1) The document discusses opportunities to apply artificial intelligence (AI) to real-world problems in both the physical world of "things" like vehicles and equipment as well as human activities. 2) Some examples of using AI to eliminate waste include improving vehicle and asset utilization, reducing production waste, and improving planning and forecasting accuracy. Examples of using AI to create surplus opportunities include accident avoidance, improved risk management, and personalized education. 3) The document evaluates AI opportunities based on their potential to create surplus or eliminate waste, as well as the availability and quality of applicable data. Top use cases identified include predictive maintenance, increasing agriculture yields, traffic control, and personalized education.

1. Making AI a reality
for the enterprise and
the physical world
Mark Phillips, McKinsey & Company
Aman Naimat, Demandbase

2. 2McKinsey & Company
Why are we doing this
• Provide motivation for application of
AI to a broader set of real world
problems
• Maximize AI’s potential impact by
application of a systematic
methodology

3. 3McKinsey & Company
A snapshot of AI
SOURCE: LinkedIn
NorthAmericanworkingpopulation
100M
Datascientists
8,900
Overstated?
Wildlifebiologists
17,900

4. 4McKinsey & Company
Machine Learning Based Study
Largest 500,000companies
Classified companies into levels of AI Maturity…
350 TB unstructured business data
10 Million business relationships
100 Million People Behavioral Data
15 Billion Page Views
Analyzed by Spiderbook

5. 5McKinsey & Company
Even where AI capability exists, maturity is low
Companies employing AI at scale
Strategic direction for business
Building applications
Lab projects / proof of concept 967
494
87
5

6. 6McKinsey & Company
Application of AI concentrated in digital and data based business
Companies investing
in AI by industry
60%
in digital and
data based
businesses
8.78%
Internet
4.19%
Telecommunications
3.37%
Research
2.66%
Retail
2.55%
Marketing and Advertising
2.35%
Financial service
2.15%
Automotive
2.04%
Government administration
1.33%
Internet
1.33%
Telecommunications
1.33%
Research
1.53%
Retail
1.63%
Marketing And Advertising
1.63%
Financial Service
1.74%
Banking
1.84%
Management Consulting
1.94%
Semiconductors
1.33%
Internet
0.92%
Internet
32%
Software
Information
Technology
Services

7. 7McKinsey & Company
Inevitably, AI talent intensity highest in a few (predictable)
companies
Top spenders per capita on AI

8. 8McKinsey & Company
IoT enables AI for the physical world
Cloud / AWS
Mobile apps
Analytics
engines
Predictive
algorithms
Data
visualization
Wearables
Electrical
equipment
Machinery
Lights
Cars
Additional revenue generation
• Allow new customer interactions
• Create new service models and
business models
• Enable new products/services
and pricing models
Better products, cheaper
• Improve existing products and
services
• Increase efficiency and optimize
resources
• Improve safety
Connectivity to ‘things’Software & data analytics
New value from
“Internet of Things”

9. 9McKinsey & Company
99% of IoT data goes unused today
Oil & Gas Case Example ~30,000 tags measured
Data
capture
100%
Infrastructure
40%
Analytics
1%
Data
Management
1%
Deployment
>1%
People &
processes
0%

10. 10McKinsey & Company
Two dimensions for the applicability of AI
in the physical world
Eliminate
waste
Create surplus
opportunity

11. 11McKinsey & Company
Eliminate waste examples
• Vehicle and equipment underutilization
• Building utilization
Asset underutilization
• Drug discovery
• Negotiation of contracts
Imperfect process outcomes
• Yield loss
• Planning and forecasting inaccuracy
Material waste

12. 12McKinsey & Company
Create surplus opportunity examples
Accident avoidance Improved outcomes
Risk management Educational outcomes

13. 13McKinsey & Company
Our framework to evaluate opportunities
Total Value
From AI In
“REAL
WORLD”
Physical world
(things)
Human activity
(people)
Applicability
of AI
Availability of
data
Quality
Accessibility
Create surplus
Eliminate waste

14. 14McKinsey & Company
Results: Top use cases
Physical world (things)
• Monitoring and treating illness
• Predictive maintenance
• Improving wellness
• Centralized and adaptive traffic control
• Increasing agriculture yield
• Autonomous Vehicles
• Real time personalized promotions
• Inventory optimization in manufacturing
• Chore automation in homes
• Video crime monitoring
• Personalized Education
• Business Concierge (sales & marketing)
• Nurses
• Autonomous Vehicles inc. commercial
• Retail Shopping
• Financial Advisors
• Customer service
• Legal/Contract Negotiations
• Government
Human activity (people)

