USE CASE SEVEN

7 Leading
Machine Learning
Use Cases
How today’s businesses are using
machine learning to achieve fast,
efficient, measurable results
INTRODUCTION

Solve your most common business challenges

with machine learning
Machine learning has moved beyond the hype to Second, you’ll want to find a use case that is rich
become a meaningful driver of value for many in data that you already have. A good business
organizations. Over half of businesses that have use case with no data leads to frustrated
deployed machine learning-powered artificial data scientists.
intelligence (AI) initiatives say the technology has
Lastly, you’ll want to evaluate whether your
increased productivity.1
business problem actually requires machine
While it’s clear that machine learning is an learning for success and whether machine
essential part of business transformation, many learning will result in better outcomes than your
organizations struggle to understand where to traditional approach. These outcomes might be
apply machine learning for the most impact. realized as cost reduction, increased employee
Selecting the right machine learning use case productivity, or improved experiences for your
requires you to consider a number of factors. end customers.

First, you need to find a balance between The best way to satisfy all of these criteria is to
optimal business value and speed. A proof of ensure that technical experts and domain experts
concept built by a siloed data scientist is not are working hand in hand on your machine
likely to generate much enthusiasm for machine learning project. Technical experts can conduct
learning in an organization. What is more apt to feasibility assessments, and domain experts will
attract the needed commitment and funding is ensure the solution is solving a real business
showing how machine learning can address the problem and will have a real impact.
practical issues your organization currently faces.
Furthermore, you’ll want to find something that
can be accomplished in 6–10 months so that you
don’t lose momentum. This is especially true if
this is your first foray into machine learning.

1
https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions.html

2
CHANGING THE GAME
INTRODUCTION

Starting with the right use case

is key to organizational buy-in 7 leading
use cases

In this eBook, we have outlined seven use cases where AWS customers 1 Improve employee productivity »
have successfully applied machine learning. These use cases will
strengthen your business case for wider adoption of machine learning, 2 Automate document data extraction and analysis »
and you can apply them to kick-start your machine learning journey or
add them to your current strategy. 3 Add contact center intelligence »

What makes a good machine learning use case? 4 Personalize customer recommendations »

• Solves a real problem for your business—one that’s important 5 Increase the value of media assets »
enough to get attention, support, and adoption
• Leverages sources of untapped data
6 Forecast demand metrics »

• Increases performance, reduces costs, and/or improves your

7 Identify fraudulent online activities »
end-customer experiences
• Includes technical experts to conduct feasibility assessments and
domain experts to ensure the solution will be used
• Can be completed in 6–10 months

When you are ready to deploy your use case, you have the choice
of using one or more fully-managed AWS AI Services to quickly get
started and easily integrate intelligence into your applications. Or, if you
want to develop your own models, you can use Amazon SageMaker—a
solution that provides you with all the machine learning tools you’ll
need in a single service.

3
USE CASE 1

Improve employee productivity IDEAL FOR ALL INDUSTRIES

by quickly and easily finding
accurate information
WOODSIDE ENERGY LTD.

Employees who have fast, easy access to accurate data are more “(With Amazon Kendra,) we’re now able to precisely search
productive. In a 2019 study by The Economist, executives identified our most valuable project engineering documents.
“ease of access to information required to get work done” as the #1 way This step-change in cognitive capability will enable
in which technology can drive productivity.2 better, faster decision-making to improve our
operations and the working lives of our people.”
Your employees can search for the information they need by asking
natural-language questions through Amazon Kendra—a highly Shelley Kalms, Chief Digital Officer

accurate intelligent search service powered by machine learning. This

is much faster and more efficient than traditional keyword search, and BAKER TILLY DIGITAL | LABS
the service is easy to deploy for businesses of all sizes. The resulting “By using (Amazon) Kendra, our clients are able to surface
boost in productivity helps accelerate research while improving
relevant information 10 times faster when compared to
decision making—and strengthens your business case for wider
SharePoint full text search, quickly surfacing an answer
machine learning adoption.
previously not possible with keyword search and connecting
them to relevant content across an enterprise-wide
repository, or providing marketing managers quick access to
crucial research on customer behavior.”
Ollie East, Director of Advanced Analytics and Data Engineering, and Tom
Puch, Sr. Manager

2
https://wtheexperienceofwork.economist.com/pdf/Citrix_The_Experience_of_Work_BriefingPaper.pdf

