Explore chart options with AI and Flourish

Tell the LLM what you’re trying to communicate and who it’s for. The more specific you are about the story: a trend, a comparison, a ranking – the better the chart suggestions will fit.

You don’t need to know which chart type you want yet; that’s what comes next.

Goal: [what you’re trying to show].

Audience: [who will see it].

Key story: [the insight or comparison].

Constraints: [any format, length, or brand requirements].

* Summarize the brief back and prompt for data upload.

Goal: show how customer churn has shifted across three product tiers over 18 months.

Audience: exec team. 

Key story: pitch compelling stories based on the data.

* Summarize the brief back and prompt for data upload.

Drop in your spreadsheet, CSV, or any supporting context like a brief or prior report. The LLM reads the structure of your data, relates it to your message, and uses it to ground every decision it makes.

If your data needs cleaning or reshaping, flag it here.

Dataset attached. 

Active sheet: [sheet name].

Key columns: [list]. 

Ignore: [columns or sheets to exclude]. 

Cleaning needed: [yes / no / specify]. 

* Confirm data structure is understood and ask to proceed with chart pitches.

Dataset attached. 

Active sheet: Sheet 1. 

Key columns: month, tier, churn_rate, cohort_size. 

Cleaning needed: no.

* Confirm data structure is understood before proceeding with chart pitches.

Download demo data: Customer churn rate

The LLM analyzes your data and narrative, then proposes a shortlist of chart types with a rationale for each. You’ll see why a slope chart might work better than a bar, or why a scatter plot surfaces a pattern a table would bury. Think of it as a senior data visualization editor giving you options.

Identify chart types that best surface each story. For each: trade-offs, suitability for identified audience, provide confidence level (1-10) based on data fit. 

* Present shortlist and ask for selection before proceeding.

Review the pitches and select the ones worth building — pick one, or pick all. You stay in control of what gets made without having to make any of it yourself.

This is the only decision point before Flourish Connector starts with the build.

Build options [list].

* Confirm selection and begin building in order.

Using Flourish, build options 1, 2, 3. 

* Confirm selection and begin building in order.

The Flourish Connector creates each selected visualization, mapping your data to the right columns and setting up the key chart details. Each result opens as a live, interactive chart, ready to review and refine.

If you want to update something across all the charts, you can instruct the LLM to apply it to every visualization at once. This is where working through the Connector pays off: batch changes that would take minutes per chart happen in one step.

Apply to all 3 charts: 

-Rewrite the title as a 6–12 word key takeaway, not a data description. Add a subtitle that describes what the data shows. 

-Highlight the story's key data point.

-Sort the data to make the insight land immediately

-Improve this chart across all five dimensions: title, chart type, metric, color, and labels — in one pass. Make every change directly and update the chart.

* Confirm all charts updated and ask if any further changes are needed before handoff.

Open any chart directly in Flourish to refine details like annotation placement, axis formatting and legend positioning. If you’d rather not adjust settings manually, use the Flourish Assistant to describe the change you want in plain language and apply it for you. Make one change at a time, or combine several refinements in a single prompt.

Connect the Flourish Connector through Claude or ChatGPT, or follow our documentation to connect it manually to another MCP-compatible tool. Then run the workflow with your own data – from dataset to a set of published, comparable charts in minutes.