Most natural-language analytics tools fail in the same place: they guess. Ask “how did we do last quarter” and a black-box assistant invents a join, picks a metric definition nobody agreed on, and returns a confident number that quietly contradicts the board deck. InVocIQ is built to remove that ambiguity rather than hide it.
InVocIQ is a voice-first business intelligence layer that sits in front of your existing warehouse. A spoken question is transcribed, mapped to intent by a large language model, and then constrained to the metrics, dimensions, and filters your team has already modelled in a semantic layer. Only governed, validated SQL reaches the warehouse — and every answer ships with the sources behind it.
What makes InVocIQ different
Semantic-layer grounding, not guesswork
InVocIQ resolves every question against your dbt MetricFlow, Cube, AtScale, or LookML definitions. Metric meaning, grain, and join logic are decided by your model — never improvised by the LLM — so departments stop arguing about whose number is right.
Bring-your-own speech and model
Plug in Whisper, Deepgram, Amazon Transcribe, or Google Speech-to-Text for the voice pipeline, and Claude or GPT for intent. Nothing is locked to a single vendor, so you keep procurement leverage and avoid platform lock-in.
Runs on the warehouse you already have
Snowflake, BigQuery, Databricks, Redshift, and Postgres are first-class targets. InVocIQ queries in place — no copy of your data into a proprietary engine, no shadow warehouse to govern.
Citations and auditability by default
Every answer carries the metric, filters, and time window that produced it. Analysts can trace a spoken question all the way to the SQL, which makes InVocIQ safe for regulated and board-level reporting.
Automatic, appropriate visuals
Results render as the right chart for the question — trends, comparisons, and rankings via Plotly or Apache Superset — with an optional spoken summary through your text-to-speech provider of choice.
Security enforced at the source
Row-level and column-level security live in the warehouse, so InVocIQ honours the access controls you already trust. Data residency stays under your control for regulated workloads.
Who InVocIQ is for
- Data and analytics leaders who want self-service that does not sacrifice metric consistency.
- Regulated organisations that need citations, auditability, and warehouse-level access control.
- Teams that want bring-your-own-model independence instead of being locked into Tableau Pulse or Power BI Copilot.
- Executives and operators who would rather ask a question out loud than build a dashboard.
Built for governed metrics — honest about its limits
Designed for
- Metric queries over modelled dimensions
- Time-window and period-over-period comparisons
- Top-N and bottom-N rankings
- Drill-downs within already-modelled metrics
Out of scope
- Free-form analysis on un-modelled, ungoverned data
- Statistical inference or causal claims
- Forecasting without an existing model
- PII access without an authorisation context
InVocIQ FAQ
How is InVocIQ different from Power BI Copilot or Tableau Pulse?
Those tools are tied to one platform and one vendor stack. InVocIQ is vendor-agnostic across speech, model, and warehouse, and it forces every question through your semantic layer so answers stay consistent and citable.
Does InVocIQ move my data?
No. It generates governed SQL that runs against your existing warehouse in place. Your data, access controls, and residency requirements stay where they are.
Which large language models does it support?
Claude and GPT for intent extraction today, constrained to your semantic-layer definitions. Because intent and execution are decoupled, you can switch models without rewriting your governance.