IBM is adding new capabilities to its AI platform Bob, enabling organizations not only to generate code but also to support the entire software development cycle. The expansion focuses on collaboration between multiple AI agents, parallel task processing, and better insight into usage, costs, and governance.
IBM introduced Bob earlier this year as a platform that goes beyond the now-common AI code assistants. The latest version emphasizes automating and coordinating the entire software development process, from initial analysis and planning through testing, deployment, and management.
The key additions include support for multiple collaborating AI agents, the ability to control different tools simultaneously, and built-in reporting on costs and usage. In addition, IBM is introducing three specialized workflows for Java modernization, the IBM i platform, and IBM Z mainframes.
According to IBM, the role of developers is shifting as AI takes over a larger share of the programming work. Writing code is increasingly less of a bottleneck; challenges now lie in analyzing existing software, planning changes, conducting tests, and monitoring quality and security.
That is why, according to the company, Bob was not developed as a traditional AI code assistant that merely makes suggestions during programming. The platform is designed to help organizations make software development as a whole faster, more consistent, and more manageable. Aspects such as governance, cost control, and insight into AI usage also play a prominent role.
AI agents divide up the work
Bob supports various operating modes for a range of tasks, such as planning, answering questions, and performing tasks independently. The platform can also select the AI model best suited for a specific task. For example, it can switch between IBM Granite and Anthropic’s Claude, depending on factors such as performance, accuracy, and cost.
A new feature is the ability to deploy multiple model-specific tools in parallel. Tasks that were previously performed sequentially now run simultaneously. According to IBM, a process that previously took about thirty seconds can now be completed in less than ten seconds. At the same time, the number of tokens required is also reduced.
Another enhancement is the use of so-called subagents. For well-defined research tasks, such as analyzing authentication in an existing codebase, Bob can temporarily deploy a separate AI agent. This agent searches through files, maps out dependencies, and returns a summary to the main agent.
Because only the summary is saved and the intermediate analysis is discarded, the context window remains clear and organized. IBM has also expanded that context window from 200,000 to 270,000 tokens, allowing larger projects to be processed in a single session.
Built-in security
In addition to the new AI functionality, IBM has also expanded its security capabilities. Bob monitors sensitive data, enforces policy rules during the development process, and integrates security tests directly into the workflow. With this, IBM aims to offer organizations a platform that not only supports developers but also meets the requirements of larger enterprise environments where compliance and control play a key role.