Cognitive Technologies in the Technology Sector from Science Fiction Vision to Real-World Value

Cognitive technologies in the technology sector From science fiction vision to real-world value

A Deloitte series on cognitive technologies Deloitte offers a broad range of services to clients in all sectors of the technology value chain, from computer hardware and software vendors to networking equipment, semiconductor, and IT services organizations. Read more about our Technology industry practice on www.deloitte.com. From science fiction vision to real-world value

About the authors

Daniel T. Pereira Daniel Pereira, Deloitte Services LLP, is a market insights manager in technology, media, and telecommunications. Prior to joining Deloitte, Pereira was the managing director and research affiliate on two corporate sponsored research projects at the Massachusetts Institute of Technology and a strategic consultant specializing in cross-sector innovation, business model generation, and lean startup methodologies.

David Schatsky David Schatsky, Deloitte LLP, tracks and analyzes emerging technology and business trends, including the growing impact of cognitive technologies, for Deloitte’s leaders and its clients. Before joining Deloitte, Schatsky led two research and advisory firms. He is the author of Signals for Strategists: Sensing Emerging Trends in Business and Technology (RosettaBooks, 2015).

Paul Sallomi Paul Sallomi is a partner in Deloitte Tax LLP and the global technology, media, and telecommunications industry leader. He is also the US and Global Technology Sector Leader. In his role with the US member firm, Sallomi helps clients transform their business and operating models to address the emergence of new, disruptive technologies. For nearly 25 years, he has worked with clients in the technology, media, and telecommunications industries to shape and execute options for growth, performance improvement, regulatory compliance, and risk mitigation.

Robert (Bob) Dalton Bob Dalton is a principal and 30-year veteran with Deloitte Consulting LLP, where he leads the global relationship with the IBM Corporation. Deloitte and IBM have a complex ecosystem involving a bi-directional client relationship as well as a significant alliance relationship. As part of the alliance relationship, there is a significant focus on business analytics, information management, and cognitive technologies. Deloitte is actively working with the core Watson business unit as well as Watson Health.

iii Cognitive technologies in the technology sector

Contents

Introduction | 1

Which cognitive technologies are hot? | 3 The tech sector’s big M&A push

Why companies embrace cognitive technologies | 5 A path to business model transformation

How companies are innovating | 8 Platforms and PaaS

Market implications and considerations | 13

Harnessing cognitive technologies | 16 Where to start

Endnotes | 18

iv From science fiction vision to real-world value

Introduction

RTIFICIAL intelligence is certainly no players—just about every technology subsector Alonger considered science fiction—or a has seen a substantial upsurge of activity in source of expensive R&D efforts with unmet this space. In fact, the race to invest in arti- potential—by major players in the technol- ficial intelligence has been described as “the ogy sector.1 Instead, we are in the midst of latest Silicon Valley arms race.”3 Since 2012, a real-world paradigm shift: the final stages there have been 100 mergers and acquisitions of a decades-long transition from the scien- (M&A) within the technology sector involv- tific discipline known as artificial intelligence ing cognitive technology companies, products, (and its various sub-disciplines) into an array and services. And this rush of M&A activity of applied cognitive technologies made more is not the only sign of the industry’s interest. widely available through innovative enterprise Many capabilities that were only just emerging architectures unique to the business culture of a few years ago are now essentially mature and the technology sector. becoming “democratized” and more readily The technology sector’s interest in these available for business applications. As a result, technologies (figure 1)2 has exploded in the last leading companies are using cognitive technol- several years. Networking companies, semi- ogies to enhance their existing products and conductor manufacturers, hardware compa- services, as well as to open up new markets. nies, IT providers, software providers, Internet

Figure 1. Widely used cognitive technologies

Natural Computer Machine language vision learning processing

Speech recognition Optimization

Rules- Robotics Planning based systems & scheduling

Graphic: Deloitte University Press | DUPress.com 1 Cognitive technologies in the technology sector

What is interesting is that the assertive large and midsized technology companies, actions of the sector’s leaders do not mirror as well as next steps for not only technology the wholesale adoption of these technologies sector companies but also technology-enabled across the industry. Many technology sector enterprises in any industry or vertical market companies have yet to turn their attention to that want to further explore cognitive tech- how cognitive technologies are changing their nologies’ potential. sector or how they—or their competitors— • Computer vision: The ability of comput- may be able to implement these technologies ers to identify objects, scenes, and activi- in their strategy or operations. ties in unconstrained (that is, naturalistic) visual environments

• Machine learning: The ability of computer systems to improve their performance by Many technology sector exposure to data without the need to follow companies have yet to explicitly programmed instructions • Natural language processing (NLP): The turn their attention to how ability of computers to work with text the way humans do—for instance, extracting cognitive technologies are meaning from text or even generating text that is readable, stylistically natural, and changing their sector or how grammatically correct they—or their competitors— • Speech recognition: The ability to automatically and accurately transcribe may be able to implement human speech • Optimization: The ability to automate these technologies in their complex decisions and trade-offs about strategy or operations. limited resources • Planning and scheduling: The ability to automatically devise a sequence of actions to meet goals and observe constraints To help leaders better understand these issues, this report considers three perspec- • Rules-based systems: The ability to use tives. First, we examine recent M&A activity to databases of knowledge and rules to determine which cognitive technology capabil- automate the process of making inferences ities are attracting the most attention. Second, about information through the examples of the IBM Watson • Robotics: The broader field of robotics is Group and Alphabet/Google, we explore how also embracing cognitive technologies to cognitive technologies can transform business create robots that can work alongside, inter- models, helping companies compete in the act with, assist, or entertain people. Such future marketplace. We also focus on the two robots can perform many different tasks main approaches technology companies are in unpredictable environments, integrat- using to pursue marketplace opportunities: ing cognitive technologies such as com- cognitive technology development platforms puter vision and automated planning with and cognitive technology platform-as-a-service tiny, high-performance sensors, actuators, (PaaS) offerings. We conclude with a discus- and hardware. sion of the go-to-market considerations for

2 From science fiction vision to real-world value

Which cognitive technologies are hot? The tech sector’s big M&A push

OGNITIVE technologies encompass a spikes in robotics company acquisitions in Cdiverse set of capabilities, and the pat- 2013 and machine learning and speech recog- tern of technology companies’ M&A activity nition company acquisitions in 2014 (figure 3). over the last three years reflects this diversity. Further analysis of the announcements for Targeted deals have recently surged, spanning these M&A deals makes it clear that, in the everything from robotics and sensor specialty overwhelming majority of cases, the acquirer’s firms to machine learning and natural lan- intent is to generate new revenue from new guage processing (NLP) companies.5 customers and new markets by improving A close analysis of 100 mergers and product and service innovation. Some com- acquisitions involving cognitive technolo- panies prefer to use cognitive technologies gies between 2012 and 2015 reveals strong to fuel innovation in existing products, while strategic investments by the tech sector in others aim to build stronger technology plat- certain technologies. forms for a wide array of solutions, offerings, Some general trends point to which cogni- and operations. tive technology capabilities are the most sought Our review of the marketplace points to after: M&A activity has been highest among three main ways in which industry leaders are machine learning and computer vision com- harnessing these technologies: panies since 2012 (figure 2), with noticeable

Figure 2. M&A transactions by cognitive technology capability 35 32 Cognitive computing capabilities by total number of transactions (2012–2015*) 30

25 24

15 14 13 12 10

5 5

0 Machine Computer Robotics Speech Natural Expert system: learning vision recognition language Planning & processing scheduling *Until Deceber 1, 2015 Source: Deloitte analysis. Graphic: Deloitte University Press | DUPress.com

3 Cognitive technologies in the technology sector

Figure 3. Technology sector M&A deals involving cognitive technologies

Cognitive technology deals 37 during 2012–2015* 3

2012 2013 2014 2015 Computer vision Machine learning Speech recognition

Robotics Natural language processing Expert system: Planning & scheduling

*Until Deceber 1, 2015 Source: Deloitte analysis. Graphic: Deloitte University Press | DUPress.com

Business model transformation: New Platform-as-a-service offerings: Modular, business units are created to increase volume extensible products are redesigned for the and grow revenue through cognitive technol- computing-intensive demands of cognitive ogy-enabled innovation. technologies, allowing both current and new Development platforms: Platforms that customers to easily transition to PaaS offerings allow for open collaboration with an extended as well as companies to rapidly position their developer community are built, accelerating PaaS offerings in new markets. the speed and scalability of product development and product release efforts.

