Public Preview For Watson Code Assistant for i Available Soon
May 21, 2025 Alex Woodie
At the COMMON POWERUp 2025 conference this week, IBM pulled the covers back a bit on Watson Code Assistant for IBM i, the new name for the AI-powered coding assistant that IBM first started talking about one year ago. A public preview of WCA for i is becoming available to a limited number of testers, with general availability expected in the second half of the year.
There has been quite a bit of development in the IBM i coding co-pilot this month. On May 7, during the IBM Think conference, two IBM executives wrote a blog post announcing WCA for i, which was then in private preview. The vice presidents also posted a link for IBM i professionals to sign up for the limited public preview, which you can access here.
More information on WCA for i came out during this week’s COMMON POWERUp 2025 conference in Anaheim. IBM i chief technology officer Steve Will and Edmund Reinhardt, the IBM i application development tools lead for IBM, presented multiple sessions on WCA for i. The sessions reveal that the RPG coding co-pilot plans first revealed at POWERUp 2024 one year ago are mostly intact, but challenges remain.

WCA for i is built on watsonx and Granite models.
There were three main goals for the product a year ago: Code explanation, code generation, and test case generation. The beta release of WCA for i, which will start in July, will satisfy the first of those goals, code explanation, while the remainder are still on the roadmap for the end of the year.
The software is on track to satisfy the four types of code explanation, including detailed explanation for a programmer, a business-level summary, an explanation of usage, and then documentation that can either inserted into the source code itself or extracted for greater summary.
“It will give you different options for how to explain it,” Will said. “Say you’re a new developer and you’d like to be able to ask an RPG expert what’s going on. Well, you can select then to have a detailed explanation of the code that you are looking at. And it will then explain that to you. And you don’t have to build the prompt. We build that for you and it’ll come back with a detailed explanation in the chat window.”
The preview of WCA for i functions as a plug-in for VS Code and utilizes IBM Granite models running in the cloud or on prem. There currently are no plans to support WCA for i on RDi, according to Will.
IBM has fed more than 10 million lines of RPG code into the Granite models, according to Reinhardt. The company is asking for more RPG code to get the product where it needs to be. In particular, IBM wants old RPG II, RPG III, and RPG OPM (original program model) source code from customers to make WCA for i better.
Most of the old RPG code that needs attention is fixed-format RPG and non-ILE, and so that is what IBM is asking customers to contribute. The older and uglier the code, the better for AI training. “I need code to be able to make this work,” Reinhardt said. “That’s exactly what we need, embarrassingly ugly code.”

This is the logo for the WCA for i product
In addition to helping newbie RPG programmers, WCA for i will help veteran programmers by reducing the learning curve, Will said. “WCA for i allows these developers to understand code they’ve never seen, and they can do it faster and more completely than if they spent much more time getting that knowledge on their own,” Will wrote in his You and i blog on TechChannel. “It’s as if they have an experienced programmer helping them – assisting them – in the task.”
The WCA for i product is not yet ready for primetime. Some of the functionality failed to work properly during a live demo by Reinhardt on Tuesday. The AI product will hallucinate, as all large language models will. “It’s not something you trust 100 percent,” Reinhardt said. But the code explanation is still useful in being able to quickly understand and document large amounts of code in a relatively short amount of time, he said.
In addition to old RPG code, IBM wants CL and COBOL, which it will eventually use to train WCA for i. It also wants examples of customer SQL code. While IBM is proud to state that its iteration of SQL for IBM i adheres to more of the ISO standard than other SQL implementations like Oracle and SQL Server, it is not necessarily the same as the other versions of SQL that mainstream LLMs have been trained on and which are widely available on LLMs in the open market. “SQL on i is the same as SQL, but not, so we need SQL on i,” Will said during the demo yesterday morning.
Most customers will run WCA for i in the cloud, but there will also be an option to run on prem, which will likely be used by customers in highly regulated environments. According to Will, IBM requires customers who want to run the other WCA products on prem to have an X86 server equipped with OpenShift and Cloud Pak for Data. That will be a non-starter for many IBM i shops, he said, so the company is exploring ways to get WCA for i to run natively on IBM i. There is a possibility that the Spyre accelerator that is being built as an adjunct to the Power11 chip will provide the computing resources necessary to accomplish this and that the matrix engines inside of Power11 and Power10 chips can do this.

A screenshot of the new WCA for i product that IBM shared at POWERUp 2025
IBM is indemnifying organizations and people who donate their RPG source code to IBM. To further convince people to donate their code, IBM is housing RPG is two repositories, one that’s visible to all project contributors and one that’s only visible to IBMers.
As the public beta for WCA for i nears, two things are clear. First, the age of AI co-pilots is upon us, and having a co-pilot for IBM i will soon become a necessity to remain competitive. Secondly, the scope of the WCA for i will likely expand in the future. In fact, IBM is considering how best to adapt to the quickly emerging world of agentic AI in a way that makes sense for its customers.
“I have grown to understand better all the different things people wanted from it so in that sense, it’s expanded,” Will told IT Jungle in an interview at POWERUp yesterday. “But I knew ahead of time I didn’t know enough. We’ve never built anything like this before, right? And we knew we needed to learn how to train a model to do these things. And we had been told we will want to train it to do one thing first, get it to be pretty good, and then figure out how to train it to do another thing. And so that’s one of the reasons we’re focusing on explain right now. And it’s actually going better than we thought. But it’s going to take a little while before we get to some of that generate stuff.”
IBM is still hoping to have RPG generation capabilities available in WCA for i by the second half of the year. The fall Technology Refresh (TR) might be a good time to announce it, although it might take a couple of months more, Will said. But one thing is for certain: IBM’s AI fortunes won’t be confined to what can be packaged and released twice a year.
“We’re not going to wait for another year and then do another big GA,” Will said. “We’re just going to roll out these things as we train the model and create a user interface to use that.”
You can track the development of WCA for i on this GitHub site. IBM also has a product page for the product here.
RELATED STORIES
Beta For RPG Coding Assistant On Track for 2Q25
RPG Code Assist Is The Killer App For AI-Enhanced Power Systems
IBM Shows Off Next-Gen AI Acceleration, On Chip DPU For Big Iron (The Next Platform)
IBM’s AI Accelerator: This Had Better Not Be Just A Science Project (The Next Platform)
Some Thoughts On Big Blue’s GenAI Strategy For IBM i