4. 4
Co-creation of Value with AI and Out-tasking
%47
2020
4th
$1.36T
Haluk Demirkan, PhD & PMP
Milgard Endowed Professor of Service Innovation & Business Analytics
Founder & Director of Center & MS Business Analytics
Milgard School of Business, UW-T
CADE 2018
Venice, Italy
June 18-20, 18
66% - 84%
5. ongoing international leadership and reputation in strategic IT, service innovation,
innovation analytics, cognitive computing & artificial intelligence
SELECTED AWARDS AND HONORS
2015 ‐ IBM Faculty Award ‐ Cognitive Assistance Framework for Watson
2014 ‐ Association for Inf. Sys. ranked 5th in Top‐100 World‐wide IS Researchers
(Center for Services Leadership, Journal of Service Research, IEEE Computing Society,
Decision Sciences Journal of Innovative Education, PMI, etc.)
ACADEMIC EXPERIENCES: Professor of Service Innovation & Business Analytics;
Founder of Center for Information Based Management, University of Washington.
15+ years higher education teaching, and inter‐ and trans‐disciplinary applied
research at U. of Washington, Arizona State U., U. of Florida, Sabanci University
PROFESSIONAL EXPERIENCES: Co‐Founder & Board of Director, International Society
of Service Innovation Professionals (with IBM, Cisco & HP); Service Innovation, IT,
Data Science & Analytics Strategist & Solution Architect. 20+ years professional work
& executive education experiences at 40+ Fortune 500 companies
SELECTED APPLIED RESEARCH ACCOMPLISHMENTS SINCE 2002
150+ publications including HBR, Informs, IEEE, ACM, and others
Co‐Editor of a Book Collection Service Systems & Innovations in Business and Society
EDUCATION: Dual degree PhD in Information Systems & Operations Management;
PME & ME in Industrial & Systems Eng.; BS in Mechanical Eng; Certified PMP
WHO I AM - Haluk Demirkan, PhD & PMP
5haluk@uw.edu
6. My research areas
T-shaped Digital
Maestro/Talents &
Organizations =
adaptive innovators
Business Analytics &
Cognitive Computing
Service Transformation
& Digital Strategy
Service Science, Service-Oriented-Enterprise
& Complex Adaptive Smart Service Systems
DATA GIG
6
9. 9
I have…
Have you noticed how the building blocks just keep getting better?
Today’s talk will explore these questions…
1. What is a cognitive computing (AI vs IA)?
2. Why should we care?
3. What types are out there?
4. How to co-create business value?
5. How can you use cognitive computing help you address
organization’s strategic challenges?
6. Reports show that 66% - 84% of these projects fail, why?
What Are Better Than Legos? Drones
Made Out Of Legos
haluk@uw.edu
12. AI is Artificial Intelligence, or
intelligence in machines (smart
machines)
Cognition as a Service (CaaS):
AI building blocks for IA solutions
IA is Intelligence
Augmentation, or
people thinking and working
together with smart
machines
12
+
haluk@uw.edu
13. 13
AI/Cognitive computing
Why we are hearing so much about cognitive computing smart machines,
and digital assistants now when one of the enabling technologies, artificial
intelligence, was defined as “the science and engineering of making
intelligent machines” John McCarthy in 1955.
haluk@uw.edu
15. 15
Smart Machines >>>>> Story of Artificial
Intelligence
There is no doubt that computers are increasingly capable
of doing things that humans could once do exclusively.
16. 16
“How can organizations
use cognitive
computing to co-create
business value?”
“It is a renaissance, it is a golden age,” Bezos said. “We
are now solving problems with machine learning and
artificial intelligence that were … in the realm of science
fiction for the last several decades. And natural
language understanding, machine vision problems, it
really is an amazing renaissance.”
haluk@uw.edu
17. What types are out there?
How to co-create business value?
17haluk@uw.edu
23. There is no value without fit for purpose and fit for use
Cognitive computing (AI) has potentials to improve
the efficiency, effectiveness, sustainability, and
innovativeness of product and service offerings, and
disrupting the business models.
