloud and Hybrid Cloud

  CIOs now recognize the importance of operating across multiple cloud vendors to reduce reliance on any one provider (so-called “vendor lock-in”) and to take advantage of differentiation in public cloud provider services. Multi-cloud refers to the use of multiple cloud computing services in a single heterogeneous architecture.7 For example, an application may use Microsoft Azure for storage, AWS for compute, IBM Watson for deep learning, and Google Cloud for image recognition.

  FIGURE 3.3

  It’s also important to be able to operate an application across private and public clouds (i.e., a “hybrid cloud” environment). Ultra-sensitive customer data might be stored in a private cloud, while public cloud infrastructure might be used for on-demand “burst capacity”—excess capacity to handle spikes in transaction processing—or other analytic processing.

  More difficult to achieve is “cloud portability” while taking advantage of native cloud provider services—i.e., the ability to easily replace various underlying cloud services that an application uses with services from another cloud vendor. For example, replacing Google’s image recognition service with Amazon’s image recognition. While the use of “containers” (technology that isolates applications from infrastructure) enables cloud portability of applications, containers do not enable portability of cloud provider services.

  Big Data

  The second technology vector driving digital transformation is “big data.” Data, of course, have always been important. But in the era of digital transformation, their value is greater than ever before. Many AI applications in particular require vast amounts of data in order to “train” the algorithm, and these applications improve as the amount of data they ingest grows.

  The term “big data” was first used in fields such as astronomy and genomics in the early 2000s. These fields generated voluminous data sets that were impossible to process cost effectively and efficiently using traditional centralized processing computer architectures, commonly referred to as “scale-up” architectures. In contrast, “scale-out” architectures use thousands or tens of thousands of processors to process data sets in parallel. Over the last dozen years or so, software technologies have emerged that are designed to use scale-out architectures for parallel processing of big data. Notable examples include the MapReduce programming paradigm (originally developed at Google in 2004) and Hadoop (Yahoo!’s implementation of the MapReduce paradigm, released in 2006). Available today under open source software license from the Apache Software Foundation, the Hadoop MapReduce framework comprises numerous software components.

  As we’ve seen, digital transformation initiatives require the capability to manage big data at petabyte scale. While the Apache Hadoop collection provides many powerful and often necessary components to help manage big data and build AI and IoT applications, organizations have found it exceedingly difficult to assemble these components into functioning applications. In chapter 10, I will discuss the need for a new technology stack, and I’ll describe how this new stack addresses the complex requirements of digital transformation. But first, there is more to the story of big data than just volume.

  Big Data Explosion

  Historically, collecting data was time consuming and labor intensive. So organizations resorted to statistics based on small samples (hundreds to thousands of data points) to make inferences about the whole population. Due to these small sample sizes, statisticians expended significant effort and time curating data sets to remove outliers that could potentially skew the analysis.

  But today, with the elastic cloud providing unlimited compute and storage capabilities, along with the emergence of software designed for parallel processing of data at massive scale, there is no longer the need for sampling nor the need to curate data. Instead, outliers or otherwise imperfect data are appropriately weighted through the analysis of large data sets. As a result, with over 20 billion internet-connected smartphones, devices, and sensors generating a stream of continuous data at a rate of zettabytes a year and rapidly growing—one zettabyte being equivalent to the data stored on about 250 billion DVDs—it is now possible for organizations to make near-real-time inferences based on all available data. As we shall see, this ability to process all the data captured is fundamental to recent advances in AI.

  FIGURE 3.4

  Another significant shift made possible by the ability to apply AI to all the data in a data set is that there is no longer the need for an expert hypothesis of an event’s cause. Instead, the AI algorithm is able to learn the behavior of complex systems directly from data generated by those systems.

  For example, rather than requiring an experienced loan officer to specify the causes of mortgage defaults, the system or machine can learn those causes and their relative importance more accurately, based on analyzing all available data for prior mortgage defaults.

  The implications are significant. An e
xperienced mechanical engineer is no longer required to predict engine failures. An experienced physician is no longer required to predict the onset of diabetes in a patient. A geological engineer is no longer required to predict oil well placement for optimal production. These can all be learned from data by the computer—more quickly and with much greater accuracy.

  Artificial Intelligence

  The third major technology driving digital transformation is artificial intelligence. AI is the science and engineering of making intelligent machines and computer programs capable of learning and problem solving in ways that normally require human intelligence.

  The types of problems tackled by AI traditionally included natural language processing and translation; image and pattern recognition (for example, fraud detection, predicting failure, or predicting risk of chronic disease onset); and decision-making support (for example, autonomous vehicles and prescriptive analytics). The number and complexity of AI applications are rapidly expanding. For example, AI is being applied to highly complex supply chain problems, such as inventory optimization; production problems, such as optimizing the yield of manufacturing assets; fleet management problems, such as maximizing asset uptime and availability; and health care problems, such as predicting drug dependency risk, to name just a few. I will cover some of these in more detail later in the book.