The Future of UI: AI to Orchestrate Visualization, Not Just Data

Traditional user interfaces, designed for predictable data structures and static visualizations, are increasingly ill-equipped to handle the dynamic and often unstructured outputs of AI agents. While large language models (LLMs) can format responses using markdown, this simple format often results in unwieldy “walls of text” within terminals or IDEs, hindering effective digestion of complex information. The fundamental challenge lies in visualizing data whose structure is unknown until the moment it appears, moving beyond predefined dashboards like Grafana which assume prior knowledge of data layout and user intent. This necessitates a re-evaluation of UI design principles, exploring methods where AI plays a more central role in how information is presented.

Two primary architectural approaches are emerging to address this. The first, exemplified by MCP apps, involves the server shipping full rendering code (HTML/JS/CSS bundles) to the client, where it’s displayed within sandboxed iframes. While offering maximal rendering flexibility for GUI-based agents, this often results in a fragmented user experience, with inconsistent mini-apps and inherent security considerations. The more promising alternative, championed by the speaker and mirroring industry trends from Google’s A2 UI to OpenAI’s Open JSON UI, advocates for AI agents returning structured data with type hints, which a client-side component library then renders. In this model, the LLM acts as an “architect,” dynamically selecting and combining predefined visualization patterns (e.g., diagrams, tables, charts) based on the data’s inherent nature, ensuring UI consistency while delivering context-aware, interactive experiences. This shifts the paradigm from pre-designed static UIs to dynamic interfaces assembled by AI on the fly, tailoring presentation to the specific demands of each interaction. The speaker’s “DevOps To” project demonstrates this principle, showcasing custom, interactive visualizations from AI-generated structured data.