Agentic Dashboards reframes visual analytics as a persistent runtime workspace bounded by a designer-defined decision space
A paper submitted on November 12 proposes the agentic dashboard: an agent selects analytical methods, views, and their coordination from natural-language goals while remaining inside a decision space defined by the designer, with results preserved in an evolving visual-analytics workspace.
Why it matters
This moves generative UI from one-off chart creation toward a continuously revised analytical environment. Designers specify what an agent may decide at runtime rather than fixing every screen, while experts retain awareness and judgment through visible state, traces, and constraints.
A closer look
The paper separates algorithm, control, and view layers. Agents can first compose API-accessible data and tools in the control layer, while coding agents may extend the algorithm layer. Proposed evaluation covers clear, ambiguous, and prohibited requests, preservation of unaffected state, and traces of specifications, tools, and rules. It should also test whether experts detect errors, understand downstream effects, and recover without rebuilding unaffected work.
Who this is for
Research, product, and engineering teams building generative analytics, BI, data workbenches, visualization agents, or human-agent decision systems.
Original sources
Revisions & corrections (1)
— New primary research found in the current arXiv HCI batch; claims and limitations verified from the paper.