Summary

• BMC has introduced AMI Agentics for Data and Ops, combining mainframe expertise with agents that investigate Db2 and IMS problems, CPU behaviour and operational alerts.
• Control-M's October capabilities include AI-assisted monitoring, Data Assurance and additional integrations. The Control-M MCP Server is generally available on SaaS, while Data Assurance is announced for both SaaS and self-hosted environments.
• Synthetic COBOL test generation and selected integrations with AI coding assistants remain in beta. BMC has not published independently verified productivity or cost savings from the new agent capabilities.

Enterprise software company BMC Software is bringing specialised AI agents into mainframe administration and expanding the automation controls used to coordinate applications, data processing and business workflows.

Announced on 20 October, BMC AMI Agentics for Data and Ops provides assistance with database performance investigations, operational alerts and established mainframe automation. The company is also introducing new Control-M capabilities for data quality, workflow monitoring and connecting AI applications with the systems on which businesses depend.

The developments are intended for environments where computing systems must operate predictably and where incorrect changes can affect important services. BMC is placing AI assistance alongside established permissions, audit records and orchestration procedures rather than presenting agents as unrestricted replacements for experienced technical staff.

The announced functionality has different availability statuses. Control-M Data Assurance is described as available across SaaS and self-hosted environments, while selected mainframe development functions remain in beta. Additional product documentation identifies native SaaS integration and a generally available Control-M MCP Server, requiring those capabilities to be distinguished rather than presented as one uniform release.

Mainframe operations retain specialised requirements

Mainframe computers continue to support high-volume transaction processing and established business applications used by large organisations, including banks and other institutions operating services that require consistent availability.

Maintaining those environments involves understanding operating systems, database structures, workload behaviour and automation procedures that may have developed over decades.

Although some organisations are modernising applications or connecting mainframes with cloud services, they must continue operating critical processes while changes are evaluated and introduced.

That can create pressure on specialist teams, particularly where technical knowledge is concentrated among experienced employees familiar with the history of an installation.

BMC AMI Agentics for Data is intended to make relevant information easier to examine through specialised agents working with established mainframe database tools.

The announced capabilities include AI Reorg Advisor, SQL Workload Analyzer for IBM Db2, Deadlock Advisor in BMC AMI Log Analyzer and IMS Exception Advisor in BMC AMI Database Advisor.

These tools address different administrative problems. Database reorganisation can affect performance and storage efficiency, while analysing SQL workloads may help engineers identify queries or operating conditions associated with unexpected resource consumption.

Deadlocks occur when database operations compete for resources in ways that prevent normal progress, requiring administrators to determine which transactions or processes are involved.

AI assistance could reduce the effort needed to assemble diagnostic information, but identifying a plausible explanation is not the same as establishing the correct operational response.

Engineers must still examine the evidence behind recommendations, particularly before changing database structures or processing arrangements affecting production transactions.

Operational agents investigate complex failures

BMC AMI Agentics for Ops extends similar assistance into systems administration and operational investigation.

The company says the agents can help teams examine increases in CPU consumption, interpret alerts and understand established REXX automation.

REXX is a scripting language used in mainframe administration and operational procedures. Scripts developed over many years may contain important business logic or system management instructions that are not fully documented elsewhere.

An agent capable of explaining those routines could help staff investigate unfamiliar code, especially when the original developers are no longer available.

However, software explanations must be checked against actual behaviour. A script may interact with other applications, depend on particular operating conditions or make changes whose consequences are not apparent from its source alone.

Investigating CPU usage involves similar uncertainty. Increased resource consumption may coincide with an application fault without being its underlying cause.

Diagnostic agents need information about workload history, scheduling, operating conditions and related events before they can provide useful recommendations.

BMC says its approach uses mainframe operational context and specialist software knowledge rather than depending entirely on general model responses.

The commercial benefit will depend on whether that context produces more reliable investigations and reduces the work required from engineers, while preserving their ability to reject incorrect conclusions.

No independently audited comparison of incident resolution times has been supplied for the newly introduced AMI Agentics capabilities.

Developer functions retain beta status

Other additions target developers maintaining mainframe applications.

BMC AMI DevX Unit Test can examine COBOL source code and copybooks within Visual Studio Code to generate synthetic test information and potential test scenarios.

Testing established business software can require substantial preparation because applications may depend on complicated record structures and relationships between different data fields.

Synthetic test generation could help developers prepare representative inputs without routinely copying sensitive production information into development environments.

However, generated cases must still exercise the business rules and failure conditions relevant to the application.