4 From science fiction vision to real-world value

Why companies embrace cognitive technologies A path to business model transformation

HY are technology companies pursu- operations, process, and business model Wing cognitive technologies so aggres- innovation as well. These new units are also sively? Research has shown that companies designed to transform the architecture of the that exhibit superior performance over the parent company over time. Such restructuring long term have two distinctive attributes: They underlines cognitive technologies’ potential tend to differentiate themselves based on value to completely revolutionize the technology rather than sector—and take price, and they many vertical seek to grow industries and revenue before The most striking finding is markets along. cutting costs. Thus a clear that technology companies IBM implication of Watson our analysis is create new business units that by prioritiz- Group ing substantial to increase volume and Perhaps the investments most clear-cut in cognitive generate revenue by using example of a technologies, technology com- companies aim cognitive technologies. pany pursuing to grow revenue business model by creating value transformation through new through cogni- or better products or services rather than by tive technologies is IBM. In January 2014, IBM cutting costs. Beyond the incentives of greater invested $1 billion to launch the IBM Watson revenue and market share, however, a deeper Group business unit; of this $1 billion, $100 reason appears to be that these companies see million was earmarked as an investment fund cognitive technologies as a way to reinvent “to support IBM’s recently launched ecosystem themselves to more effectively compete in of start-ups and businesses that are building a the future—essentially, as a basis for business new class of cognitive apps powered by Watson model transformation. in the IBM Watson Developers Cloud.” As Indeed, the most striking finding is that Ginni Rometty, IBM’s chairman and CEO, technology companies create new business noted at the new unit’s launch: “For those of units to increase volume and generate revenue you who watch us, we don’t create new units by using cognitive technologies not only for very often. But when we do, it is because we product innovation but also for structural, see something that is a major, major shift that

5 Cognitive technologies in the technology sector

we believe in.”8 This major shift is what IBM IBM’s investment in IBM Watson Group considers the dawn of a new era in the technol- and in its ecosystem-based business model ogy sector—what it is calling the “cognitive highlights that—as is the case with many cog- computing era.”9 nitive technology product innovations—com- Designed as a radical business model trans- panies are not attempting to generate revenue formation, IBM Watson Group seeks to build a directly from an easily identifiable “cognitive robust business ecosystem, which is a complex, technology market.” Instead, they are devel- dynamic, and adaptive community “of diverse oping and expanding cognitive technology players who create new value through increas- capabilities that are already generating revenue ingly productive and sophisticated models of and market share in the fast-growing areas of both collaboration and competition.”10 To this data analytics and cloud computing. As IBM’s end, the IBM Watson Group provides open- Rometty explained in a recent interview, “Now, source infrastructure for strategic partners it is about speed and scale. . . . Our data analyt- to develop cognitive technology services in ics business is a $17 billion business already; a host of different vertical markets (such as cloud, $7 billion. These are big already. We health care, financial services, media, and now scale them: more industries, more people, telecommunications). This business model more places.”14 allows the IBM Watson Developers Cloud to As evidence of this commitment to speed further expand as IBM acquires more cognitive and scale—as well as an early sign of the technology companies. transformation of the company’s very architec- An example of this acquisition strat- ture—IBM has recently made additional moves egy is IBM’s acquisition of Denver-based in the cognitive technology space. In April AlchemyAPI, a cloud computing platform, 2015, it announced the launch of IBM Watson now part of the IBM Watson Group. Health and the Watson Health Cloud plat- AlchemyAPI created an NSL-based text form. Soon after, in August 2015, IBM Watson analysis solution for developers along with Health paid $1 billion for Merge Healthcare software development kits (SDKs) in various Inc.’s medical imaging management platform. programming languages,11 as well as a platform This acquisition, while not directly involving for understanding pictures’ content and con- a cognitive technology capability, gives IBM text, which uses computer vision deep learn- access to a data and image repository that can ing12 not only for facial recognition, but also become an image training library for cognitive for image tagging and understanding complex computing solutions directed at the health care visual scenes. The IBM Watson visual recogni- market.15 In October 2015, IBM CEO Rometty tion service now utilizes this technology to proclaimed that transforming health care recognize physical settings, objects, events, and through cognitive computing is “our moon- other visuals. By January 2015, the service had shot.”16 (Moonshots are a concept fast becom- more than 2,000 classifiers and trained labels ing popular within the technology sector to for categories such as animal, food, human, describe audacious projects—projects that scene, sports, and vehicle.13 At the time of the aspire to exponential, tenfold improvements AlchemyAPI acquisition, 160 ecosystem part- [1,000 percent increases in performance]).17 ners were developing applications on the IBM Finally, in the same month, IBM formed yet Watson Group’s platform; this deal brought another cognitive technology business unit, the in an additional 40,000 developers who had Cognitive Business Solutions Group.18 already built code on top of the AlchemyAPI service platform.

6 From science fiction vision to real-world value

Alphabet, Inc. (formerly Judging from Google’s recent M&A activ- Google, Inc.) ity, cognitive technologies are integral to this business model transformation: Google In August 2015, Google Inc. announced its acquired 20 cognitive technology companies reorganization into the giant holding com- between 2012 and 2015. In retrospect, these pany Alphabet Inc., making Google (includ- acquisitions were the strategic precursor to ing Search, Android, and YouTube) a wholly the restructuring of Google into Alphabet. Of owned subsidiary of Alphabet. Like IBM these 20 deals, 8 robotics acquisitions were Watson Group, Alphabet is an ambitious struc- made in 2013 alone, all of which will be inte- tural and operational commitment to speed grated into Alphabet. Their robotics capabili- and scale—one that is fundamentally informed ties range from humanoid robotic systems, by and designed to enable the exponential the next generation of robotic arms, high-tech growth of cognitive technologies. robotic wheels that can move in any direc- “Moonshot thinking” was highly encour- tion, robots that use computer vision to better aged at Google and will continue to be encour- understand what they’re looking at and learn 19 aged at Alphabet, as the company hews to how to handle non-standard situations, and an emerging business strategy framework robotics in the fields of filmmaking, advertis- known as “exponentials,” which is based on ing, and design.24 A robotics acquisition from the exponential acceleration of technologies June 2014, the Massachusetts-based Boston such as quantum computing, artificial intel- Dynamics, has been restructured as a stand- ligence, robotics, additive manufacturing, alone Alphabet company.25 and synthetic or industrial biology. These and Another standalone Alphabet company is other exponential technologies are creating London-based DeepMind, which was acquired new competitive risks and opportunities for by Google in February 2014 for $400 million. companies that have historically maintained Besides its innovation in neural networks for 20 dominance in their industries.” deep learning, DeepMind also has natural lan- Alphabet is, by design, an exponential guage and computer vision capabilities made 21 organization (ExO) —an organization “whose available through two Google acquisitions in impact (or output) is disproportionally large, the fall of 2014 (Vision Factory AI and Dark at least 10 times larger, compared to its peers Blue Labs). Other artificial intelligence capabil- because of the use of new organizational ities integrated into Alphabet through previous 22 techniques.” Google is No. 5 on the Top Google M&A deals include facial-recognition 100 ExOs list, along with companies such as software, gesture recognition, camera-based Airbnb, Uber, Tumblr, Medium, and Twitch. language translation, image recognition via At No. 1 is the collaborative code and soft- smartphone cameras, machine learning for 23 ware development site GitHub. Alphabet is scheduling and task management, patents for making 10x growth a business priority sup- a speech interface for search engines (with a ported from inception by the new company’s system and method of modifying and updating enterprise architecture. a speech recognition program), and neural networks that improve voice and image search.26

