Journey to AI on IBM Z and LinuxONE Everything you need to get started quickly. Get started

Welcome to the Journey to AI on IBM Z and LinuxONE content solution, your homepage for technical resources.

With the introduction of an integrated accelerator for AI, organizations can scale their inferencing of transactions running on IBM z16 and IBM LinuxONE Emperor 4 while continuing to meet latency and throughput requirements. Now, AI can be directly embedded with mission critical workloads for real-time insights. Use cases from fraud detection, to risk analysis and natural language processing can be optimized like never before, allowing users to extract value from every transaction.

Explore this page to see the various use cases for AI on IBM Z and LinuxONE Emperor 4, the technologies involved, and next steps for getting hands-on with each solution.

Explore modernization patterns with AI on IBM Z and LinuxONE Emperor 4
Announcements

Do you have questions about AI on IBM Z and LinuxONE? Want to know more?

Contact us now!
Big picture 1. Determine your best-fit use case for AI on IBM Z and LinuxONE 2. Explore the technologies fitted for your use case 3. Use hands-on labs, demos, and tutorials to try out the capabilities 4. Integrate the solutions with your critical applications and middleware How to get started
Use cases

Explore use cases and some relevant capabilities in the area of fraud prevention.

IBM z16 IBM LinuxONE Emperor 4

The problem

Fraud poses a significant risk to both consumers and businesses across a variety of industries, and creating an off-platform inferencing model to detect fraud can still leave companies exposed. The inherent latency introduces timeout issues—a significant blocker when you need to rapidly approve fair transactions. Thus, a significant portion of transactions may go unscored when trying to detect fraud.

The solution

Achieve real-time scoring through co-locating an AI model with the IBM Z applications managing your transactions. This enables you to analyze 100% of transactions running through your on-premises systems, and significantly reducing risk to both your business and customers.

The business impact

Drive significant savings in exposure risk, and significant improvements in customer relationships.

Click here to learn more about this solution

The following capabilities are featured in this use case:

Watson Machine Learning for z/OS (WMLz)

Enable the rapid deployment of an AI model built for fraud detection onto the IBM Z platform, allowing real time insights without compromising security, performance, or resiliency. WMLz supports full model lifecycle management as well as capabilities for simplifying integration into business workloads.

TensorFlow

Enable the deployment of AI models built for fraud detection for use in mission critical workloads, both for real-time transactions and batch deployments.

Use TensorFlow source code for building and training credit card fraud models

IBM Deep Learning Compiler + ONNX

These capabilities together allow clients to deploy models to the IBM Z platform for inference use without requiring any of the AI frameworks used for model creation or training. Also available as part of WMLz.

Use sample code for exporting or converting a model to the ONNX format

The problem

Fraudulent claims are an impediment both to insurers and consumers with legitimate claims. The claim auditing process can be manually intensive and time-consuming, making it difficult to scale prevention, and legitimate claims are often connected to an urgent need, making delays in processing a detractor from user satisfaction.

The solution

Eliminate the manually intensive auditing process through a machine learning model trained in detecting fraud for your claims processing systems. This enables you to rapidly analyze claims and significantly reduce overall human intervention.

The business impact

Drive a significant reduction in hours spent in the auditing process, keeping customers satisfied, and allowing you to allocate resources towards higher value tasks.

Coupling the IBM Accelerator for AI with a variety of capabilities makes fraud prevention at scale easier than ever.

The following capability is featured in this use case:

Watson Machine Learning for z/OS (WMLz)

Enable the rapid deployment of an AI model built for fraud detection onto the IBM Z platform, allowing real time insights without compromising security, performance, or resiliency. WMLz supports full model lifecycle management as well as capabilities for simplifying integration into business workloads.

The problem

With regulations around anti-money laundering varying by locality, it can be very difficult for financial institutions to remain compliant with federal policies. Moreover, it can be manually intensive to identify accounts that may be associated with money laundering activities, with business analysts often relying on the time and expertise of data science counterparts to train and deploy predictive models.

The solution

Enable the analysis of potential money laundering accounts through simple querying against historical data, allowing users to independently identify problematic accounts and take appropriate action.