A test can execute successfully without demonstrating that all consequential behaviours have been examined, particularly where a program contains unusual historical requirements or interacts with other systems.

The synthetic test generation capability remains in beta and should not be presented as generally available across every customer environment.

BMC is also testing integrations connecting AI coding assistants with AMI DevX Code Pipeline.

Supported assistants include Cursor, GitHub Copilot and Claude Code, allowing developers to work with code generation and deployment activities through familiar environments.

The intended integration preserves the user’s existing mainframe permissions and records relevant pipeline activity through established audit arrangements.

That may help organisations introduce AI coding tools without creating an entirely separate change management process.

Nevertheless, generating a code change does not establish that it is suitable for production. Applications handling financial transactions or other critical business activity require testing, review and authorisation before modified software is deployed.

Techopia previously examined IBM’s approach to connecting coding agents with controlled infrastructure. BMC’s latest development concerns separate mainframe tooling, specialist diagnostics and its own deployment pipeline.

Control-M extends automation into AI workflows

Control-M coordinates activities that must be completed in a defined order, including application jobs, data processing and services operating across several computing environments.

Such systems can manage dependencies between tasks, monitor whether work has completed successfully and support recovery when a process fails.

AI applications introduce additional activities that may depend on the outputs of established business software.

A reporting process might gather information from databases, perform validation and submit the resulting records to an AI application for further analysis.

If an earlier stage produces incomplete or incorrect information, later software may generate unreliable results even when the model processes the supplied material as designed.

BMC’s Data Assurance capability adds validation within orchestrated workflows, helping organisations identify problems before information reaches downstream reporting or AI systems.

The company’s 20 October statement describes the capability as available across Control-M SaaS and self-hosted deployments.

Its accompanying October product commentary specifically identifies Data Assurance as native within Control-M SaaS. Those statements are compatible with differing deployment arrangements, but the precise feature scope and configuration for self-hosted installations should be confirmed through product documentation.

Data validation also depends on the rules organisations establish. Checking whether information is present or correctly formatted does not necessarily establish that it is factually accurate or suitable for the decision being made.

Control-M’s new AI monitoring capabilities allow operations staff to use natural language to investigate jobs, examine dependencies and understand potential causes of workflow failures.

Such assistance could reduce the effort required to assemble information from several monitoring systems, although the resulting explanation must reflect the actual state of the workflow.

Agent access remains governed by permissions

BMC’s October Control-M release also makes the Control-M MCP Server generally available for Control-M SaaS.

The server allows compatible AI agents and assistants to discover and use supported Control-M tools through the Model Context Protocol.

Depending on their permissions and the configured controls, agents can obtain job information, examine logs and initiate approved workflow operations.

The interface is intended to preserve existing security permissions and audit arrangements, with mechanisms supporting human confirmation for consequential actions.

Such controls are important because an agent interacting with operational systems may be able to affect processes beyond the application in which the original request was entered.

Restarting a failed job, for example, may be appropriate for a routine process but unsuitable where repeating the activity could duplicate a financial transaction or alter business records.

Organisations must establish which operations can be performed automatically, when confirmation is required and how unexpected behaviour can be interrupted.

BMC is extending Control-M integrations with Amazon Bedrock AgentCore, Argo Workflows, SAP Build Process Automation, Workato, IBM MQ, Oracle JD Edwards EnterpriseOne and cloud infrastructure services.

Those connections are intended to coordinate work across application, data, infrastructure and AI platforms without requiring every integration to be implemented through custom scripts.

However, connector availability does not establish that all supported services share the same permissions, recovery procedures or operational guarantees.

Predictability remains the deployment requirement

BMC also uses LangGraph within AMI Agentics to coordinate specialist agents, while Control-M supports LangGraph-based workflows through its orchestration capabilities.

Combining these technologies could make it easier for businesses to connect AI-assisted investigation with established operational processes.

The arrangements still require clear boundaries between examining a problem, recommending action and executing a change.

Enterprise systems depend on reliable processing, including predictable handling of failure and recovery. An AI model may offer several plausible answers to a problem, while the systems implementing a business transaction often require an unambiguous and auditable result.

That difference creates requirements for testing and governance, particularly when agent activity involves sensitive information or applications supporting regulated services.

Operators must also consider the cost of integrating additional software, training staff and maintaining technical connections as products change.

BMC has not published independently verified customer savings attributable to AMI Agentics or the new Control-M functionality.

The October release establishes additional tools for applying AI within controlled enterprise environments, but the commercial case will depend on whether they reduce administrative and investigation work without undermining the reliability of critical applications.