Authors: Uzair Jawaid, Horo Zhang, Ting Cheng
Abstract: The new Policy Tab in the Decision Assistant introduces a major step toward transparency and user empowerment. This feature provides clear, structured access to policy information through cross-reference citations and markdown-formatted text, ensuring accuracy and readability. Users can now preview and edit AI-generated policies in markdown, making compliance management more intuitive and collaborative. By integrating citations, the Policy Tab builds trust and reinforces ethical AI principles. In this blog, we’ll explore its purpose, editing capabilities, and benefits.
Introducing the Policy Tab: A Step Toward Transparency
As the Decision Assistant becomes an integral part of the Decision Automation landscape, transparency is key in delivering robust solutions. Users need to be able to trust the output of the large language model (’s decision-making process. An AI assistant must provide clarity to the user on why it did what it did, essentially, “showing its work”.
The Decision Assistant now includes a Policy Tab in the Workspace Canvas in the December 2025 release. This provides users with the ability to see the policy document they provided, or the AI-generated policy, in a tab alongside the AI-generated decision model and data model. This tab provides users with insight on where it got its model components. In the previous version, users were unable to verify the output of the generated decision model and data model based off their policy document and were also unable to view the policy document the LLM would generate if they did not provide a policy document. This new feature creates transparency between the user and the AI assistant.
Policy Editing and Markdown Formatting
When the user doesn’t have a policy document to provide, the Decision Assistant creates one on its own based on the initial prompt. This generated policy is then shown to the user for verification (another example of transparency). This view is a simplified version of the new policy tab where only the generated policy is shown, along with editing options. The user is empowered to either make modifications themselves or ask the Decision Assistant to modify the policy for them through the chat panel. The editing panel allows the user to directly manipulate the AI generated content, as well as preview it in markdown (MD) format. Once the changes have been made, the user can confirm the changes, and it will go through the assistant to verify the changes are valid before saving and proceeding to the next step in the decision service generation.
Figure 1 - Edit Generated Policy View
This new feature gives the user autonomy to provide the right context to the don’t have a policy document. The generated policy document can be modified however the user likes, empowering the user to have more ownership over the AI output. In return, this helps build a stronger trust between users and the Decision Assistant.
How Cross-Reference Citations Build Trust
Citations are a foundational mechanism for transparency in any AI-driven system. In the Policy Tab, every generated statement, requirement, or definition that the assistant extracts from a policy document is tied to its source through cross-reference citations. These citations allow users to trace each policy element back to the original sentence or section it came from, giving them clear visibility into the assistant’s reasoning path.
Figure 2 - Policy Tab View
This level of attribution enables users to validate the AI’s interpretation rather than relying on it as a black box. When users can immediately see why an attribute was created in the data model or which policy clause justified a business
By reinforcing the link between inputs and outputs, the Policy Tab ensures that the Decision Assistant’s work is verifiable, auditable, and compliant with enterprise-grade governance expectations. This reliability is a key step toward building long-term trust between users and AI-generated decision artifacts.
Empowering Users Through AI Transparency
One of the core principles of ethical AI is enabling users to understand, influence, and refine the system’s behavior. The Policy Tab directly supports this principle by offering a clear window into the assistant’s reasoning process. Instead of hiding the logic that shapes the data model or decision model, the assistant openly displays its interpretations and allows users to edit or correct them.
This transparency empowers users at every stage of the decision service lifecycle. Business analysts can validate whether the generated policy aligns with organizational requirements. SMEs can provide clarifications that directly improve AI outputs. Developers can trace how specific rules were formed, making downstream debugging and implementation smoother. By allowing users to engage with the policy in markdown, edit content, and submit changes back through the assistant for validation, the system creates a collaborative loop between human and AI.
Ultimately, transparency transforms the Decision Assistant from an opaque generator into a trustworthy partner—one that aligns with IBM’s broader commitment to accountable and human-centered AI.
Real-World Use Cases
1. Rapid Policy-to-Model Development for New Regulations
Organizations that must frequently update their decision services to comply with changing regulatory frameworks can leverage the Policy Tab to accelerate adaptation. When teams upload a new compliance policy, the assistant generates a structured interpretation with citations, enabling compliance officers to immediately validate correctness. This shortens the cycle from regulation to deployment.
2. Onboarding Teams Who Lack Formal Policy Documents
Many clients begin decision automation projects without a fully finalized policy document. The assistant’s ability to generate a draft policy from a business prompt—and allow users to edit it in markdown—helps teams quickly converge on a clear, standardized reference document. This is especially valuable for SMEs who know the rules but have never written them formally.
3. Cross-Functional Collaboration Across Business and Technical Teams
Policy citations enable business users, developers, and QA teams to speak the same language. Business analysts can confirm the policy intent, while technical teams can trace each element in the generated decision model back to its source. This reduces misinterpretations during implementation and testing, improving overall solution quality.
4. Governance and Audit Readiness in Regulated Industries
For industries like finance, insurance, and public sector, explainability is not optional. The Policy Tab helps organizations demonstrate exactly how each rule and decision aligns with documented policy requirements. Auditors gain a verifiable chain of reasoning, eliminating ambiguity and lowering compliance risk.
Setting the Stage for What’s Next
The new Policy tab provides users with more insight into the Decision Assistant’s content generation, creating transparency and building trust. Users can edit the AI-generated policy document as they see fit, providing them more flexibility with the
The December 2025 release of the Decision Assistant not only includes the new Policy tab, but also compatibility with IBMs Operational Decision Manager. In the next blog in this series, we’ll cover what’s new with the Decision Assistant for task model generation.