7 Cognitive technologies in the technology sector

How companies are innovating Platforms and PaaS

F business model transformation predicated enables the company to rapidly position Ion cognitive technologies is the endgame—or its cognitive technology PaaS offerings one possible endgame—how are companies in new markets. Again, the IBM Watson pursuing it? Our review of technology com- Group is illustrative: Watson Services on panies’ activity around cognitive technologies Bluemix is an open-standard, cloud-based and product innovation suggests that these PaaS for building, managing, and running innovations fall into two categories: applications of all types, such as mobile, big data, and new smart devices, provid- • Development platforms: Platforms— ing a fully integrated service for cognitive which are increasingly supported by global technology innovation.29 digital technology infrastructures that help Typically, these platforms and PaaS to scale participation and collaboration— offerings target high-value, immediately help make resources and participants more addressable markets such as analytics, cloud accessible to each other. Properly designed, computing, social, mobile, and security—all they can become powerful catalysts for rich of which grew almost 20 percent in 2014.30 ecosystems of resources and participants, Following are some examples from various defining the protocols and standards that technology subsectors. enable a loosely coupled, modular approach to business process design.27 Cognitive technology developer platforms are emerg- The platform play: ing that allow organizations to collaborate Cognitive technology with a community of passionate developers, development platforms which accelerates the speed and scalability of product development and product Intel’s RealSense Technology Platform. release efforts. An example is the IBM Launched in 2014 the RealSense Technology Watson Group’s development platform, the Development Platform and Software Watson Developers Cloud. Development Kit (SDK) are designed to con- nect developers with the tools and resources to • PaaS extensions: PaaS vendors provide a develop applications with cognitive comput- virtual IT environment that allows busi- ing technologies. At RealSense’s core are the nesses to develop and run their own capabilities of two companies Intel acquired customized applications without having to in 2013: Omek Interactive, a computer vision manage the details of a physical or cloud- company with a development platform based based data center.28 Many companies have on 3D depth sensor cameras, and Indisys, an enhanced current PaaS offerings through NLP company with a background in computa- modular, extensible products designed tional linguistics, artificial intelligence, cogni- for cognitive technologies’ computing- tive science, and machine learning. With these intensive demands. This approach allows a cognitive technology capabilities, RealSense company’s current and new customers to aims to provide a platform for the develop- easily transition to the PaaS offering and ment of “perceptual computing”—Intel’s term

8 From science fiction vision to real-world value

for gesture, touch, voice, and other sensory for manufacturing graphics processor units technologies. Intel’s objective is to transform (GPUs)35 and software for the gaming industry, the human-computer interface by empowering NVidia has released a variety of hardware and developers to “integrate hand/finger tracking, software product enhancements designed to facial analysis, speech recognition, augmented increase their applicability to cognitive tech- reality, background segmentation, and more nology product development and to support into your apps.”31 customers’ cognitive technology applications. The company is specifically positioning Nvidia is placing its bets on GPU-accelerated cognitive technologies as enabling the next deep learning, which it says will help devel- generation of more intuitive gestural controls opers bring increasingly smarter cognitive in the fast-growing mobile market.32 Patrick technologies to everything, from online image databases to house- hold items to autonomous vehicles. In July 2015, the Typically, these platforms and company announced a series of updates to Digits Deep PaaS offerings target high- Learning GPU Training System version 2 (Digits value, immediately addressable 2), its GPU-accelerated deep learning software, markets such as analytics, cloud as well as to CUDA Deep Neural Network,36 its GPU- computing, social, mobile, and accelerated library of algorithms to create and train security—all of which grew larger, more accurate neural networks through faster and almost 20 percent in 2014. more sophisticated model training and design.37 Based on company information made available in a Q2 2015 Moorehead of Forbes states that percep- corporate earnings call, Nvidia has developed tual computing “is the key to Intel’s future, GPU-accelerated software that doubles deep as it soaks up high degrees of computing learning performance for data scientists and resources.”33 If perceptual computing becomes researchers. The company also reported that inherent in the way people use their mobile it has over 3,300 developers and companies devices, Moorehead may be right: Sales of interested in machine learning and that “deep mobile devices such as the Apple iPad mobile learning is a new, exciting application.”38 digital device and Microsoft Surface are At the hardware level, the company’s Tegra expected to reach 21.5 million units this year— microprocessors are built for embedded prod- an increase of 70 percent over 2014—and 58 ucts, mobile devices, autonomous machines, million units by 2019.34 In addition, RealSense and automotive applications. In January 2015, is available as a development platform for the Tegra® X1, an innovative mobile chip with Windows 10, which has been optimized for over one teraflop of processing power39 and cross-platform development (PC, smartphone, “capabilities that open the door to unprec- tablet, and so on). edented graphics and sophisticated deep Nvidia Corporation’s Digits 2 and CUDA learning and computer vision applications,”40 Deep Neural Network. A company known was released.

9 Cognitive technologies in the technology sector

The “cloud wars”: Cognitive proposition is to view investment in Amazon technology PaaS Machine Learning as creating the future growth opportunities for AWS, which is now The so-called “arms race” in cognitive tech- positioned as a growing cloud computing nologies began long ago between the technol- capability for the company. Said one industry ogy company superpowers—Google, Amazon, observer in August 2015: Microsoft, IBM, Facebook, and Apple—in the form of investments in strategic R&D. With the It may seem odd for an e-commerce giant like Amazon to expand into machine learn- enterprise cloud market expected to grow from ing, but it’s an important way to enhance $70 billion to more than $250 billion by 2017, its growing cloud business. Amazon’s this race has escalated into full-blown combat, cloud and big data efforts belong to AWS. with machine learning as the battleground.41 Amazon doesn’t report AWS revenues Amazon Web Services (AWS) Machine separately, but analysts at Pacific Crest esti- Learning. In the case of Amazon, machine mate that AWS generated nearly $5 billion in 2014 revenues, up from an estimated learning as a service dates back to the compa- $3.1 billion in 2013. That’s just a sliver of ny’s early offerings of infrastructure as a service Amazon’s revenues of $89 billion last year, (IaaS) and software as a service (SaaS) through but it’s growing at a much faster rate than Amazon Web Services (AWS).42 Amazon’s its core product sales revenue, which rose 15 46 recent launch of a machine learning-based percent annually to $70 billion. [Emphasis PaaS offering can be seen as a direct result of added.] its significant, long-term strategic research and Soon after, Amazon did report AWS earn- development investments, notwithstanding the ings separately, and the results are impressive. criticism these investments have attracted from The Wall Street Journal recently reported that investors and Wall Street.43 in Q3 2015, the “cloud-computing unit… In April 2015, Amazon.com launched reported a 79 percent sales increase to $2.09 Amazon Machine Learning, an enhancement billion, compared with analyst expectations of to AWS that makes it easier for develop- just under $2 billion. The division’s operating ers to build predictive models for a variety profit of $521 million was nearly as much as of use cases, including detecting potentially the total for Amazon’s North America retail fraudulent transactions, reducing customer operation. Amazon has said AWS… could one churn rates, and improving customer sup- day surpass the core retail business in size.”47 port.44 Traditionally, building applications Microsoft’s Azure Machine Learning. with machine learning capabilities necessi- Like Amazon, Microsoft (through Microsoft tated expertise in statistics and data analysis, Research) has always been committed to long- as well as experience in training a model with term strategic R&D in machine learning; the machine learning algorithms. Evaluating impact of the technology is now pervasive at and refining the model and then generat- the company.48 And like Amazon, Microsoft ing analytics using the model were extremely is seeking to use machine learning as a way to labor-intensive. Amazon Machine Learning, generate new revenue from existing custom- in contrast, automates many of these labor- ers of its public cloud computing platform, intensive steps and reduces their complexity, Microsoft Azure. To this end, in June 2014, making machine learning more accessible to Microsoft launched Azure Machine Learning, software developers.45 an open, multisided platform49 designed to It is difficult to directly trace a specific accelerate and scale collaboration within a return on investment for Amazon Machine developer community and among a growing Learning back to the initial strategic R&D business ecosystem of strategic partners. As a investments (spent over a very long period). next step, in January 2015, Microsoft acquired However, one way of framing the value Revolution Analytics Inc., a major commercial

