This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Hype Cycle for Emerging Technologies, 2017
Published: 21 July 2017 ID: G00314560
Analyst(s): Mike J. Walker
Our 2017 Hype Cycle reveals three distinct technology trends that
profoundly create new experiences, with unrivaled intelligence, and offer
platforms that propel organizations to connect with new business
ecosystems in order to become competitive over the next five to 10 years.
Table of Contents
Analysis..................................................................................................................................................3
What You Need to Know.................................................................................................................. 3
The Hype Cycle................................................................................................................................ 3
Megatrends Fueled by Emerging Technologies........................................................................... 4
Major Hype Cycle Changes........................................................................................................ 5
The Priority Matrix.............................................................................................................................7
Off the Hype Cycle......................................................................................................................... 10
On the Rise.................................................................................................................................... 10
Smart Dust............................................................................................................................... 10
4D Printing............................................................................................................................... 11
Artificial General Intelligence......................................................................................................13
Deep Reinforcement Learning...................................................................................................15
Neuromorphic Hardware...........................................................................................................16
Human Augmentation...............................................................................................................17
5G............................................................................................................................................19
Serverless PaaS....................................................................................................................... 22
Digital Twin............................................................................................................................... 23
Quantum Computing................................................................................................................ 25
Volumetric Displays...................................................................................................................27
Brain-Computer Interface......................................................................................................... 28
Conversational User Interfaces................................................................................................. 30
Smart Workspace.....................................................................................................................32
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
At the Peak.....................................................................................................................................34
Augmented Data Discovery...................................................................................................... 34
Edge Computing...................................................................................................................... 36
Smart Robots........................................................................................................................... 37
IoT Platform..............................................................................................................................39
Virtual Assistants...................................................................................................................... 41
Connected Home..................................................................................................................... 43
Deep Learning.......................................................................................................................... 45
Machine Learning..................................................................................................................... 47
Autonomous Vehicles............................................................................................................... 49
Nanotube Electronics............................................................................................................... 50
Cognitive Computing................................................................................................................ 52
Blockchain................................................................................................................................53
Commercial UAVs (Drones).......................................................................................................55
Sliding Into the Trough.................................................................................................................... 57
Cognitive Expert Advisors......................................................................................................... 57
Enterprise Taxonomy and Ontology Management..................................................................... 58
Software-Defined Security........................................................................................................ 60
Augmented Reality................................................................................................................... 61
Climbing the Slope......................................................................................................................... 63
Virtual Reality............................................................................................................................63
Appendixes.................................................................................................................................... 64
Hype Cycle Phases, Benefit Ratings and Maturity Levels.......................................................... 66
Gartner Recommended Reading.......................................................................................................... 67
List of Tables
Table 1. Hype Cycle Phases................................................................................................................. 66
Table 2. Benefit Ratings........................................................................................................................66
Table 3. Maturity Levels........................................................................................................................ 67
List of Figures
Figure 1. How Emerging Technology Trends Move Along the Hype Cycle............................................... 5
Figure 2. Hype Cycle for Emerging Technologies, 2017.......................................................................... 7
Figure 3. Priority Matrix for Emerging Technologies, 2017....................................................................... 9
Page 2 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Figure 4. Hype Cycle for Emerging Technologies, 2016........................................................................ 65
Analysis
What You Need to Know
Organizations will continue to be faced with rapidly accelerating technology innovation that will
profoundly impact the way they deal with their workforces, customers and partners. In particular,
four emerging technologies are poised to be the highest priority: ecosystem-expanding
technologies such as Blockchain; Brain-Computer Interface, which further entrenches humans into
technology; Commercial UAVs (Drones), which challenge how goods and services are delivered;
and intelligent API-driven Software-Defined Security, which enables a more secure digital world.
To survive and thrive in the digital economy, enterprise architecture (EA) and technology innovation
leaders who are focused on mastering emerging and strategic trends must continue to work with
their CIOs and business leaders to look for emerging technologies that can help create competitive
advantage, generate value, overcome legal and regulatory hurdles, reduce operating costs, and
enable transformational business models. This Hype Cycle provides a high-level view of important
emerging trends that organizations must track, as well as the specific technologies that must be
monitored.
This year, three trends stand out at a high level:
■ AI Everywhere
■ Transparently Immersive Experiences
■ Digital Platforms
Enterprise architects who are focused on technology innovation must evaluate these high-level
trends and the featured technologies, as well as the potential impact (value and risk) on their
businesses. In addition to the potential impact on businesses, these trends provide a significant
opportunity for EA leaders to help senior business and IT leaders respond to the digital business
opportunities and threats by creating signature-ready actionable and diagnostic deliverables that
guide investment decisions.
The Hype Cycle
The Hype Cycle for Emerging Technologies is unique among most Gartner Hype Cycles because it
distills insights from more than 2,000 Gartner technologies into a succinct set of must-know
emerging technologies and trends. This Hype Cycle specifically focuses on the set of technologies
that is showing promise in delivering a high degree of competitive advantage over the next five to 10
years.
Gartner, Inc. | G00314560 Page 3 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Megatrends Fueled by Emerging Technologies
The emerging technologies on the 2017 Hype Cycle reveal three distinct megatrends that
profoundly create new experiences, with unrivaled intelligence, and offer platforms that allow
organizations to connect with new business ecosystems. Those three megatrends are:
■ AI Everywhere: Artificial intelligence (AI) technologies will be the most disruptive class of
technologies over the next 10 years due to radical computational power, near-endless amounts
of data, and unprecedented advances in deep neural networks; these will enable organizations
with AI technologies to harness data in order to adapt to new situations and solve problems that
no one has ever encountered previously.
Enterprises that are seeking leverage in this theme should consider the following technologies:
Deep Learning, Deep Reinforcement Learning, Artificial General Intelligence, Autonomous
Vehicles, Cognitive Computing, Commercial UAVs (Drones), Conversational User Interfaces,
Enterprise Taxonomy and Ontology Management, Machine Learning, Smart Dust, Smart
Robots, and Smart Workspace.
■ Transparently Immersive Experiences: Technology has and will continue to become more
human-centric to the point where it will introduce transparency between people, businesses
and things. This relationship will become much more entwined as the evolution of technology
becomes more adaptive, contextual and fluid within the workplace, at home, and in interacting
with businesses and other people.
Critical technologies to be considered include: 4D Printing, Augmented Reality, Brain-Computer
Interface, Connected Home, Human Augmentation, Nanotube Electronics, Virtual Reality and
Volumetric Displays.
■ Digital Platforms: Emerging technologies require revolutionizing the enabling foundations that
provide the volume of data needed, advanced compute power, and ubiquity-enabling
ecosystems. The shift from compartmentalized technical infrastructure to ecosystem-enabling
platforms is laying the foundations for entirely new business models that are forming the bridge
between humans and technology. Within these dynamic ecosystems, organizations must
proactively understand and redefine their strategy to create platform-based business models,
and to exploit internal and external algorithms in order to generate value.
Key platform-enabling technologies to track include: 5G, Digital Twin, Edge Computing,
Blockchain, IoT Platform, Neuromorphic Hardware, Quantum Computing, Serverless PaaS and
Software-Defined Security.
When we view these themes in aggregate, we can see how the human-centric enabling
technologies within Transparently Immersive Experiences (such as Smart Workspace, Connected
Home, Augmented Reality, Virtual Reality and the growing Brain-Computer Interface) are becoming
the edge technologies that are pulling the other trends along the Hype Cycle (see Figure 1 [visible
only in the noninteractive version of this research]).
Page 4 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Figure 1. How Emerging Technology Trends Move Along the Hype Cycle
Source: Gartner (July 2017)
AI Everywhere emerging technologies are moving rapidly through the Hype Cycle. These
technologies are just crossing the peak, which shows that they are a key enabler of technologies
that create transparent and immersive experiences.
Finally, Digital Platforms are rapidly moving up the Hype Cycle, illustrating the new IT realities that
are possible by providing the underlining platforms that will fuel the future. Technologies like
Quantum Computing and Blockchain are poised to create the most transformative and dramatic
impacts in the next five to 10 years.
These megatrends illustrate that the more organizations are able to make technology an integral
part of employees', partners' and customers' experiences, the more they will be able to connect
their ecosystems to platforms in new and dynamic ways.
Major Hype Cycle Changes
Understanding the new emerging technologies that are being introduced on the Hype Cycle for the
first time in 2017 provides enterprise architects with the leading indicators of what technology
trends will be strategic in the coming years.
Gartner, Inc. | G00314560 Page 5 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Also, understanding the eight new technologies below will support EA and technology innovation
leaders in building on key themes of AI Everywhere, Transparently Immersive Experiences and
Digital Platforms:
■ 5G
■ Artificial General Intelligence
■ Deep Learning
■ Deep Reinforcement Learning
■ Digital Twin
■ Edge Computing
■ Serverless PaaS
■ Cognitive Computing
In addition, EA and technology innovation leaders should evaluate the technologies that have
moved significantly along the Hype Cycle since 2016:
1. Blockchain: This concept is gaining traction because it holds the promise of transforming
industry operating models. Multiple business use cases are yet to be proved, and it is likely that
— while the hype is around the financial services industry — manufacturing, government,
healthcare and education will see more rapid evolution and acceptance.