The business impact

Drive transparent, easy-to-follow business processes that free up key data science resources for higher level needs, while rapidly identifying potential money laundering accounts and allowing for greater compliance with various regulatory bodies.

Click here to learn more about this solution

AI applications running on IBM Z can tackle money laundering through pattern recognition at scale, helping you identify similar payment accounts or currencies to that of a suspicious transaction.

The following capability is featured in this use case:

SQL Data Insights on Db2 z/OS

Use SQL queries to determine bank accounts with similar traits to a known money laundering account. SQL Data Insights on Db2 z/OS can deliver these insights without requiring a data scientist to develop a model.

The problem

Organized retail crime is a sizable burden for nearly all retailers, yet it is challenging to develop a successful model for analyzing the risk of retail crime when you only have access to your own transactional data. It's difficult to share models between merchants without potentially exposing the sensitive data of customers.

The solution

Using federated learning enables the sharing of models across retailers without exposing sensitive customer data, allowing all parties to build more effective models and improve their ability to detect bad actors.

The business impact

With federated learning, you gain opportunities to drastically improve machine learning models while contributed to a larger community of practice. In this specific use case, this enables you to more effectively combat retail crime, saving costs on inventory replenishment.

The following capabilities are featured in this use case:

IBM Watson Knowledge Catalog

Utilize a shared cloud-based metadata repository to access information to develop machine learning models for the recognition of retail crime patterns.

Take the guided demo for IBM Watson Knowledge Catalog

IBM Hyper Protect Virtual Servers

Protect your AI applications through hybrid cloud deployments, ensuring data is secured against both internal and external threats.

Get started with IBM Hyper Protect Virtual Servers

The problem

Fraud poses a significant risk to both consumers and businesses across a variety of industries, and creating an off-platform inferencing model to detect fraud can still leave companies exposed. The inherent latency introduces timeout issues—a significant blocker when you need to rapidly approve fair transactions. Thus, a significant portion of transactions may go unscored when trying to detect fraud.

The solution

Achieve real-time scoring through co-locating an AI model with the IBM LinuxONE applications managing your transactions. This enables you to analyze 100% of transactions running through your on-premises systems, and significantly reducing risk to both your business and customers.

The business impact

Drive significant savings in exposure risk, and significant improvements in customer relationships.

The following capabilities are featured in this use case:

Go, LightGBM, Python

LightGBM is a machine learning framework with python interfaces, which is used to develop and train a machine learning model - on or off the LinuxONE Emperor 4 platform. When the model is ready to be used for inference, it is deployed to LinuxONE Emperor 4 using and using a Go based serving application. This provides for a high speed and responsive fraud detection service that, on LinuxONE Emperor 4, can scale to meet demand.

The problem

Organized retail crime is a sizable burden for nearly all retailers, yet it is challenging to develop a successful model for analyzing the risk of retail crime when you only have access to your own transactional data. It's difficult to share models between merchants without potentially exposing the sensitive data of customers.

The solution

Using federated learning enables the sharing of models across retailers without exposing sensitive customer data, allowing all parties to build more effective models and improve their ability to detect bad actors.

The business impact

With federated learning, you gain opportunities to drastically improve machine learning models while contributed to a larger community of practice. In this specific use case, this enables you to more effectively combat retail crime, saving costs on inventory replenishment.

The following capabilities are featured in this use case:

IBM Cloud Pak for Data + IBM Watson Federated Learning + IBM Hyper Protect Virtual Servers

A multi-platform solution that enables the sharing of insights gathered from confidential data. Federated learning provides a mechanism to share insights across industry groups, fostering better collaboration, while not exposing confidential data outside of the owning organization.

Runtime considerations

Fraud prevention is especially notable for organizations using the following runtime environments. The examples described are just some of the many possibilities, and there are different technology solutions available - for more details, see 'Infusing AI into applications' in 'Learn more'.

A claims processing application running in CICS TS could extend fraud detection/prevention capabilities by invoking an AI model deployed to TensorFlow hosted in an IBM z/OS Container Extension (zCX) within z/OS.

A financial application running in IMS TM could call an AI model in Watson Machine Learning for z/OS (WMLz) to determine the outcome of a transaction based on a prediction of whether it is likely to be fraudulent.