10 From science fiction vision to real-world value

contributor to the open-source R project, scientist community.56 In 2012, the predictive adapting Revolution’s scalable R distribu- analytics market was a $2 billion category of tion for Azure Machine Learning, “making the overall analytics market worldwide, and it much easier and faster to analyze big data, is expected to more than triple to $6.5 billion and to operationalize R code for production worldwide in 2019.57 HPE is projecting $3 bil- applications.”50 lion in annual cloud revenue for 201558 while R is one of the first open-source program- further positioning IT operations analytics ming languages and developer communi- (data centers optimized by machine learn- ties, and is emerging as an industry standard ing algorithms for IT systems management) amongst data scientists.51 Azure Machine as a strong niche market in the enterprise Learning is now an advanced analytics service cloud market.59 that allows the development of applications in SalesforceIQ, Oracle Social Cloud, and just a few hours (a process which once took Pegasystems Pega 7: In July 2014, cloud a few weeks and required additional internal pioneer Salesforce.com acquired RelateIQ— technical resources). an American enterprise software company In an effort to move away from the slug- based in Palo Alto, California—for $390 gish PC market, cloud computing is a priority million. RelateIQ’s relationship intelligence segment for Microsoft. The company, which platform uses unstructured data (such as reported a doubling of revenue for its Azure email, smartphone calls, and appointments) cloud services in Q3 2015, will continue to rely to augment customer relationship manage- on Azure Machine Learning for ongoing PaaS- ment through data science and machine based revenues (as opposed to the single-prod- learning. In September 2015, at its annual uct sales model of other product segments).52 Dreamforce users’ conference, the company Hewlett-Packard Enterprise (HPE) Haven rolled out SalesforceIQ for Small Business Predictive Analytics. The recently formed and SalesforceIQ for Sales—both of which are HPE is yet another company that has added a based on the machine learning capabilities cognitive extension to a platform in an effort made available by the RelateIQ acquisition and to appeal to a growing developer community, designed as extensible modules to its industry- scale and accelerate product development, and leading customer relationship management boost the adoption rates of its cloud comput- (CRM) platform.60 ing platform. The platform in question is HPE’s To compete with Salesforce in an emerging Haven Big Data platform, a scalable open CRM market—social customer relationship platform for enterprise data analytics.53 management, or social CRM—Oracle Corp. In February 2015, prior to the recent and Pegasystems Inc. both made acquisitions restructuring, Hewlett Packard (HP) rolled out of cognitive companies with applications of Haven Predictive Analytics, an enhancement NLP to unstructured social media conversa- designed to accelerate large-scale machine tions. These capabilities are now a part of the learning, statistical analysis, and graph pro- Oracle Social Cloud and the Pegasystems Pega cessing,54 enabling customers—in this case, 7 platforms—automating analytics of real-time cognitive technology developers for IT opera- social media, amongst other cognitive-enabled tions—to build models for comprehensive functionalities.61 By 2019, the global market for machine learning and provide enterprise-wide social CRM is expected to surpass $13 billion statistical analysis of structured, unstructured, with a CAGR of 38 percent.62 Overall, CRM is and previously underutilized data volumes.55 a fast-growing segment of the enterprise cloud The features that differentiate this product computing market—with over 50 percent of from the competition include ease of use and CRM implementations expected to be cloud- (as in Microsoft’s case) greater focus on the based in the next five years.63 switching costs for the R programmer/data

11 Cognitive technologies in the technology sector

Cisco Systems’ Cognitive Threat sets. Instead, statistical modeling and machine Analytics. In 2014, Cisco announced new learning are used to analyze multiple param- solutions in advanced network security that eters while taking in real-time traffic data, were based, in large part, on products obtained identifying new threats, learning from the from its 2013 acquisition of the Czech com- data volumes, and adapting over time.67 pany Cognitive Security.64 The new Cisco plat- Cognitive Threat Analytics is available through form, known as Cognitive Threat Analytics, an add-on license to any Cisco Cloud Web is a cloud-based solution (which reduces Security solution. time to discovery of threats operating inside Cybersecurity is one of the largest and a network) rooted in a set of advanced algo- fastest-growing markets within the technology rithms and machine learning techniques that sector—with worldwide cybersecurity market Cognitive Security developed over 10 years estimates of $77 billion in 2015 and $170 bil- prior to the acquisition.65 lion by 2020.68 The security analytics segment Machine learning addresses the grow- is expected to grow from $2.1 billion in 2015 to ing demand for cybersecurity threat analyt- $7.1 billion by 2020, at a CAGR of 27.6 percent ics through behavioral analysis and anomaly from 2015 to 2020.69 In addition, “the market detection, which identifies disparities in the for Secure Web Gateways (SWGs) solutions outer defenses of a network that are then is still dominated by traditional on-premises used to identify the symptoms of a malware appliances. But the use of cloud-based ser- infection or data breach.66 Unlike traditional vices is growing rapidly, and advanced threat network security techniques (such as inci- protection functionality remains an important dent response systems), Cognitive Threat differentiator.”70 Analytics does not depend on manual rule

12 From science fiction vision to real-world value

Market implications and considerations

ECHNOLOGY companies looking to take Deep learning algorithms, as well as visual, Tcognitive technology capabilities to market gestural, and spatial computational needs, are may want to consider the following factors. emerging as focus areas for the technology Know where to play. Regarding the go-to- industry because they all require high levels of market activities we have discussed, the most computational resources. This opportunity to discernible trend is the commercial develop- address future processing challenges has not ment of machine learning and NLP. Today’s escaped the semiconductor subsector: Besides PaaS offerings have been designed to lower the barriers to entry to these capabilities and, more Deep learning algorithms, as well importantly, to extract value over the lifetime total value as visual, gestural, and spatial of current customers versus the average revenue per computational needs, are emerging as unit of traditional, linear unit sales and distribution. focus areas for the technology industry In the short term, we expect machine learning platforms because they all require high levels of and product innovation opportunities to continue to computational resources. generate revenue and attract new customers in the following markets: IT operations management, the previously mentioned NVidia Tegra mobile data analytics (specifically predictive analytics, chip, the IBM TrueNorth chip,74 Intel’s Xeon E7 IT operations analytics (ITOA), organizational v3 server chip,75 and Qualcomm’s Snapdragon analytics71 and advanced threat analytics), mobile processor76 are also designed to address along with customer relationship management the need for computational innovation. (CRM). NLP, meanwhile, will fuel the growth The long view. Two long-term, strategic of social CRM innovation. markets have been previously mentioned: The When entering a new market, it bears moonshot by IBM in the health care arena and keeping in mind the conventional wisdom further capabilities (besides the role speech that new markets are more difficult to enter recognition is already playing) designed into than existing markets.72 Also keep in mind the mobile devices, contributing to their ease of baseline metric for the return on investment of use and thereby enhancing their rapid adop- these license- or subscription-based business tion rate globally. Other emerging applica- models: The ratio of the lifetime total value tions and markets for cognitive development of customers for SaaS offerings relative to the platforms include wearables,77 the Internet cost of customer acquisition can be as high as of Things (IoT),78 machine-to-machine 5 to 1.73