HOW? 27haluk@uw.edu
24. Research is needed to address service as part of
the mappings of process to virtualized resource – in
a way that cuts across organizational boundaries
28
Like Driving into the Fog...
28haluk@uw.edu
25. Micro Services and Out-tasking
Service orientation is based on modularity (i.e. decomposing systems)
(Baldwin & Clark, 2000). “Service-oriented” means the independent elements
are described, discovered, and negotiated for in terms of the “services” they
provide.
OASIS (2007) definition of SOA:
– “A paradigm for organizing and utilizing distributed capabilities that may be under the
control of different ownership domains. It provides a uniform means to offer, discover,
interact with and use capabilities to produce desired effects consistent with measurable
preconditions and expectations”
Service orientation is applicable at multiple levels – to people and processes
in addition to technology. The goal is to “decouple” resources (both technical
and human) from the processes they support (Bieberstein et al., 2005; Foster,
2005).
CLAIM: By doing this, it becomes easier/smoother/faster to realign resources
into new patterns to support new business processes – hence, improving
organization agility (Erl, 2005).
Reuse: “You’re getting more value out of the dollar that you’re investing
in technology” (Koch, 2006, CIO.com)
Organization agility and flexibility
Organizations must co-create their offerings, break siloed business
processes into modular independent micro services that can be
reused on-the-fly in loosely-coupled dynamic workflows, business
processes or “out-tasked” to external smart service providers that
are enabled by artificial intelligence and cognitive applicationshaluk@uw.edu
26. Integrated Process and Technology Framework
Tier 1
BPM (business process modeling):
- reference models (Value Delivery Modeling
- Language)
- benchmarking and requirements analysis
- simulation and use case analysis
Tier 2
Conceptual Architecture:
- information model
- deployment framework (Federated Architecture)
- integration modes
• Business is represented by business
processes defined in terms of value
• The Technical Infrastructure can be
represented by a conceptual
architecture that allows mapping
collaborative process models to
components and to required
resources with the services
• Services ecosystem is an
environment with live services,
choreography & orchestration
entities, resources, key performance
indicators, and management utilities
Two independent (Business Semantics and Technology Semantics) but
reconciled process representations that facilitate the mapping of business
process to core collaboration capabilities for accurate, fast and flexible
implementations of the process models
Tier 3
Services Ecosystem:
- environment with live service entities
- process model (DISCO!)
- engagement model
30haluk@uw.edu
27. Example: Steps 1 & 2
Step 1: Search the repositories for an appropriate service pattern
Step 2 & 3: Select the “best fitting” pattern, modify as needed, or create a
new pattern
31
Service A
(Design)
a
b
Service C
(Testing)
Service D
(Implementation)
Service B
(Development)
c
a
b2
b1 a
b
c
a b
bb
Aa = Design the technical architecture
Ab = Design the systems model
Ac = Write the test conditions
Ba = Develop the technical architecture
Bb1 = Develop software component type 1
Bb2 = Develop software component type 2
Bb3 = Develop software component type 3
Bbb = Develop component code
Bbc = Develop component database
Bc = Integrate Software Components
b3
bc
bb
bb
bc
Ca = Perform test scripts
Cb = Integration testing
Cc = User testing
Da = Create user guide
Db = Training seminars
c
haluk@uw.edu
Keith, M., Demirkan, H. and Goul, M. (2016) “Advice Network Formation in
IT Project Teams: The Role of Task Uncertainty,” Decision Sciences
Journal, forthcoming (Online 24 June 2016).
28. Example: Steps 4, 5 & 6
32
• Step 4: Identify the resources needed for the pattern.
• Step 5: Reserve resources, store the new pattern, execute the project, and
store performance results.
AI
haluk@uw.edu
29. So now, we are ready for the next steps
Workload Mapping
The goal of this research topic is to better understand how
workload is mapped across process to services and
resources.
Resource prediction – i.e. the ability to create a dynamic framework to predict
workload demand based upon the process executing. Resource prediction should
support:
• The thermostat model. In this case near real time data is read from the execution
environment and additional capacity is provisioned in response.
