While AI assistants generate the code, most IBM i shops have no pipeline to get it to production safely

AI coding tools are arriving on IBM i. IBM Bob launched last month. Anthropic’s Claude and ChatGPT have shown real capability with RPG. Early results are promising. However, as IBM i shops start evaluating these tools, a practical problem keeps coming up: the development foundation most shops are running wasn’t built for this.

AI tools are real, and they work…the tools are only as good as the pipeline behind them. If you don’t have a real DevOps foundation on IBM i, AI-generated code is just more risk.”

— Daniel Magid, CEO, Eradani

Eradani is at COMMON POWERUp 2026 in New Orleans this week to talk about that problem, and what to do about it (Booth #305).

AI coding tools need source code on the developer’s local machine to work. For shops running PDM-based workflows and traditional source physical files, that’s not how things work today. Code lives on the IBM i. Getting it local, working with it, and getting it back safely is a manual process.

Marketing Technology News: MarTech Interview with Haley Trost, Group Product Marketing Manager @ Braze

Many shops are internally divided. Experienced developers want to continue using PDM, the RDi Remote System Explorer, and the Code for IBM i VS Code plugin so they can work with their libraries, source files, and source members. Their change management tools might allow them to keep a Git repository up to date, but the libraries remain the source of truth. They are not taking full advantage of Git’s power as a developer tool. On the other hand, younger developers new to IBM i want to work with the latest open-source tools and expect to use Git.

“There’s a real difference between having code in Git and actually using Git,” said Dan Magid, CEO of Eradani. “Vendors have been pushing copies of RPG source to repositories for years and calling it Git support. You can’t take advantage of the powerful parallel development support of Git; you can’t easily branch your code, stash it, trigger a pipeline from it, or run a code review against it. You can’t view changes at the branch or developer commit level, or easily roll back changes to a specific tag. You can’t make your source changes locally on your PC and then use Git to sync them to your IBM i for compiles and execution. It’s a backup dressed up as DevOps.”

Marketing Technology News: From Data to Impact: How AI is Transforming Interactive CTV Ads

Why Eradani DevOps customers are already set up for AI

Eradani DevOps gives IBM i developers true, native Git. Real branching, pull requests, merge workflows, and CI/CD pipelines. It works with GitHub, GitLab, Azure DevOps, and Bitbucket. It connects to Jira, ServiceNow, and SonarQube. It’s the same Git that 100 million developers use everywhere else, running as it was meant to on IBM i.

That matters for AI in a specific, practical way. When developers work in real Git, they can have a local clone of their repository on their machine. Clone the repo, work locally, and push your changes. That local clone is what AI coding tools need.

Eradani DevOps customers aren’t blocked by the problem that’s frustrating other shops right now. With Eradani DevOps, developers can choose to work locally or use traditional tools. Either way, they get a full Git workflow. An AI coding assistant is embedded, not an additional project.

Getting AI-generated code to production safely

Without a real pipeline, AI-generated code flows straight from the developer’s screen to production. Manual checks worked when developers wrote 10-20 lines of code per hour. With AI tools generating thousands of lines in minutes, only automated pipelines can handle the load.

With Eradani DevOps, AI-generated code goes through the same process as everything else. Committed to a branch, reviewed in a pull request, scanned by tools like SonarQube, and deployed through Eradani iDeploy with full traceability. Nothing reaches production without passing through the same gates as every other change.

“The AI tools are real, and they work,” Magid said. “What we’re telling customers at POWERUp is that the tools are only as good as the pipeline behind them. If you don’t have a real DevOps foundation on IBM i, AI-generated code is just more risk sitting between your developer and your production system.”