11/3/2019
Kick off your Journey to Data Modernization & AI
Hemanth Manda Director, IBM Cloud Pak for Data
1
2
2
1 11/3/2019
Computers are like a bicycle for the mind ‐ Steve Jobs
3
Digital transformation & AI is disrupting every enterprise & industry including our own 75% The Market of large enterprises will have digital transformation at the center of corporate strategy within the next 2 years
4
4
2 11/3/2019
AI in Enterprises Top use cases today
Automated Customer Automated Preventive Intelligent processing Service Agents Maintenance automation Fraud Analysis and Investigation
Sales process Diagnosis and IT Automation recommendation and Treatment Systems automation
5
5
Bad News: You cannot Deploy AI without an Information Architecture ready for AI.
There’s no AI without IA. —
6
3 11/3/2019
Current reality of AI deployments
Production Development PoC
17% 15% 17% $
51%
Failed
Source: AI Global Survey 2019, IDC, May 2019 7
7
AI in Enterprises Top use cases today
Data Quality, 52% Quantity, Access
Algorithms 43% Explainability & 58% $ Selection
Cost and 30% Skilled Data decision criteria Science for AI Solutions Personnel
8
8
4 11/3/2019
The AI Ladder A prescriptive approach to the journey to AI
INFUSE ‐ Operationalize AI throughout the business
AI ANALYZE ‐ Build and scale AI with trust and transparency
MODERNIZE ORGANIZE ‐ Create a business‐ready analytics foundation Make your data ready for an AI and hybrid cloud world
COLLECT ‐ Make data simple and accessible
One Platform, Any
Talent & Cloud Skills
9
9
”Cloud Pak for Data” is an integrated Data & AI Platform
Common Services
10
5 11/3/2019
Expanding 3rd Party Eco‐system Sample Below – 20+ 3rd Party Services & growing
PostgreSQL Datameer Popular Opensource NoSQL Database Open source NoSQL scalable Agile Self Service Data Prep supported through IBM database & Exploration
*
Automates Persistent storage for apps Reactive micro‐services for cloud native Kubernetes storage & human-like tasks deployed using Kubernetes application development on ICP for Data data mgmnt in ML An expanding ecosystem of Partner add-ons
IBM Cloud Pak for Data
11
11
Cloud Pak for Data – Ranked #1 by Forrester
Enterprise Insight Platforms ‐ Definition
. Enterprise insight platforms pre‐integrate most — or all — of the technology required to build systems of insight and thus help business move faster. The need to move faster and change more easily is the driving force behind customer demand for these platforms.
. Vendors that can better support all the personas of an insight team with unified experiences that feature governance and can creatively enable hybrid cloud and multi‐ cloud delivery will win.
Forrester’s commentary on “Cloud Pak for Data” IBM has an impressive portfolio of individual data management and analytics capabilities that have consistently scored well on individual component Forrester Waves. With IBM Cloud Private for Data, IBM has pre‐integrated capabilities that allow clients to be productive in a week or less. We were also impressed with its ML‐assisted data cataloging and governance tools. IBM’s platform uses Kubernetes to deploy on‐premises or into the public cloud. Lastly, IBM’s support for different insight team personas through tailored but unified experiences is commendable. Firms looking to unify the work of insight teams will do well on this platform.
Microsoft’s Perceived Weakness – Azure Cloud Platform While Microsoft offers AI services, its multimodal predictive analytics and machine learning (PAML) tools scored poorly in previous Forrester Waves. Finally, we found this offering to be too light on data governance capabilities and self‐service data preparation tooling, both of which are critical insight team capabilities. Report Preview : https://ibm.box.com/s/bry68nm9alduszvrffo105cvjhmde7pn
12
6 11/3/2019
1 Governed Data Make data simple & accessible Virtualization
2 Operationalize AI Machine Learning Ops
13
Leverage the incumbent advantage in data.