An application running in WebSphere Application Server for z/OS could invoke an AI model that uses TensorFlow on Linux on IBM Z in order to detect potential money laundering.

An electronic funds transfer system running on z/TPF could interact with a machine learning model that uses IBM Snap Machine Learning (SnapML) on Linux on IBM Z to identify potentially fraudulent transfers.

Use cases

Explore use cases and some relevant capabilities in the area of business process optimization.

The problem

When processing credit card transactions, it's critical to determine which trades or transactions are highly exposed to risk before settlement—but given the manual work involved, it can be difficult to optimize this work without potentially impacting service-level agreements.

The solution

By training a model on risk exposure and co-locating it with transactional workloads, you can unlock insights about risk on a transaction-to-transaction basis.

The business impact

Through the enhancement of a rules-based approach using the IBM Z platform, you can achieve a high level of analysis with no impact to service level agreements.

The following capabilities are featured in this use case:

PyTorch + IBM z/OS Container Extensions (zCX) + ONNX + Watson Machine Learning for z/OS (WMLz)

PyTorch provides a deep learning framework which may be trained off the IBM Z Platform by accessing data co-located on a zCX instance. The PyTorch model is then simply exported to the ONNX format and deployed using the WMLz ONNX Scoring service. This provides an endpoint that can be called from the z/OS application.

Use sample code for exporting or converting a model to the ONNX format

The problem

The loan approval process can be lengthy and manual, and customers are expecting ever-faster response times—but introducing an approval model over a distributed system can increase the risk and exposure for both lenders and customers.

The solution

An inferencing model for analyzing loan applications, alongside the automation of rules-based business decisions, can be co-located alongside transactional systems to achieve rapid insights, low latency, and minimal exposure of customer and lender data.

The business impact

Compared to models running on distributed systems, a solution built on top of the IBM Z platform can minimize loan defaults through early fraud identification, accelerate credit approval processing, and improve overall customer service and profitability.

The following capabilities are featured in this use case:

Watson Machine Learning for z/OS (WMLz) + IBM Operational Decision Manager

WMLz enables the rapid deployment of an AI model built for loan analysis onto the IBM Z platform, allowing real time insights without compromising security, performance, or resiliency. IBM Operational Decision Manager allows the analysis, automation and governance of rules-based business decisions, allowing for optimizations in the loan authorization process alongside insights produced through the deployed model.

Read about modernizing z/OS apps with ODM for z/OS and ADDI

The problem

A defaulted loan is a problem for both the lender and borrower. Without the ability to measure the factors leading to default in real time, lenders are always exposed to the risk of unfulfilled payments.

The solution

Use an optimized querying process against historical transactional data to quickly and efficiently identify problematic accounts and plan appropriate action.

The business impact

By leveraging AI capabilities on the IBM Z platform, organizations can derive value from their historical transactional data in real time, and become more proactive and agile in risk mitigation practices.

The following capability is featured in this use case:

SQL Data Insights on Db2 z/OS

Use SQL queries to determine loans with similar traits to loans that have gone into forbearance. SQL Data Insights on Db2 z/OS can deliver these insights without requiring a data scientist to develop a model.

The problem

In highly competitive insurance markets, proactivity and speed can be key to ensuring strong customer relationships—but analyzing customer circumstances to reassign insurance rates can be a manual and time intensive process. With all the time spent determining whether new rates are appropriate, lenders are limited in how much they attention they can provide to all accounts.

The solution

Use an optimized querying process against historical transactional data to quickly and efficiently identify opportunities to provide improved rates to customers.

The business impact

Insurers can proactively reassign insurance rates based on changes to customer circumstances, enabling greater user satisfaction and reducing the overall manual responses required by the insurer.

The following capability is featured in this use case:

SQL Data Insights on Db2 z/OS

Use SQL queries to determine appropriate rates for potential customers with similar traits to existing, highly qualified accounts. SQL Data Insights on Db2 z/OS can deliver these insights without requiring a data scientist to develop a model.

The problem

It can be enormously complex and time intensive to understand how weather patterns should impact insurance products. With different considerations across different regions, this becomes a problem of scale—that is, ensuring you have the expertise and computing power required to conduct such analysis, all while continuing to work at the speed of customer expectations.