4
USE CASE 2

Make faster decisions by IDEAL FOR FINANCIAL SERVICES, HEALTHCARE AND LIFE SCIENCES,

automatically extracting and ACCOUNTING, EDUCATION, GOVERNMENT, LEGAL, OIL AND GAS

analyzing data from documents

ASSENT COMPLIANCE
The millions of documents created by your organization contain a
“Amazon Textract’s OCR technology enabled us to process…
treasure trove of insights waiting to be leveraged. Unfortunately,
documents while Amazon Comprehend was able to extract
manually processing the ever-growing volumes of information to make
custom entities. (Using) Amazon Augmented AI, we were
them easy to access and search is a cumbersome, costly task. Using
machine learning, your organization can gain timely access to the able to have our teams review documents in a given
information contained in your documents, leading to new insights that accuracy range and help train our next model iteration.
inform your business decisions. Combining these services…(saved) our customers hundreds
of hours in manually reviewing documents.”
For organizations looking to easily activate intelligent document
Corey Peters, AI/ML Team Lead
processing today, AWS offers three services that can be deployed
individually or in any combination. Amazon Textract automatically
extracts handwriting, printed text, and data from scanned documents. THOMSON REUTERS
Amazon Comprehend is a natural language processing (NLP) service “Our solution required several iterations of deep learning
that uses machine learning to find insights and relationships in text. And configurations at scale. Amazon SageMaker enabled us
Amazon Augmented AI (Amazon A2I) provides built-in human review to design a natural language processing capability in the
workflows to ensure accuracy of the data.
context of a question answering application…successfully
You can also quickly and efficiently develop your own machine learning allowing (our customers) to simplify and derive more value
models for text extraction and analysis with Amazon SageMaker, a fully from their work.”
managed service that helps data scientists and developers build, train, Khalid Al-Kofahi, AI and Cognitive Computing - Thomson Reuters Center
and deploy machine learning models. This service provides several
built-in machine learning algorithms—such as BlazingText and Linear
Learner—that are optimized for text classification, NLP, and optical
character recognition (OCR).

5
USE CASE 3

Add intelligence to your contact center IDEAL FOR ALL INDUSTRIES

to improve service and reduce cost
Improving the customer service experience is one of the best ways to VONAGE
differentiate your brand—and to demonstrate the value of machine “With Amazon Lex, we can empower Vonage customers
learning. Successful organizations treat their customer contact center
to choose how and where they will engage with us—
as an asset that is crucial to success rather than viewing it solely as a
building intelligent interaction paths into existing voice
cost center.
and messaging channels.”
Machine learning can help to transform a contact center into a profit Alan Masarek, Chief Executive Officer
center by reducing call wait times, improving agent productivity
and satisfaction, lowering costs, and helping to identify business
GE APPLIANCES
improvement opportunities.
“Using Amazon Connect, Amazon Lex, and Amazon Polly,
AWS offers several flexible options to easily add intelligence to your
we can automate simple (call center) tasks, such as
contact center. Amazon Connect is an easy-to-use omnichannel cloud
looking up product information, taking down customer
contact center that helps companies provide superior customer service
details, and answering common questions before an
at a lower cost. Contact Lens for Amazon Connect is a set of machine
learning capabilities integrated into Amazon Connect that allows contact
agent answers (the phone).”
center supervisors to better understand the sentiment, trends, and Byron Guernsey, Chief Strategist
compliance risks of customer conversations to effectively train agents,
replicate successful interactions, and identify crucial company and
product feedback.

If your organization already has a contact center solution in place,

AWS Contact Center Intelligence (CCI) solutions offer a variety of
ways to quickly and cost-effectively add AI to your contact center.
These solutions, available through participating AWS Partner Network
(APN) partners, allow customers to benefit from AWS machine
learning-powered solutions to enhance self-service, analyze calls in real
time to assist agents, and learn from all contact center interactions with
post-call analytics.

6
USE CASE 4

Make personalized recommendations IDEAL FOR RETAIL, MEDIA AND ENTERTAINMENT, TRAVEL AND HOSPITALITY,

to increase customer engagement EDUCATION, FINANCIAL SERVICES, GOVERNMENT, HEALTHCARE,

SOFTWARE AND INTERNET

Consumers today expect real-time, personalized experiences

across digital channels as they consider, purchase, and use products LOTTE MART
and services.
“(With Amazon Personalize,) we have seen a 5x increase
Machine learning can help you deliver these highly personalized in response to recommended products…(increasing) the
experiences, resulting in improvements in customer engagement, number of products that the customer has never purchased
conversion, revenue, and margin. AWS offers machine learning solutions before up to 40%.”
that deliver higher-quality personalized experiences across digital
Jaehyun Shin, Big Data Team Leader
channels—all tailored to your business needs.

