At Proximity, we’re always on the lookout for new technologies that can help our clients modernise, maintain and develop their IBM i applications so they can get even more value out of their IBM i investment.
Recently, some of our IBM i development team attended the iPower conference, hosted by i-UG, where they had the opportunity to experience IBM Bob, an AI-powered software development assistant tailored specifically for the IBM i ecosystem.
Following the workshops, our developers, Darren and Dylan, took IBM Bob out for a spin in a real-world trial environment.
We sat down to discuss our findings, what we learned, and how IBM i users can responsibly start experimenting with AI.
Read on below for our hands-on review of IBM Bob.
Where IBM Bob shines: System context and modernisation
Generic AI coding assistants like ChatGPT are great for general-purpose code, but they fall short when it comes to the highly specialised nuances of the IBM i language base.
That is where IBM Bob brings a significant shift to the IBM i community.
As Bob has been developed by IBM, and specifically for the IBM i, it is deeply integrated into the IBM i file system (IFS), so it not only knows where to look, but it also knows what to look for.
It understands system dependencies, library structures, and the context of your specific environment.
During our trial, we discovered three key areas where Bob could add immense value to an IBM i development team:
- Code conversion and modernisation: One of the first tasks we threw at it was converting legacy RPG III code into modern RPG ILE. Bob handled this with, quite frankly, impressive accuracy, making it a potentially powerful tool for IBM i modernisation projects.
- Legacy documentation: If you have business-critical programs that haven’t been touched in 15 years, Bob is incredibly useful for parsing the logic. It can review a legacy program, identify the other programs it calls, list the physical or logical files it interacts with, and generate comprehensive documentation to help developers understand what the system is doing.
- SQL optimisation: When asked to look at our database interactions, Bob accurately suggested improvements to optimise our SQL statements. It predicted a 50% to 60% performance improvement, and when we ran the update, the benchmark numbers proved it right!
It is worth noting that Bob could, for instance, be used in conjunction with Fresche’s X-Analysis suite for application documentation.
Alternatively, Fresche has added AI capability natively into X-Analysis, which is a new development to the already impressive X-Analysis tool.
The challenges: Pushing the boundaries of generative code
It wasn’t all seamless, however.
When we pushed Bob past analysis and optimisation into completely writing or modifying complex chunks of execution code, we did manage to break it.
Using these AI capabilities requires “Bob Coins” (the platform’s token system).
When Bob got things wrong, we had to prompt it to fix its own errors, which quickly burned through our token allocation.
We even jokingly asked it for our Bob Coins back, but unfortunately, the AI couldn’t help us there!
For this reason, we don’t see it completely writing new code from scratch just yet, though we are highly anticipating the new Premium version to see how these capabilities mature.
A roadmap for IBM i teams: How to get started
If your organisation is looking to trial IBM Bob or similar generative AI tools on the IBM i, we highly recommend a cautious, phased approach:
- Start with read-only permissions: Do not give an AI tool write access to your code repositories or production environments right from the start. Start by configuring its permissions to ‘read-only’.
- Focus on analysis first: Use the tool to analyse your codebase, map out workflows, and document business logic. Let it show you how it understands your files and system structure before you let it modify them.
- Build team experience: Allow your developers to build familiarity with the IDE extension (it runs comfortably inside VS Code for i) in a safe sandbox environment before integrating it into active development workflows.
Accelerate your IBM i modernisation with Proximity
AI tools like IBM Bob are shifting the landscape of what’s possible, but maximising their value still requires seasoned IBM i expertise.
We’ve been using various AI tools within our IBM i support and development team for the past year-or-so.
The challenges we identified with Bob are not unique and should be a cautionary tale for any IBM i shop considering using AI within their toolset.
AI shouldn’t be used as the only or main tool for IBM i projects.
It should be seen as a tool to speed up the process. You simply can’t escape the fact that without a skilled IBM i development resource.
Tools such as IBM Bob, no matter how good they are, could get you into a lot of trouble, and fast.
Whether you are looking to modernise legacy RPG code, document decades of technical debt, or simply need expert support to keep your core systems running at peak performance, Proximity is here to help.
Ready to transform your green screens or optimise your IBM i applications?
To learn how our development and support services can future-proof your business, contact Proximity today