The solution

An inferencing model for analyzing correlations between weather patterns and insurance risk can be co-located alongside transactional systems, allowing you to achieve rapid insights.

The business impact

Insurers can size packages and evaluate exposure based meteorological data, enabling more accurate pricing and greater management of risk, while also rapidly delivering answers to awaiting customers.

The following capabilities are featured in this use case:

SciKit-Learn + Watson Machine Learning for z/OS (WMLz)

SciKit-Learn allows you to train a machine learning model based on the categorization of insurance packages against weather patterns. WMLz enables the rapid deployment of an AI model built for loan risk analysis onto the IBM Z platform, allowing real time insights without compromising security, performance, or resiliency.

Learn about accessing SciKit-Learn and other open access frameworks on IBM Z and LinuxONE

Runtime considerations

Business process optimization is especially notable for organizations using the following runtime environments. The examples described are just some of the many possibilities, and there are different technology solutions available - for more details, see 'Infusing AI into applications' in 'Learn more'.

A loan approval application running in CICS TS could improve and accelerate decision-making by scoring a loan request within a transaction, linking to an AI model deployed to the Watson Machine Learning for z/OS (WMLz) scoring engine hosted in CICS.

An insurance application running in IMS TM could enhance its existing rule-based risk assessment by referencing a model in WMLz from the IBM Operational Decision Manager (ODM) rule driven by the application. The prediction from the model could enhance the decision returned by the rule and hence mitigation taken by the application.

An insurance application running in WebSphere Liberty could enhance customer satisfaction by using an AI model hosted in WMLz within the Liberty server to determine whether to offer a discount.

A flight booking application in z/TPF could offer improved re-accommodations when flights are cancelled or delayed by invoking a REST service provided in a co-located Linux on IBM Z server to predict the likelihood of an alternative flight being acceptable. The REST service could gather additional data as input to the model, as well as calling a model deployed to PyTorch in Linux on IBM Z to make the prediction.

Use cases

Explore use cases and some relevant capabilities in the area of image and text analysis.

IBM z16 IBM LinuxONE Emperor 4

The problem

Organizations across industries need insights about changes observed in satellite imagery to properly plan the use and development of land, however, poring over such imagery can be a manually intensive process prone to human error.

The solution

Enable accurate analysis of satellite imagery with a machine learning model deployed to the IBM Z platform, ensuring the sensitive care of aerial images while driving rapid insights.

The business impact

New thresholds of accuracy, scale, and operational costs can be achieved through running a machine learning model on the IBM Z platform, allowing teams to rapidly understand changes to a given space and expedite their decision-making process.

The following capabilities are featured in this use case:

TensorFlow + z/OS Container Extentions (zCX) + ONNX + Watson Machine Learning for z/OS (WMLz)

TensorFlow provides a deep learning framework which may be trained off the IBM Z Platform by accessing data co-located on a zCX instance. The model is then simply exported to the ONNX format and deployed using the WMLz ONNX Scoring service. This provides an endpoint that can be called from the z/OS application.

Use sample code for exporting or converting a model to the ONNX format

The problem

In any industry, it's a challenge to meet the support demands of customers at scale. With queries arriving at all times, and subject matter expertise required to resolve specific scenarios, it's clear that manual responses alone may not be enough to meet demands.

The solution

Natural language processing (NLP) can be leveraged in the creation of chatbots, and the data created in conversational interactions with AI can be stored in a highly secure on-premises server.

The business impact

By scaling the use of NLP and chatbots, organizations can meet the 24/7 support demands of their customers, increasing overall satisfaction in their services while ensuring the security of customer conversations with the IBM Z platform.

The following capabilities are featured in this use case:

huggingface + RASA

huggingface allows for the building, training, and deployment of state-of-the-art models powered by references to open source machine learning communities, while RASA allows for the automation of conversational experiences at scale.

The problem

Imaging is useful across various parts of the medical industry, including both in the diagnosis of patients and in insurance determinations. But with many organizations carrying large historical records of medical images, training a model to recognize patterns and anomalies can be resource intensive.

The solution

Effective and energy efficient computer vision for medical imaging can be achieved through the training and inferencing of AI models on the IBM Z platform.