2. Commercial UAVs (Drones): Major advances in AI hardware, miniaturization of computing
power, and deep-learning algorithms that continue to be more useful are enabling drones to be
used in industries like financial services, manufacturing, retail and automotive.
3. Software-Defined Security (SDSec): Security vendors continue to shift more of the policy
management out of individual hardware elements and into a software-based management plane
for flexibility in specifying security policy, regardless of location. As a result, SDSec will bring
speed and agility to the enforcement of security policy, regardless of the location of the user, the
information or the workload.
4. Brain-Computer Interface: As wearable technology advances to become miniaturized and
more powerful, and also becomes pervasive in the commonplace, applications will benefit from
hybrid techniques that combine brain, gaze and muscle tracking to offer hands-free interaction.
Over the next five years, as virtual reality (VR) hardware develops, it is likely that noninvasive
versions of this technology will be included in VR headset designs. Brain-Computer Interface
has not only shown major progress, but also increased its impact in a transformational way.
Page 6 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Figure 2. Hype Cycle for Emerging Technologies, 2017
Source: Gartner (July 2017)
The Priority Matrix
Emerging technologies are disruptive by nature, but the competitive advantage they provide is not
yet well-known or proved in the market. However, most will take more than five to 10 years to reach
the Plateau of Productivity. These examples illustrate the impact of key emerging technologies in the
near term and the longer term.
Two to five years to mainstream adoption: The AI Everywhere trend is here, and the enabling,
emerging technologies, such as Machine Learning, are already providing widespread and significant
benefits, while Deep Learning and Commercial UAVs (Drones) are enabling Machine Learning
algorithms for the masses. The full list of emerging technologies that are two to five years to
mainstream adoption is:
■ Augmented Data Discovery
Gartner, Inc. | G00314560 Page 7 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
■ Cognitive Expert Advisors
■ Deep Learning
■ Edge Computing
■ Commercial UAVs (Drones)
■ IoT Platform
■ Machine Learning
■ Serverless PaaS
■ Software-Defined Security
■ Virtual Reality
Five to 10 years to mainstream adoption: Technologies indicate that the digital platforms are in
full force. SDSec brings speed and agility to the enforcement of security policy, regardless of the
location of the user, the information or the workload. Virtual Assistants provide unobtrusive,
ubiquitous and contextually aware advisor-based solutions, while Blockchain will expand distributed
ledger concepts that promise to transform industry operating models. The full list of emerging
technologies that are five to 10 years to mainstream adoption is:
■ 5G
■ Deep Reinforcement Learning
■ Digital Twin
■ Augmented Reality
■ Blockchain
■ Cognitive Computing
■ Connected Home
■ Conversational User Interfaces
■ Enterprise Taxonomy and Ontology Management
■ Nanotube Electronics
■ Neuromorphic Hardware
■ Smart Robots
■ Smart Workspace
■ Virtual Assistants
More than 10 years to mainstream adoption: Quantum Computing provides unprecedented
compute power. Artificial General Intelligence will drive ubiquity and AI as a service, which will
Page 8 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
ultimately be key factors in the convergence between AI Everywhere, Transparently Immersive
Experiences and Digital Platforms. The full list of emerging technologies that are more than 10 years
to mainstream adoption is:
■ 4D Printing
■ Artificial General Intelligence
■ Autonomous Vehicles
■ Brain-Computer Interface
■ Human Augmentation
■ Quantum Computing
■ Smart Dust
■ Volumetric Displays
Figure 3. Priority Matrix for Emerging Technologies, 2017
benefit years to mainstream adoption
less than 2 years 2 to 5 years 5 to 10 years more than 10 years
transformational Augmented Data
Discovery
Cognitive Expert
Advisors
Deep Learning
Edge Computing
IoT Platform
Machine Learning
Software-Defined
Security
Blockchain
Cognitive Computing
Conversational User
Interfaces
Deep Reinforcement
Learning
Digital Twin
Nanotube Electronics
Smart Workspace
Virtual Assistants
4D Printing
Artificial General
Intelligence
Autonomous Vehicles
Brain-Computer Interface
Human Augmentation
Smart Dust
high Commercial UAVs
(Drones)
5G
Augmented Reality
Connected Home
Neuromorphic Hardware
Smart Robots
Quantum Computing
moderate Serverless PaaS
Virtual Reality
Enterprise Taxonomy and
Ontology Management
Volumetric Displays
low
As of July 2017 © 2017 Gartner, Inc.
Source: Gartner (July 2017)
Gartner, Inc. | G00314560 Page 9 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Off the Hype Cycle
Because this Hype Cycle pulls from such a broad spectrum of topics, many technologies are
featured in a specific year because of their relative visibility, but are not tracked over a longer period
of time. This is not intended to imply that they are unimportant — quite the opposite. In many cases,
these technologies are no longer "emerging," but rather are becoming more integral to business and
IT (such as big data and cloud computing). In other cases, technologies have been removed from
the Hype Cycle in order to highlight other new emerging technologies.
Technology planners can refer to Gartner's broader collection of Hype Cycles for items of ongoing
interest. Some of the technologies that appeared in the "Hype Cycle for Emerging Technologies,
2016," but do not appear in this year's report, are:
■
■ Affective Computing
■ Context Brokering
■ Gesture Control Devices
■ Data Broker PaaS (dbrPaaS)
■ Micro Data Centers
■ Natural-Language Question Answering
■ Personal Analytics
■ Smart Data Discovery
■ Virtual Personal Assistants
On the Rise
Smart Dust
Analysis By: Ganesh Ramamoorthy
Definition: Smart dust refers to motes, which are tiny wireless micro-electromechanical systems
(MEMS), robots or other devices that can detect everything from light, temperature and pressure to
vibration, magnetism and chemical composition. They run on a wireless computer network and are
distributed over an area to perform tasks, usually sensing through RFID. As they do not use large
antennas, these systems have ranges measured in just a few millimeters.
Position and Adoption Speed Justification: At present, much of the activity surrounding smart
dust is concentrated in research laboratories, such as the . Defense Advanced Research
Projects Agency (DARPA)-funded project at the Robotics Research Laboratory at the University of
Southern California and JLH Labs, and more recently the University of Stuttgart, have developed a
new type of "smart dust" miniature camera smaller than the size of a grain of sand. The main
Page 10 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
purpose of the research is to make motes as small as possible, which involves both evolutionary
and revolutionary advances in miniaturization, integration and energy management. They also aim to
make motes available at as low a price as possible. Because a complete sensor/communication
system integrated into a cubic-millimeter package is still a long way off, we have yet to see major
commercial applications for smart dust. However, some reasonably small motes are commercially
available for building controls, industrial monitoring and security applications. Recently, Amphenol
Advanced Sensors announced the availability of a smart dust sensor designed to detect
particulates that decrease air quality. Given its wide range of potential applications and benefits, this
technology will, we believe, have a transformative effect on all areas of business and on people's
lives in general. However, due to the lack of any major activity in terms of commercial
implementations, smart dust remains in the same position.
User Advice: Smart dust that is available "off the shelf" can be configured with sensors that detect
and measure a variety of properties, such as temperature, barometric pressure, humidity, light
intensity, acceleration, vibration, magnetism, acoustic level and location (using GPS). The
combination of these capabilities in a well-designed sensor network could create opportunities to
deliver numerous services.
Business Impact: The potential benefits of smart dust are compelling and transformational. Given
the embryonic stage of this technology's development, vendors should stake their claims via patent
development for commercial applications, direct funding for research projects or equity funding for
companies engaged in R&D. Smart dust will transform the way humans interact with their
surroundings and create new ways for businesses to deliver services, while reducing costs in the
process. This will have wide-ranging implications for businesses' technological, social, economic
and legal practices across the globe.
Benefit Rating: Transformational
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Sample Vendors: Amphenol Advanced Sensors; Linear Technology; MEMSIC; Millennial Net; Moog
4D Printing
Analysis By: Michael Shanler; Miriam Burt
Definition: Four dimensional printing (4DP) is a technique where the materials are encoded with a
dynamic capability — either function, confirmation or properties — that can change via the
application of chemicals, electronics, particulates or nanomaterials. The printing technology has
extra functionality to sequence, mix and place specific materials that will have a calculated effect.
Position and Adoption Speed Justification: 4DP is an emerging technology that remains in the
embryonic stage, with more lab research and development continuing in the past 12 months. This
technology aims to add another dimension to the 3D printing process by creating an object
designed to change shape after it leaves the print bed, with most models relying on hydrogels to
Gartner, Inc. | G00314560 Page 11 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
execute the process. The concept of this "shape shifting" technology was triggered a few years ago
and is being developed through collaborative efforts between academia and technology firms.
While 4DP is on the radar for three-dimensional printing (3DP) technologists, smart materials have
actually been around for several decades. Recent scientific advancements in biology, chemistry,
electronics and 3D printing will accelerate the discipline. Over the next few years, 4DP research will
generate interest and hype.
Challenges persist with bringing precision to objects' transformations after they've been printed.
Material science research for 3DP is still an underserved market. Software is still a niche for both
nanoscale and human-scale programmable materials with self-assembly characteristics. Modeling
the geometries, determining interactions for changing states and calculating the energy (from heat,
shaking, pneumatics, gravity, magnetics and so on) that impacts materials is no easy task.
Engineering software vendors are just beginning to get interested.
In 2017, some exciting new advances in the newest frontier of using 4DP to grow tissues and
organs in a laboratory setting have pushed this technology up the hype curve. These include the
Harvard team's method to print transformable tissue engineering scaffolds that can support cell
growth, as well as researchers from the Wake Forest Institute of Regenerative Medicine printing 3D-
printed structures made of living cells that could replace human tissue, where the scaffolds change
shape over time. Also, NASA engineers have leveraged 4DP to print "space chain mail." The
Singapore Centre for 3D Printing and the Swiss Federal Institute of Technology in Zurich. Research
has been put into the public domain that addresses the durability of 4D printed parts and its
predictability as it relates to load-bearing of 4DP designs. It will still be over 10 years before this
technology becomes adopted as mainstream.
User Advice: 4DP offers not only tremendous opportunities for engineers, but also for designers, as
many new applications will arise. Smart materials will solve design/engineering problems, which
often arise from the limitations of current materials.
Business and R&D IT leaders with science, technology and engineering responsibilities for new
product innovation should explore the business and technical opportunities for 4D printing, and
begin to educate peers on how 4D printing can add new functions. Building an internal 4D
capability will present significant computer, scientific and engineering hurdles. Focus on strategic
partnerships to advance the technique and develop proofs of concept to build the capabilities to run
experiments and manage the entire laboratory infrastructure. The in-silico requirements can be
shared. More engineering and modeling software vendors, academic laboratories and 3DP vendors
will need to be included for sharing technical research. There are also opportunities for engaging via
open innovation or consortium approaches.
R&D groups will need to focus on the evolving intellectual property landscape. Material science is a
complicated space, and there are an immense number of scientific and formulation-based patents
that may impact business cases. Explore relationships to further improve 4DP processes through
R&D partnerships with material companies to develop and improve specifications for 4D-suitable
materials.
Page 12 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Business Impact: 4DP is an opportunity to create future technology-based products that could
disrupt your industry. Shape-shifting materials have already been leveraged in the automotive,
aerospace, defense and medical industries. Dynamic and self-assembling materials have already
begun to disrupt the way engineers think about designing components and delivering value.
Initially, the examples of "what's possible" will be technology-focused, but will have unclear revenue
impacts. Shape-shifted materials that can reduce the drag coefficient of an airplane or vehicle
during different environments might help optimize efficiency. The sole of an adaptive running shoe
may adjust to wet versus dry pavements and improve grip. A self-assembling medical stent may
reduce surgery times and improve patient outcomes. Implants will be able to change shape once
they come into contact with body heat to conform with wound areas and lead to better surgical
outcomes. A dynamic valve in an irrigation system could improve irrigation on a farm. A roof on a
house could change form to facilitate draining, and walls could increase or decrease in thickness
during the winter or summer to improve insulation values.
The business impacts for 4D are still murky, and most will be determined after 4D technology has
been refined and scaled into businesses. Until then, don't be fooled by the anticipated hype,
because the technology is still in its infancy. However, now is the time to evaluate whether it is worth
exploring the technology to build into future product and service roadmaps.
Benefit Rating: Transformational
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Sample Vendors: Autodesk; Geosyntec Consultants; MIT; Stratasys
Recommended Reading: "Maverick* Research: Make Profits by Preparing Your Business for
Global Climate Change"
"Lessons From Leaders: Insights on High Performance From 10 Years of the Supply Chain Top 25"
"What 3D Printing Means for Your Supply Chain"
Artificial General Intelligence
Analysis By: Tom Austin
Definition: Artificial general intelligence (AGI) — aka "strong AI" or "general-purpose machine
intelligence" — would handle a very broad range of use cases if it existed. It does not. Special-
purpose AI ("weak AI") is real and powerful, but limited to specific, narrower use cases. AGI exists
only in science fiction and "what if" discussions. AI technologies do not deliver AGI. Despite
appearing to have humanlike learning, reasoning, adapting and understanding, they lack common
sense, intelligence and extensive methods for self-maintenance or reproduction.
Gartner, Inc. | G00314560 Page 13 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Position and Adoption Speed Justification: Progress on AI has been limited to "weak AI."
Position and adoption speed for AGI remain unchanged year over year. (In 2016, we labelled it
"general-purpose machine intelligence." We changed it to "AGI" in 2017 to better reflect
marketplace term popularity and usage.)
Today's AI technology cannot be proven to possess the equivalent of human intelligence (the lack of
an agreed-to test is itself a problem). It may be possible to build a machine that approximates
human cognitive capabilities, but we are likely decades away from having completed the necessary
research and engineering.
AGI ("strong AI") is often entangled in cognitive computing discussions. Cognitive computing means
different things to different people: a set of AI capabilities, a specialized type of hardware (as in
neuromorphic or other highly parallel, short propagation path processors), or the use of information
and communication technology (ICT) to enhance human cognition. This latter definition is what
Gartner prefers for the term "cognitive computing."
User Advice: Focus on business results enabled by applications that exploit special-purpose
(narrow use case) AI technologies, both leading-edge and older AI technologies.
The leading edge of AI is "amazing innovations," including deep-learning tools and related natural-
language processing (NLP) capabilities. They do what we thought technology couldn't do. They are
typically research-grade tooling, still emerging from research labs, undergoing turbulent changes in
direction and not fully understood in terms of engineering principles. Over time, we learn their limits
and develop workable engineering guidelines. As the amazement wears off and ennui sets in, we
treat them as "aging innovations."
Look for business results enabled by applications exploiting either aging innovations (including
expert systems and other symbolic AI approaches, as well as simpler forms of machine learning) or
amazing innovations (typically more powerful and less well understood technologies) — or both.
Examples include autonomous transport, smart advisors and virtual assistants for customers
(VCAs), employees and individuals, focused on various missions (., wealth management) and
responsibilities (., sales or budget management). Most exploit a mix of amazing and aging
innovations.
Special-purpose AI will have a huge and disruptive impact on business and personal life.
End-user organizations should ignore AGI until such time that AGI researchers and advocates
demonstrate significant progress. Until then, ignore supplier allusions to their offerings' AGI or
artificial human intelligence attributes — these are generally programmer-created illusions.
Business Impact: AGI will likely not emerge in the next 10 years. When it does, it will likely be the
result of the combination of many special-purpose AI technologies.
We will see continued research in the next 10 years. In the long run, when AGI finally does appear,
the benefits will likely be enormous. But some of the economic, social and political implications will
be disruptive — and likely not all positive.
Page 14 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Because this is an embryonic area, there are no vendors selling systems exhibiting AGI. There is an
active area of basic research, but it has not yet advanced to the point where there are real products.
Benefit Rating: Transformational
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Recommended Reading: "Smart Machines See Major Breakthroughs After Decades of Failure"
"How to Define and Use Smart Machine Terms Effectively"
Deep Reinforcement Learning
Analysis By: Alexander Linden; Martin Reynolds
Definition: Deep reinforcement learning is the application of deep neural networks to reinforcement
learning. Reinforcement learning is a machine-learning technique where the objective is to acquire a
mapping between situations and actions. Unlike supervised learning (where there is plenty of
information for training), only sporadic rewards are available to influence a system's behavior.
Positive rewards reinforce current behavior, and negative rewards punish current or previous
behavior. The infrequency of feedback results in extended training times.
Position and Adoption Speed Justification: Reinforcement learning has been around for more
than three decades. It can be considered a heuristic form of dynamic programming, which was
introduced by Richard Bellman almost 60 years ago. The recent performance-driven success in
computer-based game playing shown by, for example, AlphaGo (developed by Google DeepMind)
and certain kinds of robot control, has driven renewed interest in a variant called deep
reinforcement learning using deep learning systems. There are a few open-source frameworks that
support the application of reinforcement learning (Google TensorFlow and those of OpenAI, for
example), but almost all commercial data science workbenches currently lack this functionality.
User Advice:
■ Don't put deep reinforcement learning on your development or deployment roadmaps unless
your problem cannot be solved in any other way.
■ Very few practical applications are available for reinforcement learning — search strategies,
game playing, robotics and control engineering are examples.
■ Deep reinforcement learning almost always requires deep expertise and, ideally, a simulation or
controlled environment where the system can search for a range of policies that will ultimately
yield the optimal evaluation.
■ Deep reinforcement learning reduces the need for labeled data, but it typically requires a
simulation, and finding good machine-learning models will require dramatically increased
training time.
Gartner, Inc. | G00314560 Page 15 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Business Impact: Deep reinforcement learning has potential primarily in the gaming and
automation industries. It has the potential to deliver incremental efficiency improvements in complex
automated processes. It may also lead to breakthroughs in robotics, including self-driving cars and
humanoid robot chassis.
Benefit Rating: Transformational
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Sample Vendors: Google; Nvidia; OpenAI
Neuromorphic Hardware
Analysis By: Chirag Dekate; Martin Reynolds; Tom Austin
Definition: Neuromorphic computing can be defined as semiconductor-based processors that are
conceptually inspired by neurobiological architectures. Neuromorphic chipsets feature non-von-
Neumann architectures and often require execution models that are dramatically different from
traditional processors.
Position and Adoption Speed Justification: Neuromorphic systems are at the very early prototype
stage. IBM has delivered a TrueNorth-based system to Lawrence Livermore National Laboratory.
BrainChip's Spiking Neuron Adaptive Processor (SNAP) technology enables fast and energy-
efficient integration of unsupervised learning. Hewlett Packard Labs are developing DotMatrix, a
neuromorphic engine designed to accelerate neural information processing. Micron's Automata
Processor is designed to deliver extreme parallelism and performance for graph analytics, pattern
matching and data analytics. There are three major barriers to the deployment of neuromorphic
hardware:
■ Accelerated computing technologies (., GPUs) are more accessible and easily programmable
than neuromorphic silicon.
■ Knowledge gaps. Programming neuromorphic hardware will require new execution models and
programming methodologies.
■ Scalability. The large numbers of neurons and deep interconnect will challenge the ability of
semiconductor manufacturers to create viable neuromorphic devices.
At the moment, these projects are not on the mainstream path for deep neural networks (DNNs), but
that could change with a surprise breakthrough in programming techniques.
User Advice: Neuromorphic computing architectures can deliver extreme performance for use
cases like deep learning, enabling real-time analytics while consuming very little energy.
Furthermore, neuromorphic architectures can enable new class of applications that have very low
temporal and spatial locality such as graph analytics. Most of the neuromorphic architectures today
are not ready for mainstream adoption. I&O leaders can prepare for neuromorphic computing
architectures by:
Page 16 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
■ Creating a roadmap plan by identifying key applications that will be impacted by neuromorphic
computing.
■ Partnering with key industry leaders in neuromorphic computing to develop testbeds using
prototype processors and software.
■ Developing applications for neuromorphic processor architectures will require new
programming skillsets. Identify new skillsets that need to be nurtured for successful
development of neuromorphic initiatives.
Business Impact: Neuromorphic hardware faces the largest barriers in advancing deep learning,
but also may unlock the most powerful results. There are likely to be major leaps forward in
hardware in the next decade, if not from neuromorphic hardware, then from other radically new
hardware designs.
We are in the midst of a "big bang"-type change in smart machines, enabled by radically new
hardware designs, suddenly practical deep neural network algorithms and huge amounts of big
data used to train these systems. This big bang will result in machines being able to tag,
contextualize and react to language, content and people's behavior; add substantial value to what
people do; and improve on some things we used to think only people could do (drive automobiles,
for example).
Every major industry will be ripe for disruption by these smart machines. Early adopters will have
the best opportunity to drive their own destiny.
Benefit Rating: High
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Sample Vendors: BrainChip; Hewlett Packard Enterprise; IBM; Micron Technology
Recommended Reading: "Cool Vendors in Novel Semiconductors for Neural Networks, 2016"
"Market Guide for Compute Platforms"
"Three Elements of High-Performance Machine Learning Infrastructure Strategy"
Human Augmentation
Analysis By: Jackie Fenn
Definition: Human augmentation creates cognitive and physical improvements as an integral part of
the human body to deliver performance that exceeds normal human limits. Augmentation examples
include increased physical strength (for example, through exoskeletons), improved perception (for
example, a hearing aid with a phone app to optimize directionality, or an implanted magnet that
Gartner, Inc. | G00314560 Page 17 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
detects electrical current) and enhanced mental focus (for example, through medication or brain
stimulation).
Position and Adoption Speed Justification: Organizations and society must confront a growing
range of opportunities and challenges relating to employees who choose — or, in some cases, are
required or are financially motivated — to enhance their bodies and minds though technology.
Increasing specialization and job competition are demanding levels of performance that will drive
more people to experiment with enhancing themselves, triggering a multibillion-dollar human
augmentation market during the next quarter-century. Based on elective augmentation trends (in
particular, the popularity of cosmetic surgery) and the growing range of augmentation opportunities
available, we are positioning human augmentation midway between the trigger and the peak, even
though it will be well over a decade before a significant number of organizations and individuals are
affected. In the meantime, some organizations will contemplate offering their staff augmentation
opportunities to increase performance, or will create policies to govern augmentation trends.
User Advice: Organizations aiming to be very early adopters of technology, particularly those
whose employees are engaged in physically or mentally demanding work, should track lab
advances and early commercialization in fields such as exoskeletons for strength, endurance and
worker safety, and sensory enhancement or transference to improve information processing.
Research advances are currently most rapid in the area of prosthetics, which are incorporating
sensory feedback, and are becoming increasing flexible and fast for users to learn to use to through
machine learning (see "Want a True Bionic Limb? Good Luck Without Machine Learning," Wired).
Once developed, advances from medical research will rapidly become available as enhancement
technologies. Cognitive enhancement through technology is already represented by the growing use
of — and dependence on — instant mobile and voice access to information and community.
Organizations must also continue to be ready for consumer- and employee-led adoption of the
latest wearable or even implantable technologies.
Ethical controversies regarding human augmentation are emerging even before the technology
becomes commonplace. Several states have already passed bills banning employers from requiring
chip implants as a condition of employment. Future legislation will need to tackle topics such as
whether a person has a right to certain types of augmentation as a medical service, and whether an
employer is allowed to prefer a candidate with augmented capabilities over a "natural" one.
Employers will need to weigh the value of human augmentation against the growing capabilities of
robot workers, particularly as robots may involve fewer ethical and legal minefields than
augmentation.
Business Impact: The impact of human augmentation — and the ethical and legal controversies
surrounding it — will first be felt in industries and endeavors demanding extreme physical
performance, such as the military, emergency services and sports, followed rapidly by those
requiring intense mental focus and stamina, such as financial trading and high-stakes sales.
Universities and some industries are already grappling with the use of nootropics, or cognitive-
enhancing drugs, typically used off label to increase focus and mental performance.
Technology and talent management leaders will find themselves at the intersection of technology,
biology and ethics as they support and manage people who are prepared or required to augment
themselves. Highly competitive work environments and performance-based incentives may require
Page 18 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
new risk measurement and monitoring techniques to detect instances of covert augmentation — for
example, by monitoring for anomalies in performance and achievements.
Benefit Rating: Transformational
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Recommended Reading: "Maverick* Research: The Future of Humans: Get Ready for Your
Digitally, Chemically and Mechanically Enhanced Workforce"
"Technology Overview: Quantified Self"
5G
Analysis By: Sylvain Fabre; Mark Hung
Definition: 5G is the next-generation cellular standard after 4G (Long Term Evolution [LTE], LTE
Advanced [LTE-A] and LTE-A Pro). It is currently being defined across several global standards
bodies — International Telecommunication Union (ITU), Third Generation Partnership Project (3GPP)
and European Telecommunications Standards Institute (ETSI). The official ITU specification,
International Mobile Telecommunications-2020 (IMT-2020), targets maximum downlink and uplink
throughputs of 20 and 1 Gbps, respectively, and latency below 5ms and massive scalability.
Position and Adoption Speed Justification: Gartner expects that by 2020, 3% of network-based
mobile communications service providers (CSPs) will launch the 5G network commercially.
In addition to the global industry bodies that are working on the 5G specification, there are regional
influencing groups (5G Forum, IMT-2020, Fifth Generation Mobile Communication Promotion Forum
[5GMF], Mobile and wireless communications Enablers for Twenty-twenty [2020] Information
Society [METIS], 5G Innovation Centre [5GIC], ETSI).
The 3GPP's Release 15 will most likely be finished in 2018. Therefore, commercial network
infrastructure with early 5G-standard compliance could be achieved by 2019.
In addition to that, a recent proposal in 3GPP called 5G New Radio (NR) is looking at enabling
mobile network operators (MNOs) to launch 5G in 2019, with only new radio access network (RAN)
deployments, leaving the existing core intact.
Examples of early CSPs' 5G plans include:
■ In 2017, Verizon will be launching fixed wireless access in network in select areas in the 28
gigahertz (GHz) spectrum (with previous trials in 15GHz, 28GHz, 39GHz, 64GHz).
■ AT&T — on , 15GHz, 28GHz. (In April 2017, AT&T also announced its plan for "5G
Evolution." With these faster speeds possible, the latest devices will be in over 20 major .
metro areas by the end of 2017. However, this is based on LTE-A Pro.)
Gartner, Inc. | G00314560 Page 19 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
■ Telstra (Australia) — 5G network for Gold Coast 2018 Commonwealth Games.
■ In South Korea — kt will showcase 5G networks at the 2018 Winter Olympics in PyeongChang.
■ T-Mobile USA — rollout expected to start in 2019 targeting national coverage in 2020 using
600MHz.
■ Sprint USA — targeting 2019 deployment on .
■ NTT Docomo — showcasing deployment for 2020 Summer Olympics in Tokyo.
