AI Control Interface Simplify your IT processes with intelligent infrastructure. Get started - Play video

You don't need to be a mainframe expert to get more from your Z operating system. The AI Control Interface for IBM z/OS (AICI) provides an approachable platform where you can leverage prebuilt AI models to automate z/OS system management tasks.

The AICI currently supports the following use cases:

AI-powered WLM batch initiator management, offered by z/OS Workload Management, proactively adjusts WLM batch initiators ahead of recurring workload spikes. By optimizing resource allocation, this use case can help reduce job queue time by up to 70%.1

AI-powered network outbound packet batching, offered by z/OS Communications Server, uses network traffic patterns to make dynamic batching adjustments, supporting reduced network latency and lower network CPU usage.

Get started with the AICI to harness the capabilities of an AI-infused z/OS.

 See what AI Control Interface can do for your business. Big picture 1. Install AI System Services for IBM z/OS. 2. Configure AI Framework for IBM z/OS. 3. Select and configure the AI use case that best aligns with your business and operational goals. 4. Use the AI Control Interface for IBM z/OS to manage and monitor your AI capabilities. How to get started
Overview

AI Framework for IBM z/OS provides the foundational components that enable AI-infused capabilities on z/OS — including data collection, AI model lifecycle management, and AI inferencing and scoring. With z/OS AI Framework, organizations can easily set up and manage AI capabilities, without needing AI or data science expertise.

A key component of z/OS AI Framework is AI System Services (AISS). AISS is a zero-charge offering that supports key AI lifecycle phases including data ingestion, model training, inference, model quality monitoring, and retraining services. AISS seamlessly integrates with other z/OS AI framework components to enable the operation of prebuilt AI models on z/OS.

After configuring z/OS AI Framework and the use cases, you can leverage the AICI to train, simulate, enable, and monitor AI-powered use cases on your systems.

To learn more, read Overview of the AI Framework for IBM z/OS.

In order to leverage z/OS AI Framework, you will need the following:

Hardware:

  • IBM z14 (all models) or higher
  • Coupling facility (internal or external; required by EzNoSQL/VSAM record level sharing (RLS))

Software:

  • IBM z/OS 3.1 or higher
  • AI System Services for IBM z/OS (Software PID #5655164)
    • AISS bundles two key components: IBM Z Common Data Provider (zCDP) and Machine Learning for IBM z/OS Core Edition.

Note: Ensure that AISS is installed and configured on the same z/OS system where z/OS Management Facility runs.

For the full list of installation hardware and software requirements, read Hardware and software requirements.

For more detailed installation guidance, read AI System Services for IBM z/OS preferred ordering and packaging scenario.

To learn more about workflows, read About workflows.

To review the procedure, read Using the AI Framework for IBM z/OS Configuration Workflow.

z/OS Workload Management z/OS Communications Server

Overview

z/OS Workload Management provides AI-powered WLM batch initiator management.

Traditionally, z/OS Workload Management dynamically manages batch initiators according to performance data and pre-defined goals, responding to workloads as they arrive.

With AI, z/OS Workload Management can analyze historical batch workload patterns to identify recurring workload spikes. As a result, it can proactively adjust the number of WLM batch initiators before an upcoming workload spike. By predicting and responding to workload behavior, the use case is designed to help organizations more quickly process workloads during peak demand periods, without needing manual tuning or expert intervention.

To learn more about the use case, read AI at the Heart of IBM z/OS 3.1 to Simplify and Optimize.

System programmers can leverage AI-powered capabilities in z/OS to predict workload demands and optimize resource allocation.

Prerequisites:

  • z/OS AI Framework, configured with z/OS Workload Management
  • z/OS 3.1 or higher
  • AI System Services for IBM z/OS
  • An internal or external coupling facility for EzNoSQL record level sharing
  • VSAM RLS for EzNoSQL

Configuration:

For more details on how to get started, refer to AI-powered WLM batch initiator management.

After configuring AI-powered WLM batch initiator management, system programmers can use the AICI to monitor, train, simulate, and enable AI capabilities.

The AICI presents a clear view of which systems are Ready to train, Active, or Inactive, allowing system programmers to quickly assess workload status and AI readiness.

  • Train: Before applying AI capabilities, all systems must first be trained. Training analyzes existing workload data and behavior to predict upcoming workload spikes.