The business impact

Organizations across the medical industry can achieve better accuracy, time to diagnosis, and energy efficiency, ultimately enabling greater care for patients and customers.

The following capabilities are featured in this use case:

IBM Deep Learning Compiler

Enables multiple deep learning frameworks to run on IBM z16

PyTorch + ONNX + Watson Machine Learning for z/OS (WMLz)

PyTorch provides a deep learning framework which may be trained off the IBM Z Platform. The model is then simply exported to the ONNX format and deployed using the WMLz ONNX Scoring service. This provides an endpoint that can be called from the z/OS application. This allows for real-time insights without compromising the security of patient data.

The problem

In any industry, it's a challenge to meet the support demands of customers at scale. With queries arriving at all times, and subject matter expertise required to resolve specific scenarios, it's clear that manual responses alone may not be enough to meet demands.

The solution

Natural language processing (NLP) can be leveraged in the creation of chatbots, and the data created in conversational interactions with AI can be stored in a highly secure on-premises server.

The business impact

By scaling the use of NLP and chatbots, organizations can meet the 24/7 support demands of their customers, increasing overall satisfaction in their services while ensuring the security of customer conversations with the IBM LinuxONE platform.

The following capabilities are featured in this use case:

Red Hat OpenShift, spaCy, RASA, TensorFlow

RASA provides an open source platform for conversational AI-like chat bots. RASA relies on highly optimized back-ends in frameworks like spaCy and TensorFlow, which can drive the execution of pre-trained Natural Language Processing (NLP) models. On LinuxONE Emperor 4, a pre-trained spaCy language model is imported to a RASA chatbot application that can provide insights to guide end users.

The problem

Imaging is useful across various parts of the medical industry, including both in the diagnosis of patients and in insurance determinations. But with many organizations carrying large historical records of medical images, training a model to recognize patterns and anomalies can be resource intensive.

The solution

Effective and energy efficient computer vision for medical imaging can be achieved through the training and inferencing of AI models on the IBM LinuxONE platform.

The business impact

Organizations across the medical industry can achieve better accuracy, time to diagnosis, and energy efficiency, ultimately enabling greater care for patients and customers.

The following capabilities are featured in this use case:

IBM Deep Learning Compiler, ONNX, PyTorch

PyTorch provides a deep learning framework which may be trained on or off the LinuxONE Emperor 4 platform. The model is then exported to the ONNX format and enabled on LinuxONE using the IBM Deep Learning Compiler, thus co-locating the model alongside key Linux workloads. This allows for real-time insights without compromising the security of patient data.

Runtime considerations

Image and text analysis use cases can be relevant to IBM Z runtime environments if part of the handling or processing of the images or text is carried out by applications running in one of CICS TS, IMS TM, IBM WebSphere Application Server for z/OS, or z/TPF.

Use cases

Explore use cases and some relevant capabilities in the area of intelligent infrastructure.

Data redaction

The problem

While organizations may rely on partnerships with vendors to analyze system diagnostic data, privacy regulations and post-processing speeds can create a significant blocker. With operational stability as a top priority, it's essential to get information in front of experts, while also taking care to protect customers' data.

The solution

Machine learning can be leveraged on the IBM Z platform to identify sensitive information and redact it from diagnostic dumps, ensuring that the resulting information allows for an inspection of system performance without creating unnecessary risk for customers.

The business impact

Organizations can be more agile in their operations with vendors and other supporting parties, allowing them to quickly produce diagnostic information without the work of manually removing PII. Thus, greater operational efficiency can be achieved on an ongoing basis.

The following capability is featured in this use case:

IBM Z Data Privacy for Diagnostics

Leverage facilities for tagging sensitive data and producing redacted diagnostic dumps which do not contain the tagged sensitive data.

Read the documentation for IBM Z Data Privacy for Diagnostics

Runtime considerations

There can be opportunities to introduce intelligence into your application infrastructure on IBM Z by leveraging AI models.

Learn more IBM Z and LinuxONE AI products and related solutions

IBM Open Data Analytics for z/OS

Available for IBM z16

Optimize the runtime of open source technologies, including Apache Spark, Anaconda and Python, and gain insights from data at its source.

