Skip to content
Anton Wolter
Research

How should work be split between people and agents?

Agentic Visualization

Design patterns for visualization systems in which AI agents take on roles while people stay in control of the analysis.

Team
Vaishali Dhanoa, Anton Wolter, Gabriela Molina León, Hans-Jörg Schulz, Niklas Elmqvist

Problem

Agents built on large language models can now act with growing autonomy. Visualization research has long kept a person in the sensemaking loop, so it has to answer how much to delegate to agents, and how people and agents coordinate, without giving up human agency.

Approach

We revisit existing visualization systems that already include semi-automated or fully automatic AI components and reinterpret them through an agentic lens.

System

This is an analysis of existing systems rather than a single new tool.

Findings

The result is a collection of design patterns for agentic visualization, covering agent roles, communication and coordination. They are meant as a foundation for systems that use agents while keeping human insight and control.

Publications