DATA
INTELLIGENCE
PLATFORM:
From Algorithms
to Intelligent
Business Processes
LUIGI DI RITO
Manager | Global Center of Excellence for SAP Platform & Technology
Data Science
Present & Future
What is changing as the technology hype enters its
maturity?
“The factory of the future will have only
two employees, a man and a dog. The
man will be there to feed the dog. The
dog will be there to keep the man from
touching the equipment”
Warren Bennis
AI adoption across industries
47% of 2135 respondents
surveyed by McKinsey for the 2018
Global Survey AI Adoption by Industry
z
and Function said they have
implemented at least one AI
technology
Source: AI adoption advances, but foundational barriers
remain, McKinsey report, 2018, URL
3
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Is a data scientist still a scarce resource?
As of 2017, annual enrollments
in data [in online learning platforms
such as Coursera or courses taught
z
by top university and industry partners
including Stanford, Princeton, and
Google] exceeded enrollments
in any other subject area
Source: Data Skills: They’re Not Just for Data Scientists,
2018, URL
4
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Biggest challenges of AI
Just 17% of respondents say their
companies have mapped out where,
across the organization, all potential
AI opportunities lie.
z
And only 18% say their companies
have a clear strategy in place for
sourcing the data that enable AI work.
Source: AI adoption advances, but foundational barriers
remain, McKinsey report, 2018, URL
5
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
The future of data science… (1/2)
Immense Data IoT and Edge Cloud Computing: We’ll Data Science/ Machine Data cleaning and
accumulation: In May of Computing: Every piece use cloud-based Learning Automation Feature engineering are
2012 we uploaded of technology is now (or solutions, even for the tools will become regular still a bottleneck for data
approximately 72 hours of will soon be) a data simplest projects. Laptops commodities that you scientists. Focus will be
video to YouTube every generator. Currently, we are not evolving fast don’t even need to know on ways of advancing and
minute. Five years later have approximately 7 enough to keep the pace how to code to use them automating modeling,
that number had billion connected IoT with the required hyperparameter tuning,
skyrocketed to 400 hours devices globally, and this computational power. and cross-validation of
every minute, which number is predicted to models
equates to 65 years of increase to 21.5 billion in
video — everyday. in 7 years.
Facebook is another huge
data generator, with over
2 billion monthly users as
of 2017, and generating
over 4 million likes every
minute.
6
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
The future of data science… (2/2)
Ethical awareness will come more to the Interpretable AI: Data Science will have to Data Science Specialization: Netflix, has
spotlight. Just as an example, USA courts are care even more about explaining machine nine Data-related roles:
already using machine learning algorithms to learning algorithms. Explaining the
sentence criminals. If the data is biased, will decisions of the so-called machine learning Business Analyst Data Analyst
the algorithm be biased as well? black boxes will be mandatory for some
industries. Quantitative
Analytics engineer
analyst
Data engineer Data scientist
Machine learning
Research scientist
scientist
Source: Data Science Trends for 2018, 2017, URL and Data Science Trends for 2019, 2018, URL Source: Beyond Interactive: Notebook Innovation at
Netflix, 2018, URL
7
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
What We See At Our
Customers
Data Driven Decisions
Data science is reshaping business processes
O Data
Customer data X Data
Product data
Click stream
Buying history Payment history
Customer Preferences External scores
Use Case 1 Pricing data Fraud scores Invoice data
Customer
Analysis Use Case 4 Payment terms
Use Case 5
Fraud
Use Case 2 Detection
Payment Risk Account statements
Search
Use Case 7
Search Select Order Pay Reconciliation
customer Use Case 3 Use Case 6
Product Accruals Receive
Recommendation prediction
INDUSTRIAL ACTION
RECOMMENDER
z
10
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Innovation eventually meets operations
I‘m not the one IT
to productize Developer
this
Here is my
python code!
?
Data IT
Scientist Operations
11
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Innovation eventually meets operations
Does our
Always the same architect know
discussion. It’s so slow. about this?
Costs money. Doesn’t
scale.
Head of And I’m half way fired or
Digital in jail if this is not
How to
compliant! transform/embed this
in our application
technology?
I‘m not the one IT
to productize Developer
this
Did anyone
Here is my check data
python code! access
rights?
How to automate
this?
How to monitor?
How to maintain this?
Data IT
Scientist Operations
12
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
How can I
The team ▪ inspire creativity and
collaboration,
▪ provide the right tools for
every one,
▪ productize the use cases and
▪ ensure adherence to
I’d like my governance
Python
Notebook
I’d like self
How can I
service data
use my
preparation.
BW and
Head of
HANA
skills…? A simple to use Digital
I’m not app…
going to Thank you!
learn SQL
for this
Data BW & BI Business
Scientist Expert Users
I just do
Business it.