IBM Z Operations Analytics

Available for IBM z16

Enable problem identification, isolation and resolution on IBM Z through analysis of structured and unstructured operational data.

IBM Z Application Performance Management Connect

Available for IBM z16

Track transaction information from z/OS subsystems to APM solutions.

IBM Operational Decision Manager

Available for IBM z16

Discover, capture, analyze, automate and govern rules-based business decisions on premises or on the cloud.

IBM Wazi as a Service

Available for IBM z16

Conduct cloud native development and testing for z/OS on IBM Cloud.

IBM Cloud Pak for Data on IBM Z

Available for IBM z16 and IBM LinuxONE Emperor 4

Confidently leverage your enterprise data within a secured, resilient IBM Z and LinuxONE private cloud infrastructure.

IBM Db2 Analytics Accelerator for z/OS

Available for IBM z16

Leverage real-time insight from data at the point of origin.

IBM Watson Machine Learning for z/OS

Available for IBM z16

Utilize unprecedented AI inferencing performance for every transaction while meeting SLAs.

SQL Data Insights on Db2 z/OS

Available for IBM z16

Uncover hidden insights in Db2 for z/OS data.

IBM DB2 AI for z/OS

Available for IBM z16

Enhance database performance with machine learning.

IBM Z Data Privacy for Diagnostics

Available for IBM z16

Leverage machine learning to detect and redact PII from diagnostic dumps.

IBM Watson AIOps

Available for IBM z16 and IBM LinuxONE Emperor 4

Deploy advanced, explainable AI across the ITOps toolchain.

IBM Z Anomaly Analytics

Available for IBM z16

Proactively identify operational issues in log and metric data.

IBM Z Integrated Accelerator for AI and IBM Integrated Accelerator for AI on LinuxONE

Available for IBM z16 and IBM LinuxONE Emperor 4

A 7 nm microprocessor engineered to meet the demands our clients face for gaining AI-based insights from their data without compromising response time for high volume transactional workloads.

IBM Z and LinuxONE Container Image Registry

Available for IBM z16 and IBM LinuxONE Emperor 4

Utilize a channel of open source container images available for use, free of charge, that can be pulled and managed through the common graphical and command line interfaces that support containerized workloads.

ONNX

Available for IBM z16 and IBM LinuxONE Emperor 4

ONNX is an open standard used for converting between different machine learning frameworks.

IBM Z Optimized for TensorFlow and IBM LinuxONE Optimized for TensorFlow

Available for IBM z16 and IBM LinuxONE Emperor 4

TensorFlow is a widely used end-to-end platform for machine learning and deep learning, including a large ecosystem of tools, libraries, and community resources.

There is an available open beta for IBM Z Optimized for TensorFlow. To request access, send an email to aionz@us.ibm.com.

PyTorch

Available for IBM z16 and IBM LinuxONE Emperor 4

PyTorch is an open-source machine learning library based on the Torch library, helping to accelerate the path from research prototyping to production deployment.

Keras

Available for IBM z16 and IBM LinuxONE Emperor 4

Keras is a Python-based deep learning library that functions as a high-level API specification for neural networks.

Anaconda Commercial Edition

Available for IBM z16 and IBM LinuxONE Emperor 4

Utilize the leading open-source data science platform, including open-source package distribution and environment management, all with IBM Z and LinuxONE workloads.

Infusing AI into applications

The Integrated Accelerator for AI offers seamless exploitation for the IBM Z runtimes, in that upgrades to the runtime environment should not be required, and applications that are already leveraging suitable deep-learning AI models deployed to IBM Z can benefit from the acceleration without change. It also opens up new possibilities for applications to incorporate AI in their processing, benefiting from the reduced latency, and using data that might only be relevant while the transaction is running.

The following sections explore how each of the application runtimes can invoke AI models deployed to the various different frameworks and environments.

This use case works well for applications running in:

  • CICS TS
  • IMS TM
  • IBM WebSphere Application Server for z/OS (Websphere traditional and WebSphere Liberty)

From the application, you can make a call that invokes an AI model deployed to the WMLz base running in z/OS or, in some cases, within the runtime itself. If you use WMLz Online Scoring Community Edition (WMLz OSCE), the model can be deployed to OSCE in z/OS Container Extentions (zCX) for scoring ONNX models. Note that this configuration using OSCE in zCX is for development or proof-of-concept purposes.

