With IBM DevOps Solution Workbench 5.2, AI-assisted spec-driven development is brought to the next level in an enterprise context. This release standardizes the AI Design Assistant through versioned Design Recipes, lets you scale customized modeling and workflows company-wide with Project Baselines, gives you full control over how your model is visualized on the canvas, and opens the Workbench to agentic AI tools through a new MCP Server. This empowers organizations in a way that a structured Design Intent can be created, applied, and reused with consistency and control across every team.
AI Design Assistant with Design Recipes
The AI Design Assistant is your companion across every design workflow in IBM DevOps Solution Workbench. In 5.2 it works with Design Recipes – versioned packages that bundle two kinds of reusable expertise:
• Skills: design know-how that the AI Design Assistant applies automatically whenever a request matches its scope. For example, how to derive out of a set of user stories the required services and components or drafting a domain model in line with your conventions.
• Commands: chat shortcuts and ready-made actions that operate on the model elements you select.
Design Recipe packages are managed and published by administrators and attached to project baselines, so the same governed expertise is instantly available to every project and every team – instead of living in individual prompts.
Key benefits
• Faster, context-aware design proposals: the assistant analyzes the full project content and proposes or applies model changes directly, instead of manual box-drawing.
• Assisted breakdown across abstraction levels: guide high-level objectives such as user stories consistently with your workflow down from idea to specification.
• Governed, reproducible AI behavior: because Skills and Commands are versioned and baseline-bound, AI assistance stays consistent and auditable across experts and teams – it no longer depends on individual prompting skill.
Ideal for architects, domain experts, and product owners who want AI support that follows their organization’s design standards rather than improvising.
Project Baselines
Project Baselines provide central management of all your customizations, turning them into company standards that every team can adopt. Define your own model element types, relationships, and rules in OML Profiles (Open Modeling Language) for your specific use cases, then bundle the profiles you need – together with recipes for Coding Assistants and project configuration – into Project Baselines.
Key benefits
• Fast scaling from pilot to production: define standards once, centrally, and roll them out to every project rather than re-creating them per project.
• Harmonized quality and consistent workflows: the same elements, relationships, and rules apply across experts and teams, preventing the divergence that results from individual prompting and ad-hoc skills.
• Controlled lifecycle: baselines follow a clear draft → published → deprecated lifecycle, so standards evolve in versioned, auditable steps.
• Best practices that scale: capture what proves valuable in a pilot project as reusable Skills and Commands, then roll them out for enterprise-wide, cross-project use.
Perfect for platform teams and centers of excellence that need to govern how dozens of teams design and build.
Customizable Visualization on Canvas
You can now tailor how each model element is rendered on the canvas. Depending on what the profile defines, an element can appear as a card, an expanded card, a customized shape, or as text (inline rich text). Switch representation at any time from the model element details or through Change visualization in the canvas context menu.
Key benefits
• A visual language that fits your organization: align element appearance with your own notation and standards, making models easier to read and to share with stakeholders.
• Mixed textual and graphical working: render elements as inline text to edit content quickly and combine prose with diagrams on a single canvas.
• The right level of detail for every audience: collapse elements to compact cards for an overview, expand them for detail, or switch to shapes for stakeholder-facing diagrams – all from the same underlying model.
• Profile-driven and consistent: because visualization is defined in the OML profile, your notation is configured once and applied uniformly everywhere it is used.
Ideal for teams that present the same model to mixed audiences – from engineers to business stakeholders.
Modeling Support for Agentic AI Tools
Agentic coding assistants can now interface with the Workbench through a new MCP Server. Agentic AI tools can read existing model data and OML profiles, create new model element content, and apply changes to the model directly inside IBM DevOps Solution Workbench. Conversely, existing model content and profiles can be extracted and handed to downstream development steps.
Because agentic AI works against your existing or customized OML profiles, the model content it generates conforms to the schema definition and aligns with your customized design process – conformance is preserved even when the authoring is done by AI.
Example use cases
• Reverse engineering: discovering software design from an existing code base (see the demo video).
• Importing user stories and business goals: Creation of document artifacts in textual form, used as the basis for further architecture breakdown inside the Workbench.
Key benefits
• Heavily reduced manual effort: let agentic AI import existing information into the Workbench instead of modeling it by hand.
• Faster discovery: analyze an existing architecture and implementation base quickly and establish the status quo in a fraction of the time.
• Schema-conformant AI output: the MCP Server constrains the AI to your OML profiles, so AI-authored content stays valid, consistent, and governable – no free-form drift.
• A connected design-to-code loop: keep Design Intent and implementation in sync as both evolve.
Perfect for application modernization, where agentic AI lifts an existing codebase into a governed, structured design.
If you want to know more about IBM DevOps Solution Workbench 5.2, book a meeting now:
https://www.ibm.com/products/devops-solution-workbench