Analysts
Does our All well and
architect know nice…
about this? but can I
monitor it End-
2-End ?!
IT Data IT
Developer Engineer Operations
13
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Data Intelligence Platform
How SAP bridges the gap between Innovation and
Enterprise Apps
Business User Business BI Consumer Data Scientist
& Customer Analyst
Data Apps Ad-hoc BI Data Lab
analysis
How can I provide
all needed
functionality
CONNECTIVITY
DATA PLATFORM
in an LAYER
enterprise ready
Supports
way…?
broad
connectivity DATA FACTORY
options and
data types
BIG DATA STORE (DATA LAKE)
Head of
Cloud On Premise
Digital
Infrastructure Infrastructure
15
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
SAP’s perspective
Business User Business BI Consumer Data Scientist
& Customer Analyst
Data Apps Ad-hoc BI Data Lab
analysis
HANA (CF, SAP Analytics SAP Analytics Any Tool
XSA)Fiori, etc. Hub Cloud (i.e. Matlab, R-Studio,
Jupyter, Zepplin, SAP PA)
Application In-Memory Data Access
SAP Services Database Integration Control
SAP HANA / DWC
Enterprise
& Integration
SAP Data Intelligence
Supports
Data broad Deploy Distributed Datapipeline Datapipeline Data
ML Model Data Transformation Design Monitoring Governance
connectivity
options and
data types
Cloud Machine
BigQuery Cloud Storage
Learning
Cloud On Premise
Infrastructure Infrastructure
16
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
How is sap enabling the intelligent enterprise
# OF USERS
MODE 1
Embedded New Intelligent INTELLIGENT APPS
Intelligence Applications
3rd Party Analytical
SAP Analytics Cloud BUSINESS ANALYST,
Tools
MODE 2
CITIZEN DATA SCIENTIST
ML SKILLS
SAP HANA ML Business Services, Open Source & 3rd Party
In-memory Libraries Conversational AI Machine Learning) DEVELOPER,
DATA SCIENTIST
DATA INTELLIGENCE | RPA | DATA QUALITY | AGILE DATA PREPARATION
17
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
SAP data intelligence: the foundation of the ai assembly line
Streaming Data Structured Data Unstructured Data
Data Hub Orchestration
Agile Data Preparation
Connect
Personalized Interaction
Machine Learning
Visualize and Respond
Deep Learning
Automate Business Processes
Deploy Models
Replace and Retire
Monitor Performance
Refine and Enhance 18
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Create and deploy
The Right Tools for the Job
Work in a JupyterLab or Pipeline context so
your Data Scientists can use the tools they
know, and IT can understand what is being
built
AI Labs as a Service
Spin up any lab environment supporting
major SAP and Open Source languages and
libraries so no time is wasted waiting for
infrastructure to be ready
Deployment Made Simple
Promote Lab environments into production
environments with just a few clicks
19
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Understand and explain
Understand Model Performance
Track and trace all elements of the machine
learning process to understand model
performance and ensure trustworthy results
Explain data impact
Profile and prepare data prior to model
creation
Track data lineage to understand data
impact
Reproduce Results
Review metadata and version history to
ensure reproducibility
and explicable results
20
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Automate and scale
Automate Low Value Tasks
Model performance and lifecycle is
automatically managed so IT and Data
Scientists can focus on the tasks that matter
Manage Everything at Once
One single interface to see all AI models
across the organization in a unified
workspace with tools to help manage
governance, auditability, and transparency
Reduce Infrastructure Costs
Unique serverless cloud architecture means
you only use and pay for what you need. No
More, No Less.
21
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
22
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Looking ahead: sap’s priorities in data science
3.
2. Easy access 4.
Hyper Scalers to SAP App Rich Industry
& Open Source data Content
ready 5.
Model integration
to business apps,
1. iRPA
Robust Data
Management
platform
23
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
How Can SAP Help My
Company Start
Data Science Experience Workshop
Data science experience workshop
▪ BI Manager
▪ Business Analysts
▪ Data Scientists
▪ Big Data Engineers
▪ Digital/Innovation Officers
▪ IT Architects
▪ Business Experts
SAP presents DS capabilities
Team will transfer into ideas (Day 2) Participants
with live demos (Day 1)
25
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
Key takeaways
Despite the hype, businesses are still
lagging behind in AI mass adoption
All the impressive
Business users want simple and trusted
achievementsz of deep
solutions with clear ROI; not bleeding edge
technology learning amount to
just curve fitting.
SAP is investing heavily in demystifying
and automating the E2E process: from
data, to insights to business applications
Judea Pearl, AI pioneer and philosopher
26
© 2019 SAP SE or an SAP affiliate company. All rights reserved. ǀ PUBLIC
THANK
YOU