InVocIQ by humaineeti -- Voice-Driven AI BI on Enterprise Data

InVocIQ by humaineeti is a voice-driven AI BI solution that transforms spoken questions into cited, visualized answers from live enterprise data -- in seconds.

The Problem: Natural-language BI fails on enterprise data without a semantic layer. "Show me revenue" can mean five different things depending on the department, grain, or time window. Naive text-to-SQL generates incorrect joins, inconsistent metric definitions, and ungoverned queries.

The Solution: InVocIQ pairs voice input and LLM-powered intent extraction with a semantic layer so the right metric, the right grain, and the right filters are unambiguous before any SQL is generated.

How It Works: InVocIQ operates as a pluggable pipeline with swappable components