Governed Data Virtualization
“Data fabric enables frictionless access and sharing of data in a distributed data environment. It enables a single and consistent data management framework, which allows seamless data access and processing by design across otherwise siloed storage. Through 2022, bespoke data fabric designs will be deployed primarily as a static infrastructure, forcing organizations into a new wave of cost to completely re‐design for more dynamic data mesh approaches.” ‐Gartner
“Tenure has its advantages. About 80 percent of the world’s data is tucked behind the firewalls of organizations. These incumbent organizations have extracted data in abundance from activities in both the online and physical domains. The data they’ve accumulated is proprietary. It’s theirs to exploit and yet most organizations admit they fall far short of utilizing it. In other words, they have big data but too little insight and value.” –IBM Institute for Business Value
14
7 11/3/2019
15
Governed AI Operationalizing AI with trust & transparency
Govern Data Build Deploy Operate Trusted AI
Data Discovery Frameworks Deployment Type Manage AI at Scale
Business KPIs Data Profiling AutoAI Validation and Feedback Scikit‐learn Online Accuracy Quality and Lineage Keras Batch Fairness and Explainability TensorFlow Data Governance Inputs for Continuous Evolution And many more Automated Anomaly and Drift Detection
Consume AI Organize Data for AI Build AI Run AI Business user Data Engineer Data Scientist App Developer
16
16
8 11/3/2019
Make Machine Learning (ML) “Creating an ML model is just a starting point. Enterprise Ready To bring the technology into production service, you need to solve various real‐world issues such Art of the Possible: as building a data pipeline for continuous ML Ops training, automated validation of the model, • Automated management version control of the model, creating a scalable serving infrastructure, and ongoing • Model Governance operation of the ML infrastructure with • Operationalize ML models monitoring and alerting.” • Ensure transparency • Remove bias in training
17
17
IBM Cloud Pak for Data One platform, any cloud
Pre‐built Use Cases Watson Applications The Ladder to AI Prepare BuildRun Manage
Watson Watson Watson Watson Machine Knowledge Studio OpenScale Catalog Learning
Unlock the value of your data and Hybrid Data Management Data Governance & Integration accelerate your journey to AI Db2 Family InfoSphere Family
Multicloud Data & AI Platform Cloud Pak for Data
IBM Cloud
18
18
9 11/3/2019
Watson Knowledge Catalog
AI Lifecycle Model Improvement Ground Truth gathering
Runtime Monitoring Data Cleansing Feature Engineering Watson Studio, Watson Machine Model Deployment Model Selection Learning, and Open Scale Model Validation Ensemble Parameter Optimization
• Search and find relevant data ⒑ • Connect and prepare data for consumption and Data Scientists analysis Enterprise Data IBM Watson Knowledge Data Analysts • Consume and analyze the data Consumption Catalog on Cloud Pak for Business Analysts • Comment, rate and share Data • Data lineage • Data ownership ☤ • Data stewardship End‐to‐End Platform for Business‐ Data Stewards • Data governance workflow Ready Data CDO • Discover metadata assets Enterprise Data Governance Integration of data quality(from LOB Risk • Classify data assets LOB Product • Build data glossary Information Analyzer) data governance • Manage metadata repository (Information Governance Catalog) and • Manage Reference Data data consumption (from Watson Knowledge Catalog) now under one • Profile data experience and brand. ␋ • Understand, monitor and remediate data quality Enterprise Data Quality Quality Analysts • Apply validation rules
19
19
Watson OpenScale will help validate and monitor AI models, deployed anywhere, to help comply with regulations and mitigate business risk
Production monitoring for compliance and safeguards Required in regulated industries and Explain and Audit model decisions use cases – FSS, HR etc. in short term; Detect and mitigate model biases others longer term* Model Validation and acceptance
Ensure that models are resilient to changing situations Required to meet Detect drift in data and anomaly in model behavior transformational goals Specific inputs and triggers to model lifecycle
Align model performance with business outcomes Foundational to all AI Correlate model metrics and business KPIs to measure business impact implementations Actionable metrics and alerts
* E.g. Fair lending practices in finance vs. GDPR across all industries
20
10 11/3/2019
21
Case Study
Customs department reduces manual effort spent in importing risk processes
Industry: Government Geography: Middle East and Africa
Benefits Client’s current rule‐based import The solution is to use AI models to • Reduction of the number of cases IBM Watson Studio risk identification system is not predict the import risk. In addition to with legal imports identified as risk accurate enough to identify scenarios that, the Customs Officers are can help save the Custom Officers IBM Watson Machine Learning where importing an item is a real risk. supported with the ability to know the time and effort. reason (explainability) for the model Most times, the system identifies • Officers can take quick and predicting risk vs. no risk. IBM Watson OpenScale cases as risk, which later on are found informed decisions on risks and to be not a real risk situation. corresponding mitigation steps Officers can make quick decisions on using explainability. It saves manual appropriate mitigation steps. The client Custom Officers spend their time inspection effort down the line. is using Cloud Pak for Data with its key more in non‐real risk cases than the features Watson Studio Local, Watson • The AI‐based risk import model can real ones. Machine Learning, Watson Open Scale, be productized and used by the Multitenancy, and also potentially Customs department of other countries for more significant Watson API Services. benefits.
22
22
11 11/3/2019
Cloud Pak for Data System True plug‐and‐play enterprise data and AI in under 4 hours right out of the box, securely behind the firewall
An all‐in‐one system pre‐integrated with all the necessary hardware and software components
Deploy a complete private cloud data and AI platform in hours, with no assembly required IBM Cloud
Dynamically scale compute, storage and software with on‐demand plug & play Hyperconverged Private Cloud System Simplify management and optimization with a unified and intuitive dashboard
23
23
Data Estate Modernization
Modernize to Cloud Pak for Data Db2 DataStage Cloud Pak for Data Cognos Analytics SPSS Modeler Cloud Pak for Data System Netezza
*MDM & Planning Analytics slated for 2020
24
24
12 11/3/2019
Modern Data Architecture is critical
25 Source : Mckinsey (see link for details) 25
Hemanth Manda Director, Offering Mgmt. Cloud Pak for Data
@hkmanda
Thank you
26
13