Wisdom

Overview

The agentic analytics assistant — ask complex, multi-stage questions in plain English; Wisdom plans, executes, explains. Far more than text-to-SQL.

Wisdom is DataGenie's agentic analytics assistant — you ask complex, multi-stage questions in plain English, and one autonomous agent plans, executes, and explains. It understands your question, pulls and verifies the numbers through DataGenie's own analytical services, and returns the answer as a single, streamed document of charts, tables, and narrative — far more than text-to-SQL.

Every number Wisdom produces comes from a deterministic service — not from a language model. Your raw data never reaches an LLM. See Responsible AI.


How Wisdom works

Ask in plain language

Pick your dataset, then ask your question in the question box — no SQL, no exact column names. Type / to pull in a skill, a KPI, or a dimension and focus the answer exactly where you want it.

Watch the answer assemble

Wisdom plans the work, fetches and verifies the data, and streams the answer back section by section — a one-line takeaway first, then the KPI against its expected range, the trend, the drivers behind any change, and a plain-language explanation. One integrated document, not a wall of tabs.

Reshape and drill in

Change how a result is shown without re-asking — swap the chart type, combine sections, or revert. Click into any chart or table to break it down by dimension and trace the numbers back to their source.

Keep the thread

Ask follow-ups in the same notebook — Wisdom holds context and builds on earlier answers. Your work is saved as a notebook you can reopen, restored to its full state.

Wisdom remembers you

Confirm the durable preferences Wisdom picks up — the metrics you care about, how you like answers framed — and it applies them to every future answer, privately to you.

The more specific your question, the sharper the answer. "What caused the drop in Conversion Rate on Mobile in the US last week?" gives Wisdom a clear scope to work with — vague questions produce broader, less actionable results.


Multi-stage queries

Wisdom handles compound prompts that span multiple analytical tasks in a single question. The agent's execution plan branches dynamically — the result of one service call determines the next service to invoke.

Invalid superlore-canvas spec — expected JSON with nodes (and optional edges/groups).

When you ask the same multi-stage question repeatedly, package it as a Wisdom Skill — invocable with arguments, deterministic output every time.


How Wisdom stays accurate

Three configuration surfaces per dataset act as guardrails. Together they make Wisdom respond within your business rules instead of inventing them.


Example questions to try

Trend analysis

"Show me the trend of Revenue over the last 3 months."

Dimensional breakdown

"Break down Conversion Rate by marketing channel."

Period comparison

"Compare Revenue this period vs last period."

Explain a story

"What caused the drop in Conversion Rate yesterday?"

Forecast

"Forecast Revenue for the next 4 weeks."

Scenario

"If Conversion Rate rises 10%, what happens to Revenue next month?"


What's next

On this page

Built withsuperlore