• The historical trending model. In this case, data is collected over time and
analytical tools predict demand on the system. Provisioning is based upon this
prediction with a mechanism to adjust over or under provisioned demand.
• The business demand model. In this case, the business software discovers
something about the job it is about to execute and communicates it to the resource
manager with the job submittal – e.g. a job used for a daily data movement batch
typically moves 10k rows of data. The same job can be used for a full refresh
containing 30 million rows of data. The ET&L job opens the data and discovers its
size and communicates what it has learned to the resource manager. Using this
information and the historical trending model, the resource manager can size the
resource demand and reconcile it against policy and SLA constraints.
33haluk@uw.edu
30. As a result of our analysis
We found out that the methods do not work well
Semantic match making between
Reservation method
Container allocation
Because
A workflow representation must facilitate workload mapping all the
way through to resources
Facilitate availability of workflow scheduling parameters for workload
mapping processes in order to orchestrate a business process
throughout the stack
Make decisions on whether resource availability should impact the
SCORE step in DISCO!, i.e., the selection of a choreography for
orchestration based on historical provisioning and currently available
resources
Recognize that activities within switch and while constructs that offer
an opportunity for control structures to be modified outside an
organization’s ecosystem have significant workload mapping
implications
34haluk@uw.edu
31. We are also developed and developing
Workflow decoration structure
Resource request schema
Service request unit
SLA design
Negotiation Process
Digital workflow management/failure recovery process
Trust security model for inter-organizational workflow
Semantic security model for inter-organizational workflow
36haluk@uw.edu
32. • Computers can help us be more objective and amplify our
intelligence.
• Technological progress can never be stopped even if it
should be better managed.
• Lamenting jobs lost to technology is little better than
complaining that antibiotics put gravediggers out of work.
• Weak human + machine + better process is superior to
strong human + machine + inferior process
Kasparov’s Law
37haluk@uw.edu
33. Escalating Project Overhead
Significant increase in the number of relationships to manage
(modularity tradeoff) (complexity ↑)
– Reuse → bottlenecks1, 2
New dependencies are required between service providers
(interdependence ↑)
– New forms of collaboration are needed between business and IT groups
Increased dynamics from inter-unit and -organizational dependencies
(complexity and interdependence ↑)
Complexity ↑ + Interdependence ↑ = Task Environment Uncertainty ↑
– (Thompson, 1967; Galbraith, 1973; Mintzberg, 1979)
38
1 http://www.infoworld.com/article/07/05/14/20FEsoabottle-intro_1.html
2. Lara et al., 2007, IJEC
When we are designing “Resource request schema,” “Service request
unit,” “SLA” and “negotiation processes, we noticed that overhead is
increasing significantly…
haluk@uw.edu
35. When the market size is keep growing for cognitive-enabled
digital transformation, some reports shows that 66% to 84% of
these projects fails
40haluk@uw.edu
38. Current research interests
In addition to developing a workflow management
system with out-tasking and AI, I am also working
on the following research areas:
AI training: How to train AI?
–Open data sets at ISSIP
–How to reward vs. punish
–When to retire
43haluk@uw.edu
39. Manual Autonomous
Human-centric Machine-centric
Human with Std.
Tools
Machine
Augmenting
Human
Collaborative
Machines
Human
Augmenting
Machine
Autonomous
Machine
Today 27% 11% 24% 35% 3%
In 5 years 4% 30% 29% 11% 26%
During this cognitive computing adoption process
Diversity of workforce types will change significantly
44haluk@uw.edu
40. Is Google a search company or a machine learning company?
“Google is not really a search company. It’s a machine-learning
company”
-Matthew Zeiler, CEO of visual search startup Clarifai | Enterprise | WIRED
46haluk@uw.edu
41. Discovery consists of seeing what everybody
has seen and thinking what nobody has
thought. Albert von Szent-Gyorgyi (1893-1986)
1937 Nobel Prize for Medicine
47
For any inquiries:
haluk.demirkan@gmail.com
http://www.linkedin.com/in/halukdemirk
an https://twitter.com/profhaluk
Thank you!
haluk@uw.edu