A FUTURE THAT WORKS:

AI, AUTOMATION, EMPLOYMENT, AND PRODUCTIVITY

JAMES MANYIKA
Extracts From McKinsey Global Institute Research, June 2017

CONFIDENTIAL AND PROPRIETARY

Any use of this material without specific permission of McKinsey & Company is strictly prohibited
Amazing progress in AI and Automation

McKinsey & Company 2

McKinsey & Company 3
 2011 2016

Humans

26% errors 5% errors 3% errors

SOURCE: Jeff Dean (Google Brain) McKinsey & Company 4
McKinsey & Company 5
McKinsey & Company 6
Why Now?

Algorithms/techniques – Neural Networks, CNNs,

1 RNNs, Deep learning, Reinforcement Learning…

Compute power – Silicon (CPUs, GPUs, Tus …);

2 Hyperscale compute capacity, cloud available …

Data – 50 exabytes (2000), 300 exabytes (2007);

3 4.4 zettabytes (2013), 44 zettabytes (2020) …

McKinsey & Company 7

Huge benefits to business,
the economy and society

McKinsey & Company 8

Machine learning has broad potential across industries and use cases
Size of bubble indicates variety Agriculture Consumer Finance Manufacturing Pharmaceuticals Telecom
of data (number of data types)
Automotive Energy Health care Media Public/social Travel, transport,
and logistics
Volume
Breadth and frequency of data
10
Lower priority
Identify Personalize
Personalize
9 fraudulent financial
Higher potential advertising
transactions products
8 Identify and
Discover new navigate roads
Personalize crops to
7 individual conditions consumer trends
Optimize pricing
Predict personalized
6 and scheduling
health outcomes
in real time
Optimize
5
merchandising strategy
Predictive
4 maintenance
(energy)
Predictive maintenance Case by case
3
(manufacturing)
2 Diagnose diseases

1 Optimize clinical trials

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
Impact score
McKinsey & Company 9
Good for business – Drives innovation, transformation and productivity

ACCURACY OPTIMIZATION
PREDICTION
SCALABILITY

THROUGHPUT CREATION
DISCOVERY

DECISIONS

McKinsey & Company 10

Good for the economy - Automation can contribute to growth in GDP per capita
FTE automation output (United States example, 2000–65)

FTEs
Millions
450 Automation will be a
significant contributor FTEs to achieve
to the productivity boost projected GDP growth
needed to projected FTE Automation output
360 GDP per capita growth in earliest scenario

FTE Automation output

270 in latest scenario

FTEs required to maintain

current GDP per capita
180
Projected FTE

Assuming zero productivity growth,

Historical FTE based on demographic trends, the
90
projected FTEs will be less than
the FTEs required to maintain
current level of GDP per capita
0
2000 05 10 15 20 25 30 35 40 45 50 55 60 65 2070
Year
McKinsey & Company 11
What about jobs?

McKinsey & Company 12

Our approach focuses on activities and capabilities of currently demonstrated technologies
Occupations Activities (retail example) Capability requirements

Retail sales- Greet Social

1 ▪ Social and emotional sensing
people customers
▪ Social and emotional reasoning
▪ Emotional and social output
Food and Answer questions about ▪ etc
2 beverage service products and services
workers Cognitive
▪ Natural language
Clean and maintain
▪ Recognizing known patterns / categories
work areas
▪ Generating novel patterns / categories
3 Teachers
▪ Logical reasoning / problem solving
Demonstrate product
▪ Optimizing and planning
features
▪ Creativity
Health ▪ Articulating/display output
4
practitioners ▪ Coordination with multiple agents
Process sales and ▪ etc
transactions
▪ ... Physical
▪ … ▪ ... ▪ Sensory perception
▪ … ▪ … ▪ Fine motor skills/dexterity
▪ … ▪ Gross motor skills
▪ Navigation
~2,000 activities assessed ▪ Mobility
~800 occupations across all occupations ▪ etc
SOURCE: Expert interviews; McKinsey analysis McKinsey & Company 13
 BASED ON
Some activities have higher technical automation potential DEMONSTRATED
Time spent on activities that can be automated by adapting currently demonstrated technology TECHNOLOGY
%