13 Cognitive technologies in the technology sector

(M2M) communication79 and the future of open developer communities. We know this transportation mobility.80 has already occurred at IBM Watson Group, Act now to avoid paralysis. As these cogni- for example. tive technology activities demonstrate, artifi- Cultivate both specialized expertise and cial intelligence is no longer a pie-in-the-sky institutional knowledge. Lowering barri- endeavor or an R&D initiative isolated from ers to entry to what have traditionally been revenue growth or the customer relationship. stand-alone, expensive, and complex artificial Arguably, the ability to tactically execute was intelligence projects is the fundamental value not broadly available across all the technology proposition we have described, with cogni- subsectors even a year ago. A product inno- tive computing now an applied technological vation approach that utilizes the traditional extension of existent, scalable IT capabili- competitive advantage or “me too” approach ties. However, one cannot assume that all IT to develop products would be an attempt to professionals have the scientific background use old tools to deploy new technologies in a that deployment of cognitive technologies may fundamentally transformed product distribu- require. With the automation of high-level pro- tion channel, and thus be unsuccessful. In this gramming tasks and the management of the interoperability of old and new data sets as examples comes the risk that people Change at this speed and scale creates will blindly use functions of which they do not have opportunities—and risk—which will a deep understanding, potentially causing sys- challenge strongly held beliefs at the tems to perform poorly when model assumptions core of a company’s business model. are violated by changes in real-world events.83 Alternatively, a data scientist may not be familiar with transformed environment, avoid paralysis and how these new technologies should be inte- denial. The art of the possible is now. Begin grated into an organization’s larger enterprise by determining which markets and business architecture and deployment capabilities. priorities are immediately addressable, then The API economy. Application program- apply them to formative cognitive-technology ming interfaces (APIs) may be where IT product-innovation value maps, technology professionals and cognitive technology experts roadmaps, and go-to-market strategies, while will find common ground. Machine learning seeking out new product innovation toolsets. APIs are readily available through both AWS84 Beware the talent squeeze. Talent is and Microsoft Azure Marketplace,85 while the becoming increasingly scarce in the various IBM Watson-powered API Harmony ser- artificial intelligence sub-disciplines we have vice on Bluemix allows developers to search discussed.81 Major US technology players are publicly available APIs based on wide-ranging not only competing for market share but for search criteria.86 Companies will also find talented individuals as well.82 The worst-case opportunities to provide support for develop- scenario is that companies will consider certain ing and applying cognitive technologies—and in-house talent as mission-critical resources APIs will be the form of and the building and thus limit access to them. In the best-case blocks of these new products and services.87 scenario, access to these experts and their Look for open platforms to support subject matter expertise is incorporated into cognitive technologies’ development. The

14 From science fiction vision to real-world value

examples cited in our analysis demonstrate architecture, open cognitive technology the importance of open platforms based on platforms will experience a rapid expansion open source technology architectures with similar to that of the mobile market since 2007. their roots in the open source movement Learn to manage the new, new thing: for software development. Recently, Google Risk. Change at this speed and scale creates announced the public availability of a portion opportunities—and risk—which will chal- of its TensorFlow AI deep learning engine88 lenge strongly held beliefs at the core of a and the IBM machine learning system, company’s business model. As a result, new SystemML, is now an Apache Incubator open value creation opportunities may exist at the source project.89 Facebook has also donated to edges, not at the core, of your organizational Torch, an open source software project that is capabilities.91 Meanwhile, exponential change focused on deep learning.90 empowers new entrants while fundamentally Open platforms enhance collaboration challenging the cost structures and delivery and create new market behaviors—and the models of incumbent companies.92 It helps prevalence of open development platforms technology companies to ask themselves: points to the technology sector’s hope that the How is your organization formulating risk adoption rate of cognitive technologies will vis-a-vis cognitive technologies? And what are mirror the success of mobile open develop- the new frameworks and methodologies for ment platforms—most notably Google’s open quantifying and mitigating risk when applying source Android mobile development platform. cognitive technologies to traditional business By replicating the open source technology problems—and utilizing these platforms and PaaS offerings?

15 Cognitive technologies in the technology sector

Harnessing cognitive technologies Where to start

RGANIZATIONS that wish to make or the scale of available resources. Other Oexpand investments in cognitive technol- C-suite executives should support these ogies may want to consider the following steps. efforts by encouraging methodologies that These steps can apply to not only technology are appropriate to each specific technology companies but also technology-enabled enter- (machine learning, computer vision, speech prises in any industry or vertical market that recognition, and so on). wants to further explore cognitive technologies’ potential to enable innovative business 2. Learn and explore: Continue to learn partnerships, platform solutions, new business about and understand these technologies, models, and product innovation. exploring various use cases where they have been applied, and whether to enable 1. Gain C-suite support: Ensure you have business model transformation or to pursue strong C-level support across your organi- product innovation. zation for the assessment of business priorities related to cognitive technologies. The 3. Flesh out your own possible use cases: CEO should make cognitive technologies Assess where and how these technologies an executive-level responsibility, increasing could positively impact your company,

Figure 4. Use the Three Vs to help identify opportunities for cognitive technologies

Screen Cognitive technology indicators Application examples

• All or part of a task, job, or workflow • Forms processing, first-tier customer service, requires low or moderate level of skill warehouse operation Viable plus human perception • Investment advice, medical diagnosis, oil • Large data sets exploration • Expertise can be expressed as rules • Scheduling maintenance operations • Workers’ cognitive abilities or training • Writing company earnings reports; e-discover, are underutilized driving/piloting Valuable • Business process has high labor costs • Health insurance utilization management • Expertise is scarce; value of improved • Medical diagnosis; aerial surveillance performance is high • Industry-standard performance requires • Online retail product recommendations use of cognitive technologies Vital • Fraud detection • A service cannot scale relying on human • Media sentiment analytics labor alone

16 From science fiction vision to real-world value

keeping in mind customer retention, partnering with organizations that already customer acquisition, and market growth have cognitive technology product offer- opportunities. Cognitive technologies are ings, as well as with consultants or systems not the solution to every problem. Our integrators who can help incorporate cogni- research on how companies are putting tive technologies for business transforma- cognitive technologies to work has revealed tion and/or product innovation. a framework that can help organizations assess their own opportunities for deploy- 6. Pursue POCs: Form a cognitive technol- ing these technologies. We suggest organi- ogy innovation team. Begin to explore and zations look across their business processes, adopt cognitive technologies through the their products, and their markets to exam- design of proof-of-concept projects. ine where the use of cognitive technologies may be viable, where it could be valuable, 7. Expand where value creation is evident: and where it may even be vital. This “Three Double down, expand, and deploy where Vs” framework is summarized in figure 4. the value proposition of cognitive technolo- Organizations can use it to screen opportu- gies is becoming evident. nities for applying cognitive technologies.93 Cognitive technologies are delivering measurable value to enterprises across mul- 4. Build your cognitive capabilities: tiple industries. Now is the time to engage Determine which part of your enterprise these technologies and to frame how best your should start to build cognitive capabili- organization will capture the value these tech- ties, possibly by initially using a center of nologies are creating. A thorough evaluation of excellence model. your company’s position with regard to cognitive technologies can help put your organiza- 5. Leverage ecosystem partners: Discuss tion on the path to reap these concrete benefits. business ecosystem and platform partner- ship options. Both of these may involve

