← Back to work

Product Design · AI Product · Data Visualisation

Prompt BI — Having a conversation with your data

From raw data to shareable insights, without needing a data analyst. A two-part system: an AI query interface and a live analytics dashboard.

Role

Design Lead

Scope

End-to-end design, leading design direction with PMs and engineers

Type

AI Analytics Platform

Prompt BI hero
"I can't make sense of my data without a data analyst."

0

Technical setup required

Plain

Language queries — no SQL needed

Live

Dashboard with auto-updating visuals

The problem

Non-technical users were locked out of their own data. Getting answers required a data analyst, writing SQL queries, or waiting for someone else to build a report. And even when insights were produced, they lived in spreadsheets — impossible to track, share, or act on over time.

The challenge: design a system that handles file uploads, enterprise database connections, natural language queries, and interactive dashboards — but stays simple when a non-technical user needs an answer fast.

Part 1 — Prompt BI: The Query Interface

Designing a flow that just makes sense

The experience starts with connecting your data. Simple as that. Once a file is added or a database is connected, Prompt BI immediately gets to work — no manual configuration needed.

Auto-generates an editable overview — the moment data is connected, users see a summary of what they're working with

🔍

Flags missing values and data issues — before the user even asks a question

🔗

Detects table relationships — automatically, so users don't have to think about joins

👁

Previews data with column names, types, and sample rows — no guesswork, no technical setup

Data connection flow

Connect your data — Prompt BI takes it from there

Insights without the intimidation

After setup, users ask questions about their data in plain language. They can mention table names for context or use smart suggestions. It's not just about data analysis — it's about having a conversation with your numbers.

Key Design Decision

Smart suggestions surface automatically as users type — reducing the intimidation of an empty query field. For users who don't know what to ask, we show them what's possible. The entry point is curiosity, not expertise.

Query interface

Ask questions in plain language — smart suggestions guide the way

Making insights actionable

Once Prompt BI generates an insight, it doesn't just display results — it makes them useful. Every response comes in both a visual chart and a summary format, giving users a clear understanding of their query.

📊

Switch chart types — bar, line, pie, table, whatever fits the data best

📌

Pin to dashboard — save important insights to review and share later

🔗

Share with teammates — directly from the result

✏️

Edit the query — refine and re-run without starting over

Insight result

Every insight is actionable — pin, share, download, or refine

Thoughtful errors, friendly fixes

Sometimes things go wrong — a file has no headers, or two tables don't match. Prompt BI guides users with clear, friendly messages and smart suggestions to fix the issue. No dead ends, no technical jargon.

Error state

Clear errors, not dead ends

Part 2 — The Dashboard: Where Insights Become a Story

From scattered insights to a cohesive picture

Pinning an insight is just the beginning. The dashboard is where everything users saved starts working together — not just a grid of charts, but a living view of what matters.

Dashboard overview

The dashboard — pinned insights working together

Customisation that scales

Saved insights unlock full customisation: edit chart types, adjust colour schemes, change labels, match branding. The dashboard grows with the user — from a quick personal view to a polished team report.

Customisation panel

Full control over how insights look and feel

Drill-downs and collaboration

Charts aren't just for viewing — they're made for exploring. Click into any visualisation to reveal the underlying data. Share dashboards with teams, download in multiple formats, or copy insights for external use. Auto-update settings keep all visuals fresh with the latest data — no manual refresh needed.

Key Design Decision

Auto-update settings were designed to be set-and-forget. Users shouldn't have to remember to refresh a dashboard. The data should always reflect reality — removing one more reason to distrust the numbers.

Drill-down view

Drill into any chart to see what's behind the numbers

Reflection

Prompt BI taught me that the hardest design problem wasn't the query interface — it was the transition between the two parts. Users needed to feel like pinning an insight was worthwhile before they ever saw the dashboard. So the dashboard had to be previewed earlier in the flow, as a destination worth building toward.

Looking ahead: more customisations for data previews, support for JSON and XML file formats, and integrations with platforms like Shopify and WooCommerce — to bring more users to their data, without needing a data team.

Next project

Dataroom Redesign →