A driving factor for 5G adoption is the global competitive landscape of next-generation broadband
access. For example, the EU's digital agenda has a target to realize 100% broadband coverage of
30 Mbps (at a minimum) by 2020 — that includes 50% of households having 100 Mbps available
subscriptions or higher by 2020.
From 2018 through 2022, organizations will mainly utilize 5G to support IoT communications, high
definition video and fixed wireless access. (See "Emerging Technology Analysis: 5G.")
Use of higher frequencies for spectrum, as well as massive capacity, will require very dense
deployments with higher frequency reuse.
As a result, Gartner expects the majority of 5G deployments to initially focus on islands of
deployment, without continuous national coverage, and typically reaching less than full parity with
existing 4G geographical coverage by 2022 in developed nations.
In addition to that, slower adoption of 5G by CSPs (compared to 4G) means less than 45% of CSPs
globally will have launched a commercial 5G network by 2025.
User Advice: CSP technology business unit leaders should:
■ Focus mobile infrastructure planning on LTE, LTE-A, LTE-A Pro, small cells and heterogeneous
networks (HetNet), as part of a planned transition toward 5G. Standards-compliant commercial
network equipment could be available by 2019, and commercial CSP rollouts occurring before
2019 are expected to leverage prestandard equipment.
■ Clarify 5G's role within the Internet of Things (IoT) ecosystem before 5G's commercial launch.
■ Ask vendors to indicate which standard they are building in order to address the risk of
increased marketing hype around 5G, until a 5G standard is actually defined.
■ Test backward compatibility to preceding generation (LTE) devices, especially with pre-5G
networks. This is necessary because initial 5G coverage may be limited, so new devices need
to be able to use at least the 4G infrastructure as a fallback.
■ Act now to secure availability and cost for pre-5G/nonstandard devices, as this is most certainly
going to be an issue, at least until 2020.
■ Focus on related architecture initiatives — such as software-defined network (SDN), network
function virtualization (NFV), wireless-edge computing, and distributed cloud architectures, as
well as end-to-end security in preparation for 5G. 4G mainly adopts cellular network
Page 20 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
architecture, but 5G will prove more complicated and a heterogeneous network (HetNet) will be
commonly adopted, so topology changes must be planned.
Enterprise business leaders should:
■ Evaluate the multiple alternatives available now that may be adequate and more cost-effective
than 5G for many use cases (for example, low-power wide-areas [LPWAs] such as NarrowBand-
Internet of Things [NB-IoT], long-range [LoRa], Sigfox, Random Phase Multiple Access [RPMA],
Wireless Smart Ubiquitous Networks [Wi-SUN]).
Business Impact: 5G requirements cover primarily three technology aspects:
■ Enhanced mobile broadband (eMBB)
■ Massive Machine Type Communications (mMTC)
■ Ultrareliable and low-latency communications (URLLC)
URLLC and mMTC will be implemented after eMBB.
Only eMBB addresses the traditional mobile handset requirement of ever higher throughput. URLLC
addresses many of the existing industrial, medical, drones and transportation requirements, where
reliability and latency requirements surpass bandwidth needs. Finally, mMTC addresses the scale
requirements of IoT applications.
5G targets up to 150,000 broadband users, or 200,000 mMTC low power IoT modules per square
kilometer (Next Generation Mobile Networks [NGMN]).
5G's increased bandwidth incremental value on top of LTE and LTE-A, as well as a mature small cell
layer and pervasive Wi-Fi, may be limited with respect to the deployment costs involved (as is the
case with every new wireless network generation). Low latency is potentially a much more critical
differentiator.
Benefit Rating: High
Market Penetration: Less than 1% of target audience
Maturity: Emerging
Sample Vendors: Cisco; Ericsson; Huawei Technologies; Intel; NEC; Nokia; Qualcomm; Samsung;
ZTE
Recommended Reading: "Emerging Technology Analysis: 5G"
"Market Guide for Proto-5G Infrastructure"
"Market Trends: Is 5G and IoT Hype or Opportunity?"
"IT Market Clock for Mobile Communications Service Provider Infrastructure, 2016"
Gartner, Inc. | G00314560 Page 21 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
"Magic Quadrant for LTE Network Infrastructure"
Serverless PaaS
Analysis By: Yefim V. Natis
Definition: A PaaS offering delivered with serverless characteristics is serverless PaaS. Serverless
is a way of delivering an IT service where the underlying service-enabling resources are opaque to
the customer, continuously available in required quantities — thus requiring no preprovisioning —
and priced in terms of the engaged IT service, not the underlying consumed resources. Function
PaaS (fPaaS) is a notable example with special constraints above the basic serverless
characteristics. It is not the definition of an etalon of serverless PaaS.
Position and Adoption Speed Justification: Serverless delivery of IT services has gained broad
notice after Amazon popularized its AWS Lambda function platform service. Although some
associate the notion of serverless exclusively with fPaaS, the significance of serverless, as seen by
the leading vendors (including Amazon, Google and Microsoft), extends beyond functions. All PaaS
capabilities can be delivered with serverless characteristics; some are already and most will in the
future. Serverless PaaS will augment, and in some cases replace, the traditional transparent model
of delivery, such as the model of Salesforce (Heroku), AWS Elastic Beanstalk or IBM Bluemix Liberty
for Java.
As the full scope of serverless delivery of PaaS capabilities rolls out, the definition will likely be
refined: relaxed in some aspects and possibly further constrained in others. Note that serverless
delivery principles also describe the common architecture of IaaS: the underlying hardware is
hidden; pricing is set for virtual compute capacity (not hardware consumption); an open-ended
number of compute units is continuously available and preprovisioning is optional. Serverless PaaS
will likely support optional preprovisioning as well, offering lower costs to many applications with
steady and predictable demand for resources. The constraints of fPaaS on time and resource
consumption per instance will likely also not be retained for general serverless PaaS practices.
The current market dynamic already reflects these trends. Adoption of fPaaS is rapidly increasing in
development of new applications, in new vendor renditions of fPaaS (including IBM, Google and
Microsoft) and the emergence of several open-source serverless programming frameworks and
platforms (Funktion, Apache OpenWhisk). The principles of serverless architecture are also
increasingly applied beyond just the fPaaS: other cloud services from various providers are
delivered serverless, including databases (SQlite, FaunaDB, DynamoDB) and other forms of cloud
platform services. Most high-productivity application platform as a service (aPaaS) (like Salesforce
[], Mendix or OutSystems) exhibit most of the characteristics of serverless delivery. So,
too, do many other current xPaaS.
fPaaS experience will become the foundation for the more general serverless PaaS. As fPaaS
evolves beyond hype — through the inevitable disappointments and toward the Plateau of
Productivity — serverless PaaS will follow, building on the fPaaS lessons learned, but also creating
its own hype and disappointments before maturity.
User Advice: CIOs, CTOs, IT leaders and planners:
Page 22 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
■ Use fPaaS offerings as representative of serverless PaaS to build in-house understanding of the
trade-offs of the new platform delivery model, but with clear understanding that some of the
constraints on design of functions (such as duration and size) are not attributes of the general
serverless model. fPaaS is a special purpose example, but not the definition of serverless.
■ When selecting platforms for cloud-native initiatives, look for platform services that closely
approximate or match the serverless delivery model to achieve improved productivity, cost-
efficiency and consistency of outcomes.
■ Avoid the serverless model if the project requires advanced and direct forms of control over
application infrastructure operations.
■ Make the cloud platform selections with an effort to minimize vendor or service lock-in — the
increasing adoption of serverless delivery model and other ongoing innovations may compel
you to consider alternative options in platforms and vendors.
Business Impact: All PaaS should have been serverless from the start to reflect the fundamentals
of design of both IaaS and SaaS, and most indeed is to some degree. Serverless PaaS represents
the true cloud-style operations for cloud platform services. Adoption of a serverless PaaS delivery
model will increase productivity and efficiency of PaaS, and help to streamline development, scale
operations and reduce infrastructure costs. It will create a more consistent and manageable
environment for cloud applications, but will require adjustments in the practices and strategies of
planning, designing and operating the PaaS-based solutions, rendering some current applications
legacy and requiring some new training and tooling.
Benefit Rating: Moderate
Market Penetration: 5% to 20% of target audience
Maturity: Emerging
Sample Vendors: Amazon Web Services; Google; IBM; ; Microsoft
Recommended Reading: "Platform as a Service: Definition, Taxonomy and Vendor Landscape,
2016"
"The Key Trends in PaaS, 2017"
"Adding Serverless Computing and fPaaS to Your Cloud-Native Architecture Toolbox"
Digital Twin
Analysis By: Marc Halpern; Alfonso Velosa; Simon F Jacobson
Definition: A digital twin is a virtual counterpart of a real object. As its purpose, a digital twin
enables other software/systems to interact with it rather than the real object directly to improve
maintenance, upgrades, repairs and operation of the actual object. The minimum elements of a
Gartner, Inc. | G00314560 Page 23 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
digital twin include the model of the object, data from the object, a unique one-to-one
correspondence to the object and the ability to monitor the object.
