Product · How it works

From business question to answer. With the interpretation visible in between.

Most analytical interfaces show you a query or chart after the meaning has already been chosen. Spotonix is the AI analyst that interprets before it queries: it makes that choice inspectable as an analysis plan before SQL executes.

See it on your data Get demo access

01 · Ground 02 · Compose 03 · Reuse

Ground / 01 · Builds the Context Graph

Grounds the question in your Context Graph.

Spotonix reads your warehouse schema, your modeling layer, BI assets, and approved query history, and builds a Context Graph of governed concepts — Segments, Calculations, dimensions, and Analysis Patterns — scoped to your deployment.

Your stack:

Warehouse Snowflake schema tables · columns · types

Transformations dbt models refs · sources · materialized

BI assets Power BI · DAX models measures · hierarchies · relationships

Activity Approved SQL history reviewed patterns · your team

Your Context Graph · retail example

Segments Calculations Answers Dimensions

18,432SQL queries analyzed

142dashboards parsed

89dbt models read

47building blocks composed

The Context Graph is built from your corpus and scoped to your deployment. It is not portable to another customer — the structure is yours. These figures are illustrative of a typical onboarding, not a benchmark.

Compose / 02 · Resolves material ambiguity

When a term is ambiguous, it asks — then can reuse your answer.

The first time Spotonix meets an ambiguous term, it surfaces the choice instead of guessing. Once you approve a definition, it can be persisted to your Context Graph. Where an approved interpretation is persisted and retrieved, a later question can reuse it instead of re-asking.

Turn 1 First time it sees "high spenders"

> What product categories do high spenders like to shop?

"High spenders" could resolve to two Segments. Pick one — or define a new threshold.

Spend > $5,000 / year Segment · approved · used in 4 prior answers Selected

Top 5% by annual spend Segment · approved · used in 1 prior answer Pick

Define a new threshold Composes a new Segment in your Context Graph Compose

✓ "Spend > $5,000 / year" accepted, and persisted to your Context Graph

Turn 2 A later question

> List the top 5 categories of products high spenders are buying this year.

Plan · retrieved from your Context Graph

"high spenders" → Spend > $5,000 / year Segment · approved earlier · retrieved from your Context Graph

"top 5 categories" → rank(sum(sales)) desc, limit 5 Calculation · Intent Algebra: ranking

"this year" → year = current_year Dimension · time filter

✓ No new clarification needed here. The plan is shown for review before SQL runs.

Approve an interpretation once. Where it's persisted and retrieved, a later question can reuse it instead of re-asking.

A prompt-only tool re-derives the term from scratch each turn, or picks one silently and never tells you. Spotonix surfaces the choice, and where the approved interpretation is persisted and retrieved it can be reused — that is how shared context accrues.See it on your data →

Compose / 03 · Shows the plan

Read the plan before any SQL runs.

A real, multi-concept question. The interpretation is a graph — named nodes, named edges — you can read in seconds. The plan is the review surface: the business terms, calculations, filters, grain, and scope the generated query is expected to honor.

> Which stores are losing habitual buying customers over the last 4 quarters?

Answer Stores with Declining Habitual Buyers ranked by QoQ drop

resolves to

Segment Habitual Buyers the risky word: one team means ≥ 4 purchases per quarter, another means a spend threshold — the plan exposes the choice

grounded on

Calculation · Intent Algebra: growth QoQ Customer Count Change (current − prior) / prior · current vs. prior quarter

scoped by

Dimensions Store × Quarter over the last four completed quarters

✓ Plan accepted. The generated SQL is validated against the accepted plan's bindings before it runs. you approve the interpretation before SQL executes

Every node points to a definition you can read. Every edge is a relationship your team can review. The plan is editable — and when the data shape qualifies the result, Spotonix can say so.

Illustrative example, not a performance benchmark. Spotonix does not claim a deterministic plan-to-query compiler or identical SQL across repeated runs; the claim is plan-governed generation with material bindings validated before execution.

Reuse / 04 · Approved work becomes shared context

Approved work can become shared context.

Where approved interpretations are persisted and retrieved, they can become shared context for later analysis — so a later question can start from Segments and Calculations your team already accepted, rather than from a blank prompt.

Then a discovery question — answered from what's already there

“Show me how the customer dimension is segmented with respect to sales.”

Nothing new to define. Spotonix answers from the Segments and Calculations your team already accepted — Premium Customers, High Spenders, and the High-Sales cohort, each already related to Sales in the Context Graph — and shows how the customer dimension breaks down before any SQL runs.

Approve an interpretation once. Where it is persisted and retrieved, a later question can reuse it instead of re-asking.See it on your data →

Current capability boundary

What you can evaluate today.

Visible analysis plan

Live

Inspect the business concepts and analytical structure selected for the question.

Copilot confirmation gate

Live

Pause after interpretation and resume only after plan confirmation.

Binding validation

Live

Check material query bindings against the accepted plan before execution.

Watched signals & briefs

Live

Part of the Insights capability, available today.

Recurring Reviews

Live

Decision and outcome memory for recurring meetings, so each review starts from what the last one settled.

Native DAX / PBIX ingestion

Private beta

Interpret your own Power BI models. Available in private beta today.