17 Cognitive technologies in the technology sector

Endnotes

1. The technology sector, or “tech sector,”is 2013, http://www.scientificamerican.com/ composed of companies that design, develop, article/will-ibms-watson-usher-in-cognitive- or manufacture electronic components, computing/, last accessed November 23, 2015. products and/or solutions—and is further 10. Eamonn Kelly et al., Business ecosystems classified across multiple sub-sectors, including come of age, Deloitte University Press, April computer hardware, IT services, networking, 15, 2015, http://dupress.com/articles/ Internet, semiconductor, and software. business-ecosystems-boundaries-business- 2. David Schatsky, Craig Muraskin, and Ragu trends/, last accessed November 23, 2015. Gurumurthy, Demystifying artificial intel- 11. AlchemyAPI, “AlchemyAPI Products,” ligence: What business leaders need to know http://www.alchemyapi.com/products/, about cognitive technologies, Deloitte University last accessed November 24, 2015. Press, November 4, 2014, http://dupress. com/articles/what-is-cognitive-technology/. 12. “The deep learning algorithm is a powerful This article is also an overview of cognitive general-purpose form of machine learning that technologies and their improving performance. is moving into the mainstream. It uses a variant of neural networks to perform high-level 3. Brandon Bailey, “Google, Facebook and abstractions such as voice or image pattern other tech companies race to develop artificial recognition. Pioneered in 2006 by Geoffrey intelligence,” San Jose Mercury News, April Hinton, a professor at the University of Toronto 23, 2014, http://www.mercurynews.com/ and researcher at Google, deep learning is business/ci_25627092/google-facebook- proving its metal across a diverse spectrum of and-other-tech-companies-race-develop/. challenges—from new drug development to 4. The M&A data is a technology sector analysis translating human conversations from English of activity in the United States and Canada to Chinese. Google is using deep learning with and does not include global M&A data. The its Android phone to recognize voice com- final 100 deals that make up the analysis mands and on its social network to identify include cross-sector (media and telecom- and tag images. Facebook is exploring deep munications) and cross-industry activity learning as a means to target ads and identify when appropriate in an effort to take into faces and objects. A company called DeepMind account cross-sector convergence issues, . . . recently announced a Neural Turing vertical market digitization trends, and broader Machine. This computing architecture utilizes a trends in sensors, actuators, and hardware. form of recurrent neural networks that can do 5. Deloitte analysis of M&A data gathered by end-to-end processing from a sensory percep- financial data provider S&P Capital IQ. tion data stream to interpretation and action. This system has been used to learn games 6. Michael E. Raynor and Mumtaz Ahmed, from scratch, and in some cases has demon- The Three Rules: How Exceptional Compa- strated better-than-human-level performance.” nies Think (New York: Penguin, 2013). Source: Bill Briggs and Marcus Shingles, 7. KurzweilAI, “IBM forms new Watson group for Exponentials, Deloitte University Press, cloud-delivered cognitive innovations, with $1 January 29, 2015, http://dupress.com/articles/ billion investment,” http://www.kurzweilai.net/ tech-trends-2015-exponential-technologies/. ibm-forms-new-watson-group-for-cloud-de- 13. Janet Wagner, “The rapid rise of deep learning livered-cognitive-innovations-with-1-billion- computer vision technology,” Programmable investment/, last accessed November 23, 2015. Web, June 6, 2015, http://www.program- 8. YouTube, “IBM Watson Group launch event mableweb.com/news/rapid-rise-deep- Jan. 9 in New York,” https://youtu.be/rB7Vk- learning-computer-vision-technology/analy- rUYCAg, last accessed October 2, 2015. sis/2015/06/19, last accessed October 20, 2015. 9. Larry Greenemeier, “Will IBM’s Watson 14. Charlierose.com, “Ginni Rometty, Chairman usher in a new era of cognitive comput- and CEO of IBM,” http://charlierose.com/ ing?” Scientific American, November 13, watch/60546698/, last accessed June 12, 2015.

18 From science fiction vision to real-world value

15. IBM, “IBM closes deal to acquire merge 29. IBM, “IBM Bluemix,” http://www.ibm.com/ healthcare,” http://www-03.ibm.com/ bluemix, last accessed October 13, 2015. press/us/en/pressrelease/47839.wss, 30. Gil Press, “IBM reinvents itself by establish- last accessed November 23, 2015. ing a frontier homestead for cognitive 16. Gadgets360, “IBM plays down earnings computing,” Forbes, October 27, 2014, http:// miss as part of evolution,” http://gadgets. www.forbes.com/sites/gilpress/2014/10/27/ ndtv.com/laptops/news/ibm-plays-down- ibm-reinvents-itself-by-establishing-a- earnings-miss-as-part-of-evolution-755236/, frontier-homestead-for-cognitive-computing/, last accessed November 24, 2015. last accessed October 13, 2015. 17. Peter H. Diamandis and Steven Kotler, 31. Intel, “Intel Real Sense SDK,” https:// Bold: How to Go Big, Create Wealth software.intel.com/en-us/intel-realsense- and Impact the World, (New York: sdk, last accessed November 23, 2015. Simon and Schuster, 2015), p. 82. 32. Jean Philippe Bouchard et al., “Worldwide 18. Robert McMillan, “IBM’s new unit bets on smartphone 2014–2018 forecast and analysis,” boom in artificial intelligence,” Wall Street IDC, March 2014, http://www.idc.com/getdoc. Journal, October 6, 2015, http://www.wsj. jsp?containerId=247140, “The pace of mobile com/articles/ibms-new-unit-bets-on-boom- adoption is growing significantly; smartphone in-artificial-intelligence-1444104061. shipments are expected to grow from 1.2 19. Alphabet.com, “G is for Google,” https:// billion in 2014 to 1.7 billion units by 2018—a abc.xyz/, last accessed October 19, 2015. CAGR of 11.5 percent during 2014-18.” 20. Briggs and Shingles, Exponentials. 33. Patrick Moorehead, “Perceptual computing: Critical to Intel’s future,” Forbes, 21. Salim Ismail, Exponential Organizations: Why September 20, 2012, http://www.forbes. New Organizations are Ten Times Better, Faster, com/sites/patrickmoorhead/2012/09/21/ and Cheaper than Yours (and What to Do about perceptual-computing-the-key-to-intels- It), (New York: Diversion Books, 2014), p. 18– future/, last accessed November 23, 2015. 19. Ismael mentions Google Ventures (which will continue as the venture capital business 34. Fool.com, “Why surging sales of hybrid devices unit of the Alphabet holding company), “an is great news for Microsoft and Intel,” http:// almost perfect exponential organization.” www.fool.com/investing/general/2015/09/17/ why-surging-sales-of-hybrid-devices-is-great- 22. Ibid, p. 18. news.aspx, last accessed October 13, 2015. 23. Exponential Organizations.com, “Top iPad is a trademark of Apple Inc., registered 100ExOs,” http://top100.exponentialorgs. in the United States and other countries. The com/, last accessed October 19, 2015. current report is an independent publication and has not been authorized, sponsored, 24. Deloitte analysis of M&A data gathered by or otherwise approved by Apple Inc. financial data provider S&P Capital IQ. 35. The Graphics Processor Unit (GPU) is “a single 25. QZ.com, “All of Google’s—er, Alphabet’s— chip processor with integrated transform, light- companies and products from A to Z,” http:// ing, triangle setup/clipping, and rendering en- qz.com/476460/here-are-all-the-alphabet- gines that is capable of processing a minimum formerly-google-companies-and-products- of 10 million polygons per second.” Source: from-a-to-z/, last accessed October 19, 2015. http://www.nvidia.com/object/gpu.html. 26. Deloitte analysis of M&A data gathered by 36. NVIDIA® CUDA® is a parallel computing financial data provider S&P Capital IQ. platform and programming model that 27. John Hagel, The power of platforms, Deloitte enables dramatic increases in computing University Press, April 15, 2015, http://dupress. performance by harnessing the power of com/articles/platform-strategy-new-level- the graphics processing unit (GPU). Source: business-trends/, last accessed October 9, 2015. https://developer.nvidia.com/about-cuda. 28. Paul Clemmons, Jason Geller, and Adam 37. Jeffrey Burt, “Nvidia enhances DIG- Messer, Cloud hits the enterprise, Deloitte ITS, CUDA software for deep learn- University Press, July 1 2012, http://dupress. ing,” eWeek, July 8, 2015, http://www. com/articles/cloud-hits-the-enterprise- theinquirer.net/inquirer/news/2416690/ the-scoop-on-software-as-a-service/, nvidia-updates-digits-and-cudnn-gpu- last accessed October 9, 2015. accelerated-deep-learning-software.

