Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Mainframe developers can use AI to explain code, draft routines, suggest optimizations, and support selected modernization tasks—but should treat it as an assistant, not an authority. The practical choice is not whether to let AI write production code unchecked; it is whether a specific task can benefit from a draft or explanation that a developer can verify against requirements, security rules, and tests.
What can AI do for mainframe development?
Mainframe-focused tools describe capabilities for working with COBOL and other established languages, alongside broader assistance for modernization. The useful boundary is task-specific: ask for help with an understandable, reviewable unit of work, then check the result in the context of the application and its operating environment.
Explain unfamiliar code
A developer can ask an assistant to explain a COBOL or JCL segment, summarize its apparent control flow, or clarify how a section relates to a stated input and output. Treat the response as a reading aid, not as authoritative documentation: compare it with the code, surrounding modules, job definitions, and any available specifications.
Draft code from a clear specification
AI can produce a first draft of a routine from a precise description of inputs, outputs, edge cases, and constraints. The developer still needs to check that the draft follows local conventions, handles data and error conditions correctly, and fits the program’s existing interfaces. A plausible-looking routine is not proof that the requirements were understood.
#1 Best Overall
- Murach's Mainframe COBOL
- Mike Murach & Associates
- ABIS BOOK
Explore optimization and modernization
An assistant can suggest possible improvements or help teams investigate selective modernization. IBM Research describes an IBM Z approach involving application discovery, dependency analysis, refactoring support, and automated testing in a selective COBOL-to-Java workflow. IBM’s earlier announcement also describes discovery, refactoring, transformation, and validation as lifecycle stages. These are descriptions of IBM’s product approach, not independent evidence that a transformation is correct or will save a particular amount of time.
What do the available results say about productivity?
The evidence does not establish a dependable productivity increase for mainframe developers specifically. IBM Research’s 26 April 2025 enterprise study examined IBM’s internal deployment of watsonx Code Assistant using surveys of two user cohorts (669 people in total) and unmoderated usability testing with 15 participants. The researchers reported that assistants can deliver net productivity increases, but not all users experience those benefits. The study describes user experience in one company; it is not a controlled mainframe-only trial and does not provide a causal productivity estimate for mainframe coding.
Rank #2
IBM reported more than 250 completed or ongoing generative AI client projects in its 2025 z17 overview. That is IBM’s project count, not a measurement of coding speed, code quality, or mainframe developer productivity. Neither figure justifies promising a percentage gain to an individual team.
Free tools Windows power users keep installed
One-click scans. No signup required.
How should a developer use AI-generated mainframe code?
Keep a developer responsible for requirements, design decisions, security review, and validation. A safe workflow makes the assistant’s contribution easy to inspect and keeps generated code out of production until it passes the same engineering controls as other code.
- Choose a bounded task. Start with an explanation, a small draft routine, or an optimization idea rather than an open-ended request to rewrite a business-critical application.
- State the constraints. Include the intended behavior, interfaces, edge cases, relevant language and platform context, and applicable local coding standards. Do not supply sensitive or regulated information unless the tool’s approved data-handling terms and organizational policy permit it.
- Review the output line by line. Confirm that it meets the requirement, fits the application’s dependencies and conventions, and does not introduce unsafe behavior. For explanations, verify the interpretation against the source code and surrounding context.
- Validate with the normal engineering process. Compile or otherwise check the code in the target environment, run appropriate unit and integration tests, and evaluate behavior against representative cases. For changes involving business logic or dependencies, use review and testing that cover those impacts—not only whether the generated code appears syntactically plausible.
- Record decisions and provenance where policy requires it. Make clear which parts were generated or materially assisted, and preserve the review and test evidence expected by the team’s change-control process.
What should a team evaluate before adopting a tool?
Compare tools against the work the team actually needs to do. Vendor feature descriptions can identify what a product claims to support, but they do not substitute for an evaluation in the team’s own environment.
- Language and task coverage: Check the exact languages, code explanation, generation, and modernization functions supported in the version under consideration.
- Workflow fit: Confirm compatibility with the team’s IDE, repositories, review process, and mainframe development practices.
- Deployment and data handling: Understand where prompts and code are processed, what controls apply to data, and whether the deployment meets organizational requirements.
- Traceability and review: Determine how developers can inspect suggestions, understand their context, and document changes.
- Testing support: Check whether the tool assists with test creation or execution, and distinguish that assistance from proof that the code is correct.
- Governance and security: Set acceptable-use rules, access controls, and review requirements. IBM’s general developer guidance identifies inaccuracy, overreliance, security vulnerabilities, and the need for more comprehensive code review as relevant risks.
- Evidence of outcomes: Run a bounded pilot with representative tasks and define what to measure—such as review effort, defect findings, or task completion—before drawing conclusions about value.
How mature are the mainframe-specific options?
Capabilities and availability depend on product and version, so verify current documentation before choosing a tool. IBM announced that watsonx Code Assistant for Z version 2.6 would be available to clients starting 27 June 2025. In that announcement, IBM described natural-language COBOL generation through a VS Code chat interface and context-aware inline suggestions; it said code explanation supported COBOL, JCL, PL/I, and REXX, with Assembler explanation in preview. These are IBM’s product statements, and preview status or availability may have changed since the announcement.
IBM’s z17 overview also described AI-supported development and IBM Test Accelerator for Z, including planned natural-language test generation and AI in unit and integration testing. Planned functions should not be treated as generally available without checking current product documentation. More broadly, an assistant that generates or explains code and a modernization workflow that analyzes dependencies solve different problems; evaluate each against the task rather than assuming one tool covers the whole lifecycle.
Recommended Free Tools
So, should mainframe developers code with AI?
Yes, when AI is used for bounded assistance and developers retain control of specification, review, security, and testing. It can help make unfamiliar code easier to inspect and provide drafts or ideas worth evaluating. The evidence does not support treating AI as a replacement for mainframe expertise or promising a universal productivity gain. Adopt it where the team can verify its output and measure whether it helps.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