Position and Adoption Speed Justification: The idea of modeling many things including cars,
buildings and consumer products, with functional behavior embedded in the virtual models is just
emerging. Until now, not even 1% of such assets are modeled such that the models capture and
mimic behavior. Digital twins today have gained tremendous mind share but remain the purview of
relatively few professional communities in select manufacturing industries or utilities.
In high-value asset-intensive industry sectors (., transportation and manufacturing) and mission-
critical sectors (., aerospace and defense), it is fairly common to instrument and model complex
things (., vehicles, aircraft, spacecraft, machines) but even so, digital twins are still rare. To date,
Gartner estimates that only 5% of such complex assets are modeled.
In consumer-oriented industries such as consumer electronics, simple digital twins are beginning to
proliferate to differentiate the products (., stereo systems, smart lighting, etc.) among consumers.
For example, a digital twin could be a model of a home sound system that enables a remote user to
manipulate the physical system with virtual sliders and buttons on a mobile device. The features of
digital twins can be criteria for selecting which consumer electronics brands to buy. Also, software
updates intended to update user interfaces for millions of products could first be instantiated and
validated on the digital twins. Gartner expects such simple digital twins to proliferate rapidly.
Increasingly, organizations will use more detailed digital twins to avert equipment failure and plan for
equipment service, to plan manufacturing processes, to operate factories, to predict equipment
failure or increase operational efficiency, and to perform enhanced product development (based on
simulating the behavior of new products based on the digital twin insight from prior products, taking
into consideration their cost, environment and performance). These more complicated digital twins
will proliferate at a slower rate due to the difficulty and expense of creating them. But, over the next
10 years, they will be adopted by operations managers for assets where the cost-benefit analysis of
risks in operations makes the case for digital twins compelling. It is also possible that organizations
might invest in simple digital twins such as thermostats or relays that are a critical part of more
complicated systems, which are not fully modeled as digital twins yet software interconnects exist
between the digital twins, their actual counterparts, and the complicated systems they connect to.
User Advice: IT strategists whose companies must manage assets, products, or systems over
multiple years should be identifying and prioritizing opportunities to enable digital twins that improve
customer experiences and business operations. Early evidence suggests that customers delight in
the convenient ability to monitor and control their consumer electronics remotely through relatively
simple digital twins that exist on their mobile devices. Therefore, companies and entities with lower-
value assets should consider whether simpler digital twins can be used, economically, to help
improve the reliability and user experience of those assets. In industries such as manufacturing and
utilities, the shift from preventative to predictive (condition-based) maintenance is a well-
established, high-value use case for digital twins. Ideally, a digital twin implements one-for-one
monitoring and control for each, distinct physical asset, and the digital twin counterpart can be
queried or controlled with impact on the actual counterpart by authorized parties.
Page 24 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
IT strategists do not have to choose between simple or complicated digital twins. Digital twins can
range anywhere from simple to complicated. Or, simple digital twins can be systematically
enhanced to duplicate increasingly complicated aspects of actual systems.
Digital twin planners must factor culture change as part of adopting digital twin strategies. For
example, technicians, engineers and operations personnel who operate real-world things will
increasingly need to work with data scientists and other IT professionals who have an expanding
role in improving safety, reliability and performance by enabling digital twins.
When seeking the technology to adopt, IT strategists should look for IoT solutions, either IoT
devices or IoT software, that provide digital twin templates that can be easily leveraged to create
digital twins for your particular requirements and assets. They should also adopt IT that ensures all
aspects of a digital twin — the sensors, metadata, data and analytics — are secure since digital
twins are proxies for real-world systems.
Business Impact: Digital twins are transformational because hundreds of millions of things will
most likely have digital twins within three to five years. They will compel business to operate
differently. Benefits include superior asset utilization, service optimization and improved user
experience across nearly all industries.
Benefit Rating: Transformational
Market Penetration: 1% to 5% of target audience
Maturity: Emerging
Sample Vendors: Amazon; Autodesk; Bosch Software Innovations; Dassault Systèmes; GE Digital;
IBM; LogMeIn; Microsoft; PTC; Siemens PLM Software
Recommended Reading: "Innovation Insight for Digital Twins — Driving Better IoT-Fueled
Decisions"
"Top 10 Strategic Technology Trends for 2017: A Gartner Trend Insight Report"
"Digital Business Is Transforming New Product Development Priorities"
"Enhance Business and Manage Risks With Appropriate Simulation and Computer-Aided
Engineering Use"
Quantum Computing
Analysis By: Martin Reynolds; Matthew Brisse; Chirag Dekate
Definition: Quantum computing is a type of nonclassical computing that is based on the quantum
state of subatomic particles. The state of the particles represents information, denoted in single
elements known as qubits (quantum bits). A qubit can hold all possible results simultaneously until
read, an attribute known as superposition. Qubits can also be linked with other qubits, a property
Gartner, Inc. | G00314560 Page 25 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
known as entanglement. Quantum computers manipulate linked qubits to solve a problem,
observing (reading) the final result in the qubits.
Position and Adoption Speed Justification: Quantum computers are not general-purpose
computers. Rather, they are accelerators capable of running a limited number of algorithms with
orders of magnitude of speedup over conventional computers. These problems fall into a broad
category of search, where a traditional algorithm would take impossibly long to find a solution.
Quantum computing is probabilistic. Grover's algorithm can complete a search with a 90%
probability of a correct result in the square root of the time that a conventional computer takes. This
example expresses the speedup, and also the care with which results must be handled.
Hardware based on quantum technology is unconventional, complex and leading-edge. To date, the
largest demonstration of entanglement is about 17 qubits, which is little more than a lab curiosity.
Even so, most researchers agree that hardware is not the core problem. Effective quantum
computing will require the development of new quantum algorithms that will solve real-world
problems, while operating in the quantum state. The lack of these algorithms is a significant
problem. Researchers are trying to optimize new quantum algorithms to the specific design
characteristics of quantum computers. IBM recently opened its quantum platform for external use,
with the goal of raising awareness of quantum computing. Today, with 17 qubits, the system solves
only trivial problems, but IBM expects to continue to increase its scale by increasing the number of
qubits and decreasing the error rates.
Another emerging approach is that of trapped ions, instead of electrons. Ions are thousands of
times more massive than electrons, which makes them less susceptible to noise, and easier to
manage. Trapped ion advocates hope to achieve tens of linked cubits within the decade.
The technology continues to attract significant funding, and a great deal of research is underway at
many university and corporate labs. D-Wave Systems, a manufacturer of annealing based quantum
computers, currently leverages 2,000 qubits but does not rely on fully entangled qubits. Google, a
user of a D-Wave quantum computer, believes that it might accelerate deep learning using the
machine. Microsoft's Quantum Architectures and Computation Group (QuArC) is working on
developing quantum algorithms as well as developing a software architecture for programming
future algorithms.
User Advice: In the few known applications, quantum computers can operate exponentially faster
than conventional computers. One example, noted above, is known as Grover's algorithm. However,
Grover's algorithm is worthless for computers with a small qubit count.
Given the focus and achievements of research in quantum computing, our view is that general-
purpose quantum computers will never be realized; they will instead be dedicated to a narrow class
of use. This suggests architectures where traditional computers offload specific calculations to
dedicated quantum acceleration engines. A lack of programming tools, such as compilers, is
another factor restricting the broader potential of the technology. Specific applications include
optimization, code breaking (as prime number factoring), image analysis and encryption.
If a quantum computer offering appears, check its usefulness across the range of applications that
you require. It will probably be dedicated to a specific application and this is likely to be too narrow
Page 26 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
to justify a purchase. For those customers interested in quantum computing, Gartner recommends
the use of quantum as a service (QaaS). QaaS providers such as IBM's Q cloud and Quantum
Experience enable developers and programmers the ability to work with a quantum machine.
Quantum code is even available on github.
Business Impact: Quantum computing could have a huge effect, especially in areas such as
optimization, machine learning, cryptography, DNA and other forms of molecular modeling, large
database access, encryption, stress analysis for mechanical systems, pattern matching, image
analysis, and (possibly) weather forecasting. Analytics is likely to be a primary driver as the
technology becomes useful, but this is outside the planning horizon of most enterprises.
Benefit Rating: High
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Sample Vendors: D-Wave Systems; Google; Harris Computer Systems; IBM; Microsoft
Volumetric Displays
Analysis By: Brian Blau
Definition: Volumetric displays create visual representations of objects in three dimensions, with a
360-degree spherical viewing angle in which the image changes as the viewer moves. Unlike most
3D planar displays, which create the illusion of depth through stereoscopic or autostereoscopic
technique, volumetric displays create lifelike images in 3D space.
Position and Adoption Speed Justification: Volumetric displays have emerged from the laboratory
but are often thought of as the iconic volumetric image of Princess Leia created by R2-D2 in the first
Star Wars movie. Volumetric displays remain an elusive yet aspirational goal.
Volumetric displays fall into two categories: swept volume and static volume. Swept volume uses
the persistence of human vision to recreate volumetric images from rapidly projected 2D "slices."