81
69
64

26
18 20
9

Time spent
in all US 7 14 16 12 17 16 18
occupations
%
Manage Expertise Interface Unpredictable Collect Process
Process Predictable
Predictable
physical data data
data physical
physical

Total wages 596 1,190 896 504 1,030 931 766

in United
States, 2014
$ billion
Most 51% of US wages
susceptible
activities $2.7 trillion in wages
McKinsey & Company 14
Some sectors have more automatable activities than others BASED ON DEMONSTRATED TECHNOLOGY

Size of bubble indicates % of Ability to automate (%)

time spent in US occupations
0 50 100
Unpredictable Collect Process Predictable Automation potential
Sectors by activity type Manage Expertise Interface physical data data physical %

Accommodation and food services 73

Most automatable

Manufacturing 60
Transportation and warehousing 60
Agriculture 57
Retail trade 53
Mining 51
Other services 49
Construction 47
In the middle

Utilities 44
Wholesale trade 44
Finance and insurance 43
Arts, entertainment, and recreation 41
Real estate 40
Administrative 39
Least automatable

Health care and social assistances 36

Information 36
Professionals 35
Management 35
Educational services 27

McKinsey & Company 15

 Employee weighted overall % of activities that can be
All countries could be impacted by automation automated by adapting currently demonstrated technologies

<45 45–47 47–49 49–51 >51 No data

Automatability across economies Million FTE

Employee weighted overall % of activities that can be automated $ trillion

Remaining
countries
China

100% =
1,156M FTEs
$14.6 trillion

Japan
Big 5 in Europe
United States
India

McKinsey & Company 16

A small percentage of occupations can be fully automated by adapting current technologies, but
almost all occupations have some activities that could be automated

100
91
73
62
51
% of roles 42
34
(100% = 26
18
820 roles) 8
1
100 >90 >80 >70 >60 >50 >40 >30 >20 >10 >0%

Percent of automation potential

Example  Sewing machine  Stock clerks  Bus drivers  Fashion designers  Psychiatrists
occupations operators  Travel agents  Nursing assistants  Chief executives  Legislators
 Assembly line workers  Dental lab technicians  Web developers

While about More occupations will have portions of their tasks automated e.g.

5% 60%
of occupations could have of occupations could have
close to 100% 30%
of tasks automated, of tasks automated
SOURCE: US Bureau of Labor Statistics; McKinsey Global Institute analysis McKinsey & Company 17
Automation potential spans from high to low wage occupations BASED ON DEMONSTRATED TECHNOLOGY

Ability to technically automate

Percentage of time on activities that can be automated
by adapting currently demonstrated technology

100
File
clerks
80
Landscaping and
grounds-keeping workers
60
Chief
executives
40

20

0
0 20 40 60 80 100 120

Hourly wage
$ per hour
McKinsey & Company 18
Several factors affect the pace and extent of AI and automation

Technical Cost of Cost of labor Benefits Regulatory

feasibility and developing and related including and social
pace of and supply- and beyond factors
breakthroughs deploying demand labor
technologies dynamics substitution

McKinsey & Company 19

In summary…

McKinsey & Company 20

We’ve seen this before—but is this time different?
Distribution of labor share by sector in the United States, 1840–2010
%

90
Rest of the economy
80

70

60

50

40

30

20

10 Manufacturing
Agriculture
0
1840 50 60 70 80 90 1900 10 20 30 40 50 60 70 80 90 2000 2010

McKinsey & Company 21

So with huge benefits, some real challenges to address

Benefits Challenges
 Faster innovation and business  Jobs and wages
transformation  Skills and training
 Better performance, outcomes, Social and  Dislocation and
For quality, speed
businesses economic transitions
and users  Overcome human limits; Solve  Distributional issues
new problems, create new
opportunities and innovations
 Acceptance

 Safety, utility, quality of life

 Boost productivity growth,  Transparency, openness

GDP growth and prosperity and competition
For
economies
 Counter aging or shrinking
Other issues  Biases
workforce
and society  Safety, Cybersecurity
 Solve “moonshot” problems
(e.g., climate)  Ethics

McKinsey & Company 22