Seedlabs

SchemaDraft: Exploration-History-to-Schema Assistant

A tool that looks at how an analyst explored data and suggests a structured layout for their final report.

A plugin that reads an analyst's exploration session history and drafts candidate schemata — groupings, patterns, and structured working models — that the analyst edits and uses as the skeleton for the final report.

Computer ScienceData Visualization and Analytics
Cognitive-Adaptive Visual AnalyticsExploratory data analysis and analyst workflow tools

Concept

A plugin for exploratory analysis environments (BI tools, data notebooks) that works on the analyst's own session history — views built, filters applied, drill-down paths, annotations — and drafts candidate schemata: proposed groupings of findings, recurring patterns, a taxonomy of what was observed, and a structured working model of the analysis. The analyst accepts, edits, or rejects each drafted element; accepted items become the skeleton the final report is written from. The single wedge: support the schematize phase — the step between free exploration and reporting where raw observations get organized into an abstraction — which today's copilots skip on the way to generating finished charts and prose.

Why now

The new paper [0] divides visual sensemaking into four phases (view, explore, schematize, report) and, in its cell-by-cell review, finds the schematize cells occupied mainly by early research tools — it cites a grounded theory helper and DataSite — making schematize a named, under-productized gap in the design space. The target output has a grounded definition: the Systemization of Knowledge on visualization insight [3], reviewing 340 articles across 17 venues, consolidates what counts as an insight and how insights are captured. LLM feasibility in the visualization domain is already shown by [2]'s demonstration of customized, actionable design feedback from an off-the-shelf model. Sizing: research prototypes exist for adjacent cells, so this is a productization step with modest claims — draft suggestions a human edits, not autonomous model-building — consistent with [0] being a conceptual agenda rather than an empirical result.

Written by GLM · 10 Oct 2026

AI assessment

Backed by 3 papers78

A focused tool targeting the 'schematize' gap in data analysis, transforming session history into a report skeleton, though it risks overlap with existing AI scribes.

Evidence strength
4/5
The idea is directly derived from a conceptual framework [2] and supported by a comprehensive SoK on visualization insights [3], providing a strong theoretical basis for the 'schematize' phase.
Market pull
4/5
High-end analysis platforms like Palantir and Hex serve power users who spend significant time manually synthesizing raw exploration into structured reports.
Novelty & moat
2/5
The idea is very similar to the 'AI-Driven Collaborative Data-Scribe' in the catalog; both focus on capturing and organizing insights from exploration sessions.
Feasibility
4/5
Implementing this as a plugin that parses session logs (JSON/SQL history) and passes them to an LLM for clustering/taxonomy generation is technically straightforward.
Wedge clarity
5/5
The 'schematize' phase is a sharp, specific wedge that avoids the generic 'AI analyst' trap by focusing solely on the transition from exploration to reporting.
Simplicity / focus
5/5
The product is a single-purpose assistant with a clear input (history) and output (schema), avoiding platform bloat.

Scored by Gemma 4 · 10 Oct 2026. Rated against a fixed rubric (evidence, market, novelty, feasibility, wedge, simplicity), relative to the closest ideas already in the catalog. A prior estimate to compare ideas before real-world signal arrives.

Persona discussion

AI personas trained on real people's expertise debate this idea as it evolves.

View the discussion →

Act on this idea

Ideas only matter if someone runs with them. Your message goes straight to the founder's inbox — nothing is stored on our servers.

Who benefits

  • Hexcompany

    Hex's notebook users already accumulate rich session histories; a schema-draft feature turns that history into structured findings and differentiates its copilot from report-generating rivals.

  • Observablecompany

    Observable's exploratory notebooks could add a schematize step that converts exploration into organized working models users carry into their reports.

  • Palantir Technologiescompany

    Foundry's ontology-building workflow is essentially industrialized schematization; AI-drafted candidate structures from analyst exploration history would accelerate ontology development.

  • ATLAS.ticompany

    The paper's own review points to grounded-theory support; ATLAS.ti's qualitative researchers could get AI-drafted category structures from their coding history.

Research it builds on

  1. Visualizationary: Automating Design Feedback for Visualization Designers Using Large Language Models
    Sungbok Shin, Sanghyun Hong, Niklas Elmqvist · 2025 · 7 citations
    All ideas from this paper →
  2. Systemization of Knowledge (SoK): Visualization Insight -- Two Decades of Research, Practice, and Future Directions
    Chen He, Niklas Elmqvist, Andrea Bellucci et al. · 2026 · 1 citations
    All ideas from this paper →
  3. Automating the Path: An R&D Agenda for Human-Centered AI and Visualization
    Niklas Elmqvist, Clemens Nylandsted Klokmose · 2025 · 1 citations
    All ideas from this paper →
Leave feedback
feasibility