Depending on the runtime, there are a number of ways in which the WMLz scoring can be invoked:

  • Using CICS API commands: For CICS TS, the WMLz scoring engine can be hosted in a Liberty server within the CICS runtime, and invoked via an EXEC CICS LINK call that passes the data using CICS channels and containers.
  • Using Java API: For WebSphere traditional or WebSphere Liberty, a Java API can be used to call the WMLz scoring feature configured in the same WebSphere server as the application runs, or configured in a separate WebSphere server.
  • Using REST API: For IMS TM, and as an alternative for CICS TS and for WebSphere Application Server, the model can be invoked using a REST API.

This use case works well for applications running in:

  • CICS TS
  • IMS TM

An ODM rule driven by the runtime application can be enhanced to reference a model deployed to WMLz, and then use the prediction from the model in the rule. ODM uses a highly efficient interface between ODM and WMLz.

If the application already uses ODM rules, then a rule called by the application can be enhanced with machine learning to drive an AI model deployed to WMLz. If the application does not currently use ODM rules, it could be updated to use an ODM rule that drives the AI model via WMLz, and hence include additional insight in the result from the rule. This use case has the added benefit of transparency in the use of the AI prediction, due to it being visible in the rule.

This use case works well for applications running in:

  • z/TPF
  • CICS TS
  • IMS TM
  • IBM WebSphere Application Server for z/OS (Websphere traditional and WebSphere Liberty)

From the application, you can make a REST call to an AI model deployed in a framework such as IBM Snap Machine Learning (SnapML), TensorFlow, or PyTorch, that might be hosted in a z/OS Container Extentions (zCX) instance within the z/OS environment, or hosted in Linux on IBM Z. When using zCX, the call uses an optimized form of access within z/OS. When using Linux on IBM Z, the call can use Shared Memory Communications (SMC) for efficient access.

The REST APIs provided by the AI model can be driven from the runtimes in a number of ways, including:

  • Using z/OS Connect EE: CICS, IMS and batch applications can take advantage of z/OS Connect EE and the API requester functionality to drive the REST APIs.
  • Using Java libraries: WebSphere traditional and WebSphere Liberty applications can use Java libraries such as ‘Rest Client for MicroProfile’ to make REST API calls.
  • Using z/TPF REST consumer support: z/TPF applications can make REST calls by using z/TPF REST consumer support.
  • Using CICS API commands: CICS TS applications can use EXEC CICS WEB SEND and WEB RECEIVE commands.
  • Using the z/OS client web enablement toolkit: IMS applications can take advantage of the z/OS client web enablement toolkit provided with z/OS, using the HTTP/HTTPS protocol enabler APIs to invoke the AI model and the JSON parser to interact with, create, or parse the corresponding JSON payloads.
Technical resources Expert perspectives

Learn the value of leveraging AI on the IBM Z and LinuxONE platform

Read the blog
Expert perspectives

Learn how to jumpstart your experience with AI on IBM Z

Read the blog
Use Db2 for z/OS with SQL Data Insights for AML

Watch a demo on how AI applications running on IBM zSystem can tackle the money laundering problem in various ways including solving the scatter gather problem.

Watch the video
Use Watson Machine Learning for z/OS to detect fraudulent transactions

Watch this video to see a demo on real-time detection and the prevention of credit card fraud with AI on IBM zSystems.

Watch the video
Solving fraud scenarios in real time

Determine business insights during transactions using IBM z16, harnessing the scalability and speed needed to address fraud.

Read the brief
Solving anti-money laundering

Learn how to use AI application on IBM z16 to identify money laundering patterns and prevent them in real time.

Read the brief
AI on IBM Z & LinuxONE Community

Join a community of practitioners and experts using AI on the IBM Z and LinuxONE platform

Join the community
Data and AI on IBM Z

Transform and modernize your IT approach to turn valuable IBM Z data into business opportunities

See the value of Data and AI technology on IBM Z
IBM Db2 AI for z/OS

IBM Db2 AI for z/OS (Db2ZAI) is built on the services that are provided by Watson Machine Learning for z/OS to help optimize performance.