19 Cognitive technologies in the technology sector

38. NVidia Corp., “Q3 2015 NVidia article/2886132/machine-learning/ Corp conference call,” Thomson how-machine-learning-ate-microsoft. StreetEvents, August 6, 2015. html/, last accessed September 29, 2015. 39. CIO Review, “NVidia release next genera- 49. Multisided platforms (MSPs) are tech- tion Mobile Super Chip,” http://technology. nologies, products or services that create cioreview.com/news/nvidia-releases-next-gen- value primarily by enabling direct interactions eration-mobile-super-chip-nid-4601-cid-19. between two or more customer or participant html, last accessed November 19, 2015. groups. See Andrei Hagiu, “Strategic deci- 40. Nvidia, “NVIDIA launches Tegra X1 Mobile sions for multisided platforms,” MIT Super Chip,” http://nvidianews.nvidia.com/ Sloan Management Review, December 19, news/nvidia-launches-tegra-x1-mobile- 2013, http://sloanreview.mit.edu/article/ super-chip, last accessed October 19, 2015. strategic-decisions-for-multisided-platforms/. 41. Jagdish Rebelo, Enterprise cloud comput- 50. Microsoft, “Machine learning blog: Mi- ing: Future market size, growth and crosoft closes acquisition of Revolution competitive landscape, IHS Quarterly, May Analytics,” http://blogs.technet.com/b/ 20, 2014, http://blog.ihs.com/q12-enterprise- machinelearning/archive/2015/04/06/ cloud-computing-future-market-size- microsoft-closes-acquisition-of-revolution- growth-and-competitive-landscape. analytics.aspx/, last accessed October 19, 2015. 42. For a history of the development of AWS, 51. DeZyre.com, “R Programming Language— Infrastructure as a Service—and the cre- A powerful statistical programming tool ation of the cloud computing market—all written by statisticians for statisticians,” of which are the origins of the Amazon https://www.dezyre.com/article/-why-r- Machine Learning innovation at Amazon, programming-language-still-rules-data- see Brad Stone, “Chapter 7: A technology science/161, last accessed December 1, 2015. company, not a retailer,” The Everything Store: 52. Wall Street Journal, “Microsoft profit tops esti- Jeff Bezos and the Age of Amazon, (New mates,” http://www.wsj.com/articles/microsoft- York: Back Bay Books, 2013), p. 193. profit-climbs-1445545134?mod=djemCIO_h, 43. MobyLives, “Wall Street finally says maybe last accessed October 23, 2015. Amazon should, you know, make some money,” 53. Hewlett Packard, Inc. “HP Haven Big Data http://www.mhpbooks.com/wall-street-finally- Platform,” http://www8.hp.com/us/en/ says-maybe-amazon-should-you-know-make- software-solutions/big-data-platform-haven/ some-money/, last accessed December 1, 2015. index.html, last accessed September 29, 2015. 44. Amazon.com, “Amazon Machine Learn- 54. Hewlett Packard, Inc. “Haven Predictive ing,” http://aws.amazon.com/machine- Analytics,” http://www8.hp.com/us/en/ learning/, last accessed August 13, 2015. software-solutions/predictive-analytics-r- 45. Marketwatch.com, “Amazon Web Services An- big-data/, last accessed September 29, 2015. nounces Amazon Machine Learning,” http:// 55. Adrian Bridgewater, “HP Haven Predictive www.marketwatch.com/story/amazon-web- Analytics: Operationalizing large-scale services-announces-amazon-machine-learn- machine learning,” Computer Weekly, February ing-2015-04-09/, last accessed October 2, 2015. 2015, http://www.computerweekly.com/blogs/ 46. Fool.com, “The cloud battle heats up between cwdn/2015/02/hp-haven-predictive-analytics- Amazon.com Inc., Google Inc., Microsoft operationalising-large-scale-machine-learning. Corp, and IBM,” http://www.fool.com/ html/, last accessed September 29, 2015. investing/general/2015/04/16/the-cloud- 56. Ibid. As Adrian Bridgwater of Computer battle-heats-up-between-amazoncom-inc-go. Weekly also noted: “Now the global developer aspx/, last accessed August 8, 2015. community can employ R to scale to more 47. Wall Street Journal, “Amazon posts than a billion predictive records of data—and surprise profit, while sales surge,” http:// this is said to be ‘an order of magnitude www.wsj.com/articles/amazon-posts- improvement’ over traditional R-based surprise-profit-while-sales-surge- performance. This offering from HP also 1445544477?mod=djemCIO_h&alg=y, retains the consistency with R and enables data last accessed October 23, 2015. scientists to use their familiar R console and RStudio to work with Distributed R—and this 48. Mary Branscombe, “How machine learn- could indeed be important for R converts.” ing ate Microsoft,” Infoworld, February 19, 2015, http://www.infoworld.com/

20 From science fiction vision to real-world value

57. Techcrunch, “The next wave of enterprise 68. Steve Morgan, “Dell’s secret cybersecurity software powered by machine learning,” http:// IPO,” Forbes, October 14, 2015, http://www. techcrunch.com/2015/07/27/the-next-wave- forbes.com/sites/stevemorgan/2015/10/14/ of-enterprise-software-powered-by-machine- dells-secret-cybersecurity-ipo/print/, learning/, last accessed October 19, 2015. last accessed October 16, 2015. 58. Larry Dignan, “HP Enterprise: Can it 69. Markets and Markets, “Security analytics mar- overcome layoffs, split growing pains, and ket by application—global forecast to 2020,” FUD,” ZDNet, September 16, 2015, http:// http://www.marketsandmarkets.com/Market- www.zdnet.com/article/hp-enterprise-can- Reports/security-analytics-market-1026. it-overcome-layoff-split-growing-pains- html, last accessed November 19, 2015. and-fud/, last accessed October 19, 2015. 70. Lawrence Orans and Peter Firstbrook, Magic 59. Hewlett Packard, Inc., “Operations Analytics,” quadrant for secure web gateways, Gartner, http://www8.hp.com/us/en/software-solutions/ Inc., May 28, 2015. Gartner does not endorse operations-analytics-operations-analysis/, any vendor, product or service depicted in last accessed December 3, 2015. its research publications, and does not advise 60. Charles Babcock, “SalesforceIQ gathers technology users to select only those vendors intelligence for customers big and small,” with the highest ratings or other designation. Information Week, September 17, 2015, Gartner research publications consist of the http://www.informationweek.com/cloud/ opinions of Gartner’s research organization software-as-a-service/salesforceiq-gathers- and should not be construed as statements intelligence-for-customers-big-and-small/, of fact. Gartner disclaims all warranties, last accessed November 23, 2015. expressed or implied, with respect to this research, including any warranties of mer- 61. Pegasystems, “Pegasystems transforms chantability or fitness for a particular purpose. customer service with advanced social collaboration and predictive analytics,” http:// 71. Venturebeat, “Microsoft acquires people www.pega.com/about/news/press-releases/ analytics startup VoloMetrix, will integrate pegasystems-transforms-customer-service-ad- organizational tech into Office 365,” http:// vanced-social-collaboration-and, last accessed venturebeat.com/2015/09/03/microsoft- November 23, 2015; Oracle, “Oracle social acquires-people-analytics-startup-volometrix- cloud,” http://www.oracle.com/us/solutions/ will-integrate-organizational-tech-into- social/overview/index.html/, last accessed office-365/, last accessed November 23, 2014. November 23, 2015; Pegasystems, “Marketing 72. Steve Blank, The Four Steps to the Epiphany: analytics,” http://www.pega.com/marketing- Successful Strategies for Products that Win analytics/, last accessed December 1, 2015. (San Mateo: Café Press, 2005), p. 133. 62. Markets and Markets, “Social Customer 73. Thomas Steenburgh, Jill Avery, and Nadeem Relationship Management (CRM) Market Dahod, “Hubspot: Inbound marketing by Solution—Global Forecasts to 2019,” and Web 2.0,” Harvard Business School http://www.marketsandmarkets.com/ Publishing, January 24, 2011; interview Market-Reports/social-customer-relationship- with Michael A. Cusumano, Sloan School of management-market-124012616.html, Management, Sloan Management Review last accessed December 11, 2015. (SRM) professor in management. 63. CRMSearch, “Cloud computing CRM,” 74. Nextbigfuture, “IBM wants to commercialize http://www.crmsearch.com/cloud.php/, silicon brains with deep learning for smarter last accessed November 23, 2015. phones,” http://nextbigfuture.com/2015/10/ 64. Cisco Systems, Inc. “Introducing Cisco ibm-wants-to-commercialize-silicon. Cognitive Threat Analytics,” http://blogs.cisco. html, last accessed November 23, 2015. com/security/introducing-cisco-cognitive- 75. Agam Shah, “Intel’s 18-core Xeon chips threat-analytics, last accessed August 28, 2015. tuned for machine learning, analytics,” PC 65. Ibid. World, May 5, 2015, www.pcworld.com/ article/2918852/intels-18core-xeon-chips- 66. Ibid. tuned-for-machine-learning-analytics. 67. Cisco Systems, “Cisco Cognitive Threat Analyt- html, last accessed December 1, 2015. ics,” http://www.cisco.com/c/en/us/solutions/ 76. Aaron Tilley, “Intel and Qualcomm battle enterprise-networks/cognitive-threat-analytics/ to bring computer vision into phones and index.html, last accessed August 28, 2015. drones,” Forbes, August 26, 2015, http://www. forbes.com/sites/aarontilley/2015/08/26/