If you want to quickly get started with personalization today, you can ZAPPOS
use Amazon Personalize—a fully managed service that leverages over
“We are…using analytics and machine learning solutions
20 years of personalization experience at Amazon. Amazon Personalize
to personalize sizing and search results for individual users.
makes it easy to develop applications for a wide array of personalization
AWS services (including Amazon SageMaker) allow (our)
use cases, including product or content recommendations, individualized
search results, and customized marketing communications—with no
engineers to focus on improving performance and results
machine learning expertise required. rather than DevOps overhead.”
Ameen Kazerouni, Head of Machine Learning Research and Platforms
Or, if you want to develop your own machine learning models for
recommendation engines, you can use Amazon SageMaker—a fully
managed service that helps data scientists and developers build, train,
and deploy machine learning models quickly. You can use built-in
algorithms such as factorization machines or XGBoost, both of which
are optimized for personalization, to readily train and deploy models.
You can also bring your own algorithms or models or select from
the hundreds of algorithms and pre-trained models available at the
AWS Marketplace.

7
USE CASE 5

Analyze media assets to increase IDEAL FOR ENTERTAINMENT AND MEDIA, SOFTWARE AND

value and create new insights TECHNOLOGY, TRAVEL AND HOSPITALITY

Media assets, such as audio and video, are invaluable in offering an

NASCAR
enhanced customer experience across entertainment, professional sports,
education, and other industries. The value of these assets can be greatly “We selected Amazon Transcribe to power the
increased through targeting, personalization, and improved monetization. captioning of NASCAR’s VOD content…spanning
Unfortunately, many companies struggle to optimize their media to fully take 195 countries and 29 languages. Since implementing
advantage of this content. Amazon Transcribe, we automatically add captions
to 99% of our VOD content and spend 97% less than
Applying machine learning to this problem can provide benefits across
what we had originally estimated.”
four key areas—easily enable better content search and discovery,
increase accessibility through captioning and localization, improve content Patrick Carroll, Senior Director, Development
monetization, and improve media compliance and moderation. AWS offers
several machine learning solutions that can help you intelligently manage NFL MEDIA
your media assets.
“Amazon Rekognition…significantly improves
AWS Media Insights Engine leverages Amazon Rekognition, Amazon the speed in which we can search for content and
Transcribe, and Amazon Translate to dramatically accelerate your ability enables us to automatically tag elements that
to index rich content—so you can quickly make it available for content required manual efforts before.”
searches, ad optimization, subtitling and localization, content moderation,
Brad Boim, Senior Director, Post Production and Asset Management
compliance adherence, and highlight generation. Amazon Rekognition
enables faster, easier image and video analysis with custom labeling to further
enhance content indexing and retrieval. Amazon Transcribe enables efficient
conversion of audio speech into text to create rich metadata indexes for
search and discovery. Combining Amazon Transcribe with Amazon Translate
provides content publishers a workstream to easily caption and translate
videos to remove barriers and increase reach and accessibility. Media2Cloud
helps you set up a serverless, end-to-end content ingest and analysis
workflow, streamlining the process of moving media assets to the cloud. It
uses machine learning to extract and analyze the metadata needed to build
the viewing experiences your customers want.

8
CHANGING
USE CASE 6 THE GAME

Forecast key demand metrics IDEAL FOR ENERGY AND UTILITIES, INDUSTRIAL, MANUFACTURING,