Watch the video
IBM Watson Machine Learning for z/OS Hands-on Lab

Learn how to do data visualization, model creation, deployment and evaluation, and scoring with WMLz.

Try the lab
Streaming SMF data real time through IzODA

Sample code and hints for how you could utilize a real time data consumption strategy with your IzODA applications using the SMF Realtime Interface on z/OS

Read the blog
IBM Open Data Analytics for z/OS Lab

Choose Your Path: Spark and Scala or Anaconda and Python

View the labs
AI on IBM Z 101

Find samples, containers, and other relevant resources for using AI on the IBM Z platform

Visit the site
GitHub repository: credit card fraud detection

View sample code for running a credit card fraud detection scenario with AI on IBM Z

View the repository
GitHub repository: sample containers for s390x environments

Access sample container build files for AI software that can be utilized in s390x environments.

View the repository
Github repository: sample demonstration code

Access small, useful examples to demonstrate some of the technologies available for use on IBM Z and LinuxONE systems.

View the repository
IBM Open Data Analytics View the documentation IBM Z Operations Analytics View the documentation IBM Z Application Performance Management Connect View the documentation IBM Operational Decision Manager View the documentation IBM Wazi as a Service View the documentation IBM Cloud Pak for Data View the documentation DB2 Analytics Accelerator for z/OS View the documentation IBM Watson Machine Learning for z/OS View the documentation Db2 AI for z/OS View the documentation Data Privacy for Diagnostics View the documentation IBM Watson AIOps View the documentation IBM Z Anomaly Analytics with Watson View the documentation Infusing AI into applications on IBM zSystems

Technical guide for planning the infusion of AI with applications running in CICS, IMS, WebSphere Application Server for z/OS, and z/TPF.

Read the guide
Using Watson Machine Learning for z/OS scoring service

Embedding the Watson Machine Learning for z/OS scoring service in a CICS region using the WMLz ALNSCORE program.

Learn more
Using Watson Machine Learning for z/OS OSCE

Redpaper on Optimized Inferencing and Integration with AI on IBM Z - shows a CICS application using a model in WMLz to predict a credit risk score

Read the Redpaper
Using ODM with WMLz

View a tutorial that steps through enhancing ODM rules with IBM Watson Machine Learning for z/OS predictions

View the tutorial
Using z/OS Connect and REST API

Learn how to call REST APIs from z/OS applications using z/OS Connect

Read the documentation
Using z/OS Connect and REST API

View a Github sample showing a REST API call from IMS using z/OS Connect EE

View the sample
Using z/OS Connect and REST API

View a Github sample showing a REST API call from CICS using z/OS Connect EE

View the sample
Interact with REST APIs provided by AI models

Use the z/OS client web enablement toolkit provided with z/OS to interact with the REST APIs provided by AI models

Read the documentation
Using z/TPF to call a REST service

Use the z/TPF client to call a REST service from a z/TPF application

Read the documentation
Related solutions z/OS Container Extensions (zCX)

Run Linux on IBM Z applications as Docker container images on z/OS.

Journey to hybrid cloud with IBM Z

Build your hybrid cloud with IBM Z® for data privacy, security, cyber resiliency and speed to modernization.

IBM Open Data Analytics for z/OS

Integrate data into a single source and share it on z/OS to simplify data analysis and development.

What's new Last updated September 13, 2022

The use cases and products/related solutions showcased on this page have been updated to reflect the announcement of the IBM LinuxONE Emperor 4.

Information about the open beta for IBM Z Optimized for TensorFlow has been added to the 'Featured open source communities' section.

The technical resources section has been updated to include a technical guide for planning the infusion of AI with applications running in CICS, IMS, WebSphere Application Server for z/OS, and z/TPF.

The page has been updated to feature a host of use cases broken out across the 'Get Started' section, with additional considerations for relevant runtimes provided.

Additionally, two new 'Learn more' accordions have been added, one featuring the various products and capabilities available for AI on the IBM Z platform, and the other featuring methods for implementing AI with relevant runtimes.

Lastly, a host of new resources have been added, including new demonstration videos for featured use cases, product documentation, and solution briefs.

AI and Data Science availability on IBM Z

Rate this content solution