21 Cognitive technologies in the technology sector

intel-and-qualcomm-battle-to-bring- 87. George Collins and David Sisk, API computer-vision-into-phones-and-drones/, economy: From systems to business services, last accessed November 23, 2015. Deloitte University Press, January 29, 2015, 77. GSMA Mobile Economy, “The mo- http://dupress.com/articles/tech-trends- bile economy 2015,” http://www. 2015-what-is-api-economy/; Thomas H. gsmamobileeconomy.com/GSMA_Glob- Davenport and Bala Iyer, “Move beyond al_Mobile_Economy_Report_2015.pdf/, enterprise IT to an API strategy,” Harvard last accessed November 16, 2015. Business Review, August 6, 2013, https:// hbr.org/2013/08/move-beyond-enterprise- 78. CIO Journal, “Connected device data an it-to-a/, last accessed November 23, 2015. enterprise windfall,” http://deloitte.wsj. com/cio/2015/10/05/connected-device- 88. Cade Metz, “Google’s open source AI Tensor- data-an-enterprise-windfall/, last flow signals fast changing hardware world,” accessed December 1, 2015. Wired, November 10, 2015, http://www. wired.com/2015/11/googles-open-source-ai- 79. Darrell M. West, The state of the mobile tensorflow-signals-fast-changing-hardware- economy, 2014: Its impact and future, The world/, last accessed November 23, 2015. Brookings Institution, September 2014, http:// www.brookings.edu/~/media/research/files/ 89. ZDNET.com, “IBM’s SystemML machine papers/2014/09/10-state-of-mobile-economy- learning system becomes Apache Incuba- west/state-of-mobile-economy_v13.pdf/. tor project,” http://www.zdnet.com/ article/ibms-systemml-machine-learning- 80. Scott Corwin, Joe Vitale, Eamonn Kelly, and system-becomes-apache-incubator-project/, Elizabeth Cathles, The future of mobility: last accessed December 1, 2015. How transportation technology and social trends are creating a new business ecosys- 90. Quentin Hardy, “Facebook offers artificial tem, Deloitte University Press, September intelligence tech to open source group,” New 24, 2015, http://dupress.com/articles/ York Times, January 16, 2015, http://bits.blogs. future-of-mobility-transportation-technology/. nytimes.com/2015/01/16/facebook-offers- artificial-intelligence-tech-to-open-source- 81. GSMA Mobile Economy, “The mo- group/, last accessed November 23, 2015. bile economy 2015,” http://www. gsmamobileeconomy.com/GSMA_Glob- 91. Jacob Bruun-Jensen and John Hagel III, al_Mobile_Economy_Report_2015.pdf/, Minimum viable transformation, Deloitte last accessed November 16, 2015. University Press, April 15, 2015, http://dupress. com/articles/minimum-viable-business- 82. Amir Mizroch, “Artificial intelligence model-transformation-business-trends/; experts are in high demand,” Wall Street John Hagel, John Seely Brown, and Duleesha Journal, May 1, 2015, http://www.wsj.com/ Kulasooriya, Scaling edges: A pragmatic articles/artificial-intelligence-experts-are- pathway to broad internal change, Deloitte in-high-demand-1430472782?alg=y. University Press, http://www2.deloitte.com/ 83. Interview with Daniel Byler, Deloitte us/en/pages/center-for-the-edge/articles/ data scientist and manager, strategic scaling-edges-methodology-to-create-growth. planning, October 16, 2015. html/, last accessed November 24, 2015. 84. Amazon Web Services, “Welcome to 92. Deloitte, “Exponentials: Brace yourself for Amazon Machine Learning,” http://docs. the next wave of disruption,” http://www2. aws.amazon.com/machine-learning/ deloitte.com/us/en/pages/dbriefs-webcasts/ latest/APIReference/Welcome.html/, events/august/2015/dbriefs-exponentials- last accessed December 1, 2015. brace-yourself-for-the-next-wave-of-disruption.html/, last accessed December 1, 2015; 85. Microsoft Azure, “Machine Learning interview with Andrew Blau, director, Deloitte APIs,” https://azure.microsoft.com/ US, Enterprise Risk Services, October 16, 2015. en-us/services/machine-learning/api/, last accessed December 1, 2015. 93. David Schatsky, Craig Muraskin, and Ragu Gurumurthy, Cognitive 86. IBM, “IBM unveils matchmaking technology technologies: The real opportunities for to navigate API economy,” http://www-03. business, Deloitte University Press, ibm.com/press/us/en/pressrelease/48026. January 26, 2015, http://dupress.com/articles/ wss, last accessed December 1, 2015. cognitive-technologies-business-applications/.

22 From science fiction vision to real-world value

Acknowledgements

The authors would like to thank Elina Ianchulev and Karthik Ramachandran for their contributions. In addition, many Deloitte colleagues contributed with their time and industry-specific insight: A special thanks to Andrew Blau and Daniel Byler.

Significant research contributions were provided by Deepan Kumar Pathy, Negina Rood, and Prathima Shetty. Additional research support was provided by Gaurav Khetan, Alok Ranjan, Shankar Lakshman, Chandrakala Veerabomma, Amit Kumar Sudrania, Divya Ravichandran, Abhimanyu Singh Yadav, Vasudev Neehar Chinnam, and Gitanjali Venkatesh.

For their contributions to the development and review of this article, the authors extend their thanks to John Shumadine, Dr. Preeta Banerjee, Anupam Narula, Sandy Rogers, Junko Kaji, Jon Warshawsky, William Eggers, Kristin Martin, and Lynette DeWitt. Finally, for their integrated media production expertise, special thanks to Troy Bishop and Alok Nookraj Pepakayala.

23 Cognitive technologies in the technology sector

Contacts

Daniel Pereira Manager Market Insights Deloitte Services LP +1 617 437 3429 [email protected]

David Schatsky Senior manager Deloitte Innovation Deloitte LLP +1 646 582 5209 [email protected]

Paul Sallomi Vice chairman Global technology, media, and telecommunications industry leader Global/US technology sector leader Partner Deloitte Tax LLP +1 408 704 4100 [email protected]

Bob Dalton Analytics alliances leader Principal Deloitte Consulting LLP +1 404 631 3939 [email protected]

Follow @DU_Press Sign up for Deloitte University Press updates at DUPress.com.

About Deloitte University Press Deloitte University Press publishes original articles, reports and periodicals that provide insights for businesses, the public sector and NGOs. Our goal is to draw upon research and experience from throughout our professional services organization, and that of coauthors in academia and business, to advance the conversation on a broad spectrum of topics of interest to executives and government leaders. Deloitte University Press is an imprint of Deloitte Development LLC.

About this publication This publication contains general information only, and none of Deloitte Touche Tohmatsu Limited, its member firms, or their related entities (collectively the “Deloitte Network”) is, by means of this publication, rendering professional advice or services. Before making any decision or taking any action that may affect your finances or your business, you should consult a qualified professional adviser. No entity in the Deloitte Network shall be responsible for any loss whatsoever sustained by any person who relies on this publication.

About Deloitte Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients. Please see www.deloitte.com/about for a more detailed description of DTTL and its member firms. Deloitte provides audit, tax, consulting, and financial advisory services to public and private clients spanning multiple industries. With a globally connected network of member firms in more than 150 countries and territories, Deloitte brings world-class capabilities and high-quality service to clients, delivering the insights they need to address their most complex business challenges. Deloitte’s more than 200,000 professionals are committed to becoming the standard of excellence. © 2015. For information, contact Deloitte Touche Tohmatsu Limited.