Static volume displays rely on a 3D volume of active elements. Swept and static volumetric displays
suffer from the significant dangers of rapidly moving parts or ionized particles in the vicinity of
people, especially because the volumetric nature of the generated image convinces the brain that it
is solid and "real" and, therefore, can be touched. In all cases, the physical volume of data required
to generate a volumetric image is considerable, which will limits its overall advancement in the
coming years.
User Advice: Outside of specialized areas where budgets are not significant constraints, with few
exceptions, this technology remains firmly in lab rather than commercial applications. Current
technologies limit the size of volumetric space that can be displayed, and the mechanical solutions
create potentially dangerous, rapidly moving parts. Until alternative approaches can be delivered
(which seems unlikely in the near future), volumetric displays will remain an extremely niche product,
Gartner, Inc. | G00314560 Page 27 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
but eventually could be used as a product display, or to view objects or even people who are not at
the same location.
Businesses interested in experimenting with volumetric displays should investigate Looking Glass
Factory, a new entrant in the holographic display market. Its product produces a volumetric image
inside a small desktop-based display device; it was announced in 2016, but has yet to ship.
Alternative devices, such as the HoloLamp, or even simple mirrors such as the ones used in the
Tupac Shakur performance art display at Cochella 2012, could provide quality volumetric
experiences using projectors compared to swept or static volume displays.
Business Impact: General applications are not well-developed for business use with volumetric
displays. To date, simple applications in marketing have been deployed — usually targeted at high-
end retail environments. There are some specialized applications for geospatial imaging to enhance
2D maps, and for use in architectural rendering. However, most of these can be achieved at much
lower costs using other more commercialized technologies, such as 3D displays. Concurrently, the
rapid growth and continuing development of head-mounted displays and light field displays
threaten to overwhelm the continuing development of volumetric displays outside of specialized
markets. Potential application areas include medical imaging, consumer entertainment, gaming and
design, but costs will need to fall dramatically for these to be viable options for using true volumetric
displays.
Benefit Rating: Moderate
Market Penetration: Less than 1% of target audience
Maturity: Emerging
Sample Vendors: HoloLamp; Leia; Looking Glass Factory; Musion; Realfiction; Voxiebox; Zebra
Imaging
Recommended Reading: "Market Trends: Head-Mounted Displays for Virtual Reality and
Augmented Reality"
Brain-Computer Interface
Analysis By: Anthony Mullen
Definition: A brain-computer interface (BCI) is a type of user interface whereby the user's distinct
brain patterns are interpreted by a computer. Data is either passively observed for research or used
as commands to control an application or device. There are three approaches:
1. Invasive, where electrodes directly connect to the brain.
2. Partially invasive, where the skull is penetrated, but the brain is not.
3. Noninvasive, where commercially available caps or headbands are worn to detect the signals
from outside the skull.
Page 28 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Position and Adoption Speed Justification: Noninvasive methods cannot use higher-frequency
signals as the skull blocks and disperses electromagnetic waves. A major challenge for this
approach is obtaining sufficiently distinct brain patterns to perform a range of commands. While
control today is not very smooth or continuous, it is possible to control virtual objects in multiple
dimensions, play interactive games and control hardware. Notably, the world's first mind-controlled
drone race was held by the University of Florida in 2016, showing a potential path for consumer
robotics. Today there are still major issues of latency from thought to detection, making real-time
control challenging.
Currently, the best neural interfaces are used for limb prosthetics, and use 100 channels to distill the
neural signals of the brain. The Defense Advanced Research Projects Agency (DARPA) is investing
$60 million over four years to improve this to a million channels with Neural Engineering System
Design (NESD), which would see a one cubic centimeter device implanted in the human brain,
which allows neurons to transfer data to electronics. This would be a transformational step for this
technology with wide-reaching implications on not just more nuanced interfacing, but in deeply
understanding the brain from a physical and psychological dimension. Initiatives such as the Obama
administration's decade-long Brain Activity Map project will also drive forward knowledge benefiting
this field.
While invasive techniques provide better results, it is expected that the noninvasive BCIs will grow
at a quicker rate as the method has no issues with infection and discomfort, and can be more easily
accommodated by institutions, patients and consumers. Noninvasive methods make up the majority
of research; however, to date there is no large corpus of data available or standards between
providers and hardware. As a result, determining accuracy of readings based on user characteristics
such as demographic traits and state of mind and wider machine learning, has not flourished.
Brain-computer interfaces remain at an embryonic level of maturity, although we have positioned
them at the prepeak point of the Hype Cycle in recognition of the gains made in prosthetics control,
maturing open-source communities, new use cases such as drone control and increased usage for
customer behavior research. Larger commercial investments by major technology investors such as
Elon Musk (see "Elon Musk Launches Neuralink, a Venture to Merge the Human Brain With AI," The
Verge) are trying their hand in this space as well while Facebook's Building 8 research group
recently announced at their F8 developer conference a noninvasive project that allows users to
interact and type just using thoughts with a goal of 100 words per minute.
User Advice: Today, outside the medical domain, speech recognition, gaze tracking or muscle-
computer interfaces offer faster and more-flexible interaction than brain-computer interfaces. The
need to wear a headband or cap to recognize the signals is also a serious limitation in most
consumer or business contexts. As a result, there is no significant market for the use of these
devices in mainstream business IT. Ultimately, most users outside of the medical and rehabilitation
domain should treat brain-computer interfaces as a research activity and experiment with
noninvasive tools. Undertaking these projects will require a considered investment of time and
expertise. Open source communities for brain-computer interfaces and knowledge sharing are
maturing with both OpenBCI and NeuroTechX building a much-needed international network for
neurotechnology in 2015.
Gartner, Inc. | G00314560 Page 29 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Hardware manufacturers developing drones, robotics, virtual reality headsets and professional
sports devices should explore the benefits of noninvasive methods to improve performance and
experience immersion.
Platform developers in the physical and mental wellness space should consider these devices as
part of innovation programs to better understand contextual conditions that give rise to mind states
such as attention, joy and frustration.
Marketers, customer experience professionals and interaction designers can use these devices now
to add more quantitative signals on mind state to better understand how consumers use products
and view messaging.
Business Impact: The BCI market is typically segmented into neurogaming, neuroprosthetics,
defense and neuroanalysis (psychology). Neuroanalysis and neuroprosthetics are the largest
commercial segments driven by hospitals and rehabilitation centers. Psychological research centers
and military applications are next, with neurogaming is mostly nascent. These market sizes are likely
to persist for five to 10 years.
As wearable technology becomes more commonplace, applications will benefit from hybrid
techniques that combine brain, gaze and muscle tracking to offer hands-free interaction. Over the
next five years, as virtual reality (VR) hardware develops, it is likely that noninvasive versions of this
technology will be included in VR headset designs.
Benefit Rating: Transformational
Market Penetration: Less than 1% of target audience
Maturity: Embryonic
Sample Vendors: ANT Neuro; Blackrock Microsystems; Emotiv; InteraXon; MindMaze; neurowear;
NeuroSky; OpenBCI; Personal Neuro Devices
Recommended Reading: "The Future of Customer Feedback in Marketing"
"Supply Chain Brief: The Use of Wearable Technology in Transportation"
Conversational User Interfaces
Analysis By: Magnus Revang; Van L. Baker; Tom Austin
Definition: Conversational UI (CUI) is a high-level design model in which user and machine
interactions primarily occur in the user's spoken or written natural language. Typically informal and
bidirectional, these interactions range from simple utterances (like "Stop," "OK" or "What time is
it?") through to highly complex interactions (for example, collecting oral testimony from crime
witnesses) and highly complex results. As design models, CUI depends on implementation via
applications or related services or on a conversational platform.
Page 30 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Position and Adoption Speed Justification: CUIs have seen an explosive growth in 2016/17 with
chatbots, messaging platforms and virtual assistants, especially home speakers such as Amazon
Echo and Google Home, all contributing to the increased hype. The promise of CUIs is a dramatic
shift in responsibility between user and interface — where the responsibility shifts from the user
having to learn the software, to the interface learning what the user wants. This promise warrants a
transformational impact — even if current CUIs are far from living up to this promise.
Over the last year, there has been an explosion in the availability of conversational platforms used to
implement CUI. These tools have made it a lot easier for developers to build CUIs. We have, as a
consequence, also seen CUIs being implemented inside popular applications as an alternative to
GUI, and even in application suites. We expect application suite vendors to bring to market CUIs in
front of their business applications — which can quickly lead to hundreds of different chat interfaces
being available to employees of a large enterprise — on multiple messaging platforms. The
emerging pattern of chatbots acting as a guide or concierge in front of these conversational
interfaces will likely gain a lot of traction over the next year.
Most CUI implementations are still primitive, and thus are not able to respond to complex queries.
Increases in capabilities will, at first, largely come from improvements in natural-language
understanding (NLU) and speech recognition, which will bring CUIs closer to the promise and hype.
Additional capabilities around context handling, user identification and intent handling will likely
arrive within the next year, but will still not be good enough to avoid a disillusionment phase in two-
to-three years' time.
User Advice: CUIs shift the responsibility for learning from the user to the software, so the software
learns what the user wants. The impact on training, onboarding and expansion of use cases is
profound. The need for literacy-related training and tools will thus significantly diminish during the
next decade. Plan on CUIs becoming the dominant model. By 2020, at least 40% of people working
in new applications will primarily interact with CUIs there, removing much of the perceived need to
invest further in improving "computer literacy."