    Requirement: You need at least 30 days of contiguous training data before enabling AI-powered WLM batch initiator management.

  • Simulate: To see how the AI works before enabling, you can enter simulation mode, which compares results between AI-powered and traditional batch workload management.
  • Enable AI: After training, AI-enhanced batch initiator management can be applied to eligible service classes. Using the workload patterns learned during training, AI automatically adjusts the number of batch initiators to meet the demand. By optimizing your system resources, AI can reduce job backlog, enabling your critical batch workloads to finish on time.

Refer to Controlling and managing AI-powered WLM batch initiator management for more guidance on managing AI capabilities in the use case.

Overview

z/OS Communications Server offers AI-powered network outbound packet batching.

Traditionally, z/OS Communications Server batches outbound packets before sending them to the network interface — optimizing packet sizing and delivery based on current network conditions. Since these decisions rely on real-time traffic analysis, network changes must first be observed before the batching can be adjusted.

With AI, z/OS Communications Server can use training data and historical traffic patterns to anticipate future network conditions before they occur. By intelligently predicting and anticipating network changes in advance, AI-powered network outbound packet batching is designed to enhance application responsiveness, reduce CPU utilization, and accelerate operations.

To learn more about the use case, check out AI-powered network outbound packet batching explained.

System programmers can enable AI-powered capabilities in z/OS Communications Server to intelligently forecast future network conditions.

Prerequisites:

  • z/OS AI Framework, configured with z/OS Communications Server
  • z/OS 3.2 or higher
  • AI System Services for IBM z/OS
  • An internal or external coupling facility for EzNoSQL record level sharing
  • VSAM RLS for EzNoSQL
  • OSA-Express / Network Express interface support

Restriction: Only OSA-Express and Network Express network interfaces (OSD and OSH CHPID types) are supported for AI-powered network outbound packet batching.

Configuration:

For more details on how to get started, refer to AI-powered network outbound packet batching.

To help system programmers quickly understand workload status and AI readiness, the AICI offers an at-a-glance view of which systems are Ready to train, Active, or Inactive. With clear status visibility and guided actions, you can confidently move workloads from training to AI-powered optimization.

  • Train: Before applying AI capabilities, all systems must first be trained. Training analyzes existing workload data and behavior to predict when to send network packets and how to batch them efficiently.

    Requirement: You need at least 28 days of training data before enabling AI capabilities.

  • Enable AI: After training, you can apply AI-enhanced outbound packet batching to eligible TCP/IP stacks. Using the workload insights learned during training, AI helps z/OS Communications Server make automatic and informed packet-batching decisions. Rather than reacting to changes in traffic patterns, AI proactively optimizes network behavior to improve efficiency, responsiveness, and performance for your most business-critical applications.

Refer to Using AI Control Interface for z/OS to manage AI-powered network outbound packet batching for more guidance on managing AI capabilities in the use case.
Documentation IBM Documentation

Read more about z/OS AI Framework.

Information about z/OS AI Framework in IBM Documentation. Technical resources IBM z/OS 3.1 AI-powered WLM batch initiator management Resource and Tuning Guidelines

Find information for planning and implementation of the AI-powered WLM batch initiator management offering.

Learn about AI-powered WLM What's new

Page structure was changed, various sections were edited, links were updated, and new content was added.

Page was restructued from AI Infusion into z/OS to AI Control Interface. Various sections were edited and new content was added. z/OS Communications Server use case was added.

Links to documentation were updated for currency.

The notification about support for AI System Services for IBM z/OS 1.2 in the Introduction was updated.

A notification about support for AI System Services for IBM z/OS 1.2 was added to the Introduction.

A link to a related solution was added for the EzNoSQL content solution.

 Footnotes

1 DISCLAIMER: The performance of WLM managed batch initiators with AI inferencing active was compared with that of WLM managed batch initiators with the AI inferencing disabled. The workload was a synthetic batch workload in which 100 jobs in a WLM service class were scheduled to be submitted at the top of each hour while other background batch work ran in other WLM service classes. The system under test was an LPAR on an IBM z17 model 793 running z/OS 3.2 and the AI Framework with AI-powered WLM batch initiator management. Measurements were collected using a model trained based on 30 days of workload history and conducted in a controlled environment. Results may vary.