faster and more accurately to meet RETAIL, SOFTWARE AND INTERNET, STARTUPS, TELECOMMUNICATIONS,
TRAVEL AND HOSPITALITY
customer demand and reduce waste
Forecasting what customers want, how much of it they want, and when DOMINO'S PIZZA ENTERPRISES LIMITED
they will want it is vital to any organization’s success. Supply chain, sales, “We knew that by using AWS Services (including Amazon
finance, and other business units are dependent upon accurate demand SageMaker) we could…give our stores a glimpse into
metrics to satisfy customers, better manage inventory, and optimize the future by predicting what pizzas would be ordered
cash flow. You can use machine learning to discover how time-series
next. Customers are getting their pizza faster, hotter,
data and other variables like product features and location affect each
and fresher because of the improvements we’ve
other to generate forecasts such as product demand and resource
put into place.”
needs performance.
Michael Gillespie, Chief Digital and Technology Officer
In the past, machine learning-powered forecasting tools have been too
expensive and complex for many businesses to adopt in a meaningful
HEROLEADS
way. Amazon Forecast changes the equation, making it easy to
generate fast and accurate forecasts by combining time-series data with “By integrating Amazon Forecast, we will free up the
additional variables that are relevant to each customer’s unique needs. team to focus on more value-added work, expand the
Automated processes help you create a custom machine learning model reach of our models to be used by other teams, and
in hours—with no machine learning experience required. improve our forecast model accuracy to 99%. (Amazon
Forecast provides) faster insights, improved predictability,
Organizations that want to develop their own machine learning models
performance alerting systems, dynamic budget planning,
for forecasting can use Amazon SageMaker, a fully managed service
and more accurate investment models.”
that helps data scientists and developers build, train, and deploy
machine learning models quickly. The service includes several built-in Amit Das, Lead Data Engineer

machine learning algorithms, such as DeepAR, that are optimized for

forecasting. Amazon SageMaker removes the heavy lifting from each
step of machine learning to make it easier to develop high-quality
models for forecasting.

9
USE CASE 7
SEVEN

Make it easy to identify potential IDEAL FOR E-COMMERCE, FINANCE, RETAIL

fraudulent online activities
Around the globe, billions of dollars are lost each year to online fraud.3 EULER HERMES
Many applications that are designed to protect against potential
“With administrative and financial data of more than 30
online fraud rely on business rules that are not keeping pace with the
million companies, it can be challenging to detect cyber
ever-changing tactics of bad actors.
fraud before it impacts business operations. (Using Amazon
Fraud detection is a good application for machine learning for three SageMaker,) we were able to launch a new internal service
primary reasons. First, it addresses a problem that’s rich in data and in seven months and can now identify URL squatting fraud
can benefit from pattern identification within datasets. Second, it can within 24 hours.”
achieve results that are nearly impossible to accomplish through human Luis Leon, IT Innovation Advisor
input alone. Finally, these results are easily quantifiable in financial
terms, which can help foster executive buy-in for machine learning
across the organization. VACASA

“We’re excited about the introduction of Amazon

Amazon Fraud Detector leverages machine learning and more than
Fraud Detector because it means we can more easily
20 years of fraud detection expertise from Amazon to catch more
use advanced machine learning techniques to
potential online fraud faster and easier. It puts your data at the
center of your solution and makes it simpler to identify and prevent
accurately detect fraudulent (vacation) reservations.
fraud—with no prior machine learning experience necessary. Protecting our ‘front door’ from potential harm enables us
to focus on making the vacation rental experience seamless
For fraud detection, Amazon SageMaker offers built-in algorithms, such and worry-free.”
as Random Cut Forest and XGBoost, and hundreds of algorithms and
Eric Breon, Founder and CEO
pre-trained models available through the AWS Marketplace, allowing
you to develop fraud detection solutions in days.

3
https://www.javelinstrategy.com/coverage-area/2019-identity-fraud-report-
fraudsters-seek-new-targets-and-victims-bear-brunt

10
CHANGING THE GAME
CONCLUSION

Start or expand your machine learning

journey now
With the use cases in this eBook, you can leverage machine learning
to boost productivity, make smarter use of your data, meet customer
demands more effectively and efficiently, enhance customer experiences
and satisfaction, make better decisions faster, and reduce the frequency
and impact of fraud.

We chose to highlight these seven use cases because our customers

are achieving success with them today—and because they fulfill all the
requirements you should look for when identifying a suitable application
for machine learning. These use cases can be completed in a matter of
months, solve real business problems, leverage sources of untapped data,
and increase performance, reduce costs, and/or improve end-customer
experiences. They lend themselves to the inclusion of technical and
domain experts, and—when properly executed—generate results
that gain attention and foster executive support for wider adoption of
machine learning.

The business potential of machine learning goes far beyond these seven
use cases. With the broadest and deepest set of machine learning services
available today, AWS can help you apply machine learning in a wide variety
of ways that transform your business—allowing you to push innovation
to new heights and reimagine the possibilities of what your organization
can achieve.

Learn more about AWS machine learning »

11