A Composable Agentic System for Automated Visual Data Reporting

TL;DR

Addresses brittleness in monolithic AI agents by proposing a human-in-the-loop partnership for automated visual data reporting. The paper presents a DSPy-based multi-agent system that externalizes core reasoning to deterministic modules (e.g., Dataset Profiler, Dataset Visualizer powered by Draco) and produces dual outputs: an interactive Mosaic-powered Observable 2.0 report and executable Marimo notebooks, with full execution traces recorded in Langfuse. The approach demonstrates generalization across eight diverse datasets and enables durable analyst steerability through traceable tool logic, as well as reader Explorability via linked data in Mosaic. The work highlights a path toward auditable, modular, open-standards-based AI-assisted data storytelling that preserves human agency and reproducibility.

Abstract

To address the brittleness of monolithic AI agents, our prototype for automated visual data reporting explores a Human-AI Partnership model. Its hybrid, multi-agent architecture strategically externalizes logic from LLMs to deterministic modules, leveraging the rule-based system Draco for principled visualization design. The system delivers a dual-output: an interactive Observable report with Mosaic for reader exploration, and executable Marimo notebooks for deep, analyst-facing traceability. This granular architecture yields a fully automatic yet auditable and steerable system, charting a path toward a more synergistic partnership between human experts and AI. For reproducibility, our implementation and examples are available at https://peter-gy.github.io/VISxGenAI-2025/.