Abstract

Techniques are proposed herein for using visualization in multi-agent enterprise artificial intelligence (AI) systems. The core approach treats visualization as a specialized agent that a supervisor can invoke at multiple points during a live execution trajectory. Rather than generating only a final static widget, the system can insert intermediate visual checkpoints that explain what an agent team has discovered, which hypotheses remain, which sub-agents contributed, and why a recommendation should be trusted. The techniques can support complex, cross-domain troubleshooting and remediation workflows in which a user supervises, understands, and approves a next step. By stitching prompt-and-response pairs, evidence widgets, executor-agent lineage, and provenance into a story-like canvas, the system turns opaque multi-agent execution into an explainable operational narrative. The supervisor can dynamically decide when visualization should appear, send semantic context and why-now reasoning to a visualization agent, integrate multiple intermediate and final visualizations into the same conversation, provide provenance-rich cards showing agent lineage and data origin, and learn successful visual-storytelling patterns as reusable skills.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS