An AI coding assistant only earns its place in the Software Development Life Cycle (SDLC) after a developer has tried to break it. That is why we sent two of our software developers to IBM’s Bobathon in May 2026, a hands-on developer day where IBM invites teams to stress-test Bob against real coding problems. Reema Sharma and Dennis Baffour spent a day with IBM Bob. They tested it on the kind of work they would normally do by hand. The short version: it held up. In a few places, it surprised them.
Bob is IBM’s enterprise AI assistant for software development. In other words, it is not just an autocomplete tool inside the editor. It sits across requirements, design, build, test and CI/CD.
What our developers actually tested
The Bobathon was hands-on, not a sales presentation. Reema built a working application from end to end. She walked Bob through requirements, design, code generation, unit tests with MOQ, and the CI/CD pipeline. As a result, within a couple of hours she had a functioning app that touched every stage of the SDLC.
Dennis took a different approach – he asked Bob to transform a sample SOAP web service into a modern .NET API using only a WSDL that he generated on the fly. Bob produced a full migration pathway, deployment steps, and supporting documentation. In addition, the team watched Bob lift a COBOL green-screen system into a modern web application. That kind of work usually takes weeks of careful reverse engineering.
Where this kind of AI coding assistant lands hardest
After the Bobathon, our developers were clear about where this category of tool pays back. In their view, day-to-day editor productivity is well served by Copilot, which has strong code context inside Visual Studio. By contrast, Bob’s real strength is somewhere different. Bob provides a far broader capability by extending AI support across the full SDLC lifecycle, helping to improve end-to-end delivery efficiency and automation. Bob is also particularly strong at legacy transformation and reverse engineering. Lifting old SOAP services into .NET. Pulling a COBOL system out of a green-screen and into a browser. Generating documentation that did not exist when the code was first written.
That distinction matters. In practice, the hype around AI coding assistants often treats them as one category. However, they are not. Some are autocomplete on steroids. Others, like Bob, sit higher up the stack and reason about whole systems.
Security flagged at build time, not bolted on later
The feature Dennis kept coming back to was Bob’s ability to identify and highlight security flaws as part of the generation process. In short, security as a build-time concern, not a post-build audit.
For anyone who has worked on retrofitting security into legacy code, the value is obvious. Picking up an insecure pattern at the moment it is written is cheaper, by orders of magnitude, than picking it up six months later in a pen test.
In addition, this is the part of the Bob proposition that deserves more attention. It is not just an accelerator for writing code. It is also an accelerator for writing code that does not fail a security review.
What this means for IT and engineering leaders
A few practical points from the day, for any leader weighing up an AI coding assistant.
- The real skill shifts. Reema and Dennis both said the same thing. With an AI coding assistant in the loop, the differentiator is how well a developer defines requirements, plans the approach, and writes good prompts. That is a craft, and it can be taught.
- Tools are not interchangeable. Copilot for productivity inside the editor. Bob for transformation, modernisation, and reverse engineering. In practice, picking the right one for the job matters.
- Cost models differ. Bob is a metered enterprise platform, not a flat SaaS seat. As a result, usage scales with SDLC tasks rather than with headcount.
- Security and AI assistance belong together. The strongest argument for enterprise-grade over consumer-grade is exactly this. Secure by default, at the point the code is written.
What our Developers had to say
“I was particularly impressed by its ability to identify and highlight security flaws. This is an area I’ve worked on previously and know can be challenging, so I really value this feature.”
Dennis Baffour, Northdoor developer
“What was particularly impressive was that within a couple of hours, BOB was able to support and streamline the entire process very effectively, from requirements through to development and testing. It demonstrated how AI can significantly improve productivity, accelerate delivery, and help maintain code quality across the SDLC.”
Reema Sharma, Northdoor developer
Where Northdoor sits
We have worked with IBM for three decades. That is part of why we were at the Bobathon. The other part is that we want our own developers to have hands on these tools before we recommend them to a client. As a result, Reema and Dennis came away convinced that Bob has a real role in legacy modernisation. It also has a role in API rewriting work, and in any environment where security cannot be an afterthought.
If you are weighing up where an AI coding assistant fits in your delivery pipeline, or if your modernisation backlog is bigger than your team, we would be happy to share what we have learned. In the meantime, you can read more on our IBM Bob page or speak to our team about a discovery session. Further context on the underlying capabilities is available from IBM’s documentation on watsonx Code Assistant.