Be wary, however, of committing to CUIs too deeply. Conversational interfaces can make machines
smarter and improve the ability of people to handle novel situations (people and machines
collaborating will be better than either working alone), but they also carry an extra burden. For well-
developed, repetitive skills that can be performed almost effortlessly, injecting conversation can
degrade performance — unless the technology is able to recognize the repetitive patterns and is
able to invoke many steps of a routine process with a single, user-generated command.
Avoid retrofitting CUI front ends to existing applications unless this improves usability and user
delight.
Business Impact: CUIs are the interaction pattern of many chatbots and virtual assistants — both
will be significant contributors to the impact of CUIs.
Outside of this, CUIs will appear primarily in new applications. Enterprise IT leaders should be on
the lookout for (and biased toward) CUIs to improve employee (and customer) effectiveness, as well
as to cut operating expenses and time spent learning arcane computer semantics.
Gartner, Inc. | G00314560 Page 31 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
There will also be some retrofitting. Over the next five years, we do not expect large enterprises to
invest heavily in retrofitting existing systems of record where the employee base is experienced and
stable, and the feature set well-known to the user base. However, where there is high employee
turnover or significant rapid changes in feature sets, or where enterprises face a continuing burden
of providing computer literacy training, enterprise IT leaders need to consider creating people-
literate front ends to make it easier for employees to adapt and excel.
Benefit Rating: Transformational
Market Penetration: 1% to 5% of target audience
Maturity: Emerging
Sample Vendors: Amazon; Baidu; Facebook; Google; IBM; IPsoft; Microsoft; Next IT; Salesforce
Recommended Reading: "Conversational AI to Shake Up Your Technical and Business Worlds"
"Architecture of Conversational Platforms"
"Market Insight: How to Collaborate and Compete in the Emerging VPA, VCA, VEA and Chatbot
Ecosystems"
Smart Workspace
Analysis By: Mike Gotta; Carol Rozwell
Definition: A smart workspace exploits the growing digitalization of physical objects brought about
by the Internet of Things (IoT) to deliver new ways of working, scheduling resources, coordinating
facility services, sharing information and collaborating. The programmability of physical
environments enables smart workspaces to work contextually with mobile devices, software
applications, enterprise social graphs and smart machines to improve workforce efficiency and
effectiveness. Any location where people work can be a smart workspace.
Position and Adoption Speed Justification: Smart workspaces primarily reflect advances in, and
synergies between, six trends:
1. The IoT
2. Enterprise social graphs (and other types of graph)
3. Artificial-intelligence-related technologies
4. Digital signage/electronic whiteboards
5. Indoor mapping
6. Smart buildings (including trends in integrated workplace management systems)
A smart workspace is a key aspect of a digital workplace initiative, as it involves strategists involved
in facilities and real estate as key stakeholders. It applies to physical environments such as:
Page 32 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
■ Building and campus environments, including in-building open spaces
■ Office and desk spaces
■ Conference rooms
■ Huddle rooms (small spaces where people congregate)
■ Retail and shop floors
■ Manufacturing assembly lines
"Things" participate in a smart workspace. Examples include applications and devices such as
electronic whiteboards, building interfaces (HVAC), large digital displays, smart badges,
workstations, mobile devices and wearables.
Taking full advantage of a smart workspace will require organizations to revisit design strategies, to
include methods for gaining a better understanding of how people participate in physical spaces.
Adoption rates will vary, based on organizations' requirements to support flexible work models that
optimize the physical and interactive aspects of places and things (as well as employees' privacy
concerns).
Technological advances in nonenterprise environments — in consumer electronics and appliances,
as well as in homes, cities, transportation, fashion, security and so on — will influence smart
workspace innovation. Conversely, a lack of advances in these areas will constrain progression of
smart workspace technologies.
User Advice: Enterprise strategists focusing on a digital workplace strategy and digitalized
business processes should follow smart workspace trends and look for deployment opportunities,
such as meeting rooms, huddle rooms and in-building open spaces. Emerging applications will
expand beyond traditional productivity scenarios to include situations that are more industry- and
process-specific, such as an insurance professional using a digital pen that interacts directly with
back-end processing systems, or a patient being remotely monitored via a wearable interface in
their home that interfaces with diagnostic systems and advises healthcare professionals to improve
care delivery. IT organizations will need to work much more closely with real-estate and facilities
teams, and vice versa. Identity, access management, privacy and security teams will also play a
critical role.
Additionally, electronic whiteboards are becoming integrated with traditional collaboration and
content software systems, providing more opportunities for experimentation. Meeting artifacts can
be better captured and connected to digital workplace graphs, to become more widely searchable.
Beacons and sensors placed in key locations within a workplace can interact with mobile apps to
deliver personalized information to workers, based on proximity. These can be used to improve
employee learning, provide relevant information on products, or communicate safety procedures
based on employee location.
The smart workspace will emerge at an uneven pace as organizations prioritize potential solutions
independently of one another. For instance, building upgrades may take longer than expected, and
Gartner, Inc. | G00314560 Page 33 of 68
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
some market sectors will be laggards in terms of smart workspace adoption. Localization needs will
also influence smart workspace adoption.
Business Impact: The business impact of smart workspaces will be diverse, ranging from improved
employee productivity and cultural perception of the workplace by workers, to improved customer
experience as employees make better use of smart workspaces to serve clients. The results of
these changes will often be a reduction in cost, because office utilization data will guide decisions
about what types of workspace are most conducive to employee effectiveness.
The digitalization and programmatic evolution of places and things will impact IT methodologies
related to system design, requiring new skills for design teams to understand how people use
places and things. Smart workspaces will also have organizational impacts as traditional software
teams now need to work with facilities management teams in ways not previously envisioned. The
digitalization and programmability of the workplace will create new integration opportunities. For
instance, smart workspace activities will signal information to digital workplace graphs and smart
machines, and vice versa. Finally, adoption of smart workspaces will trigger a form of
consumerization — "bring your own thing" (BYOx) — as employees add their own objects to smart
workspace environments. Organizations will need to formulate and adjust BYOx policies
accordingly.
Benefit Rating: Transformational
Market Penetration: 5% to 20% of target audience
Maturity: Adolescent
Sample Vendors: AgilQuest; Condeco; Estimote; MCS; Microsoft; Oblong; Planon; Prysm; Trimble
Recommended Reading: "Market Guide for Integrated Workplace Management Systems"
"Create a Catalog of Activity-Based Spaces in the Digital Workplace to Improve the Employee
Experience"
"The Rebirth of Office Space: What Every CIO Needs to Know and Do"
At the Peak
Augmented Data Discovery
Analysis By: Rita L. Sallam; Cindi Howson; Carlie J. Idoine
Definition: Augmented data discovery (formerly smart data discovery), a key feature of next-
generation modern BI and analytics platforms, enables business users and citizen data scientists to
automatically find, visualize and narrate relevant findings, such as correlations, exceptions, clusters
and predictions, without having to build models or write algorithms. Users explore data via
visualizations, search and natural-language query technologies, supported by natural-language
generated narration interpretation of results.
Page 34 of 68 Gartner, Inc. | G00314560
This research note is restricted to the personal use of zyb@.
This research note is restricted to the personal use of zyb@.
Position and Adoption Speed Justification: Over the past five years, visual-based data discovery
tools have disrupted the traditional business intelligence (BI) and analytics market, as they are easy
to use and enable users to assemble data rapidly, and explore hypotheses visually, to find new
insights in data. Although visual-based data discovery has been transformative in the way it enables
business users to explore data (in comparison with traditional BI technologies), many of the
activities associated with preparing data, finding patterns in large, complex combinations of data,
and sharing insights with others remain highly manual. Visual-based data discovery tools are easy
to use, but since users analyze data manually by creating queries to investigate a hypothesis, it is
not possible for them to explore every possible pattern combination, let alone determine whether
their findings are the most relevant, significant and actionable.
Relying on business users to find patterns manually may result in users exploring their own biased
hypotheses, missing key findings and drawing their own incorrect or incomplete conclusions, which
may adversely affect decisions and outcomes.
Augmented data discovery can reduce time-consuming exploration and the false identification of
less-relevant insights. Instead of an analyst manually testing all the combinations of data, algorithms
for detecting correlations, segments, clusters, outliers and relationships are automatically applied to
the data, with only the most statistically significant and relevant result presented to the user in smart
visualizations and/or natural-language narration that are optimized based on the user's context.
Applying a range of algorithms to the data in parallel and explaining actionable findings to users
reduces the risk of missing important insights in the data versus manual exploration and optimizes
the resulting action or decision.
Augmented data discovery capabilities will advance rapidly along the Hype Cycle to mainstream
adoption, as a key feature of modern BI and analytics and data science platforms. More importantly,
automated insights from augmented data discovery will also be embedded in enterprise
applications — expanding its reach beyond the citizen data scientist to operational workers for
greater business impact.
By 2018, augmented data discovery, which includes natural-lang