15. 15McKinsey & Company
Example: Supply chain planning and forecasting
100s – 1000s internal
and external demand
variables
Bullwhip effect:
translation of variability
throughout the supply
chain
Need for faster, granular
and precise supply
chains
• Demand sensing
• Profit-optimized
production planning
• Automated/closed loop
planning
-35+%
-35 - 65%
Inventories
Lost sales

16. 16McKinsey & Company
Example: Centralized and adaptive traffic control
Analog and de-
centralized approach to
traffic management
• Dynamic optimization
of traffic flow
Urbanization leading to
severe congestion
• Predictive incident
prevention
Increasing amounts of
data – video cameras,
mobile phones,
connected cars
• Emergency vehicle
preemption
• Multi -modal public
transport optimization
-10-20%
Transit time
-15-30%
Accidents

17. 17McKinsey & Company
Example: Business Concierge
ALEXA for B2B :
Personalized sales,
success & service
person
• Manages entire
relationship
• How product can help
their business + ROI
calculation
• Answers pricing, billing,
upgrades questions
• Logs tickets and
reports on status
Sales
Service
Success

18. 18McKinsey & Company
Example: Streaming Music Service
Need for personalized
taste & experience,
without user input
• Gets the Next Song
right
• Balances songs you
love with ones you will
• Educates, informs and
entertains
• Matches context with
content
Conver-
sions
Subscrip-
tions
NPS

19. 19McKinsey & Company
Example: Legal/Contract Negotiations
Already a Billion dollar
opportunity with early
products. A Fancy IR.
Communication and
collaboration among
stakeholders using
legalese to maximize
benefit & minimize risk
No need in the world to
ever create a new
contract from scratch! Big
Natural Language
Generation Opportunity.
Note: CODEX @Stanford doing early work in this space for past 10 years
E-Discovery Contract Law Negotiations

20. 20McKinsey & Company
How to build a solution : Government
Problem Problem Problem
Most citizens do not
know what/how state or
local bills impact their
lives
Direct Democracy
doesn’t work because it
requires citizens to have
in-depth knowledge of
issues and policy
Representative
Democracy suffers from
The Principal-Agent
Problem
This is too hard?
Too futuristic?

21. 21McKinsey & Company
Wait a second…Is it really that far out?
• A Bill is just complicated text
• Thousands of +/- feedback from Special-Interest (SIG), Non-profits, Citizens…
• All publically available
The Training Data
• Probability-of-Vote (Citizen, Bill)
• Citizen likes a dozen SIGs as preferences
• NLP + Netflix-style Matrix Factorization on SIG/Citizen feedback to derive p
(citizen, bill)
The Algorithm
• Machinery knows how every citizen will vote on every bill
• Allocates taxes optimizing preferences across all citizens
AI Representative Democracy

22. 22McKinsey & Company
Challenges in
reaching outcomes
Technology building blocks
Open Source “Real-world Data”
Coordination between technical solutions
(Interoperability)
Business model, incentives and
alignment of stakeholders

23. 23McKinsey & Company
Technology building blocks
Natural language generation/processing
Deep learning (text, speech, vision)
Evolutionary systems
Access to, and accelerated processing of, training data

24. 24McKinsey & Company
Open source ‘real world’ data: No Data è No AI
Dirty Data Need
Most public/academic datasets limited and pure. Real world data is dirty.
Thus, academic results aren’t dependable.
Silo’d
Data use limited to specific scope or domain: single-thread use cases or
historic analysis
Proprietary
6 Orgs manage most of the worlds data (Google, Facebook, Amazon,
Microsoft, Government, AT&T)

25. 25McKinsey & Company
Coordination between technical solutions (Interoperability)
40% of value from IoT derived from
interoperability of systems
Greater for AI?

26. 26McKinsey & Company
Business model, incentives and alignment of stakeholders
New outcomes based on AI
Business model
disruption
• Value realized on
reduction of waste or
creation of surplus
• Customers value
transparency
Incentives redefined
• Intermediate players in
value chain must enable
data, promote
transparency
New stakeholders vs
incumbents
• Value created by entrants
that provide value from
data
• Incumbents threatened

27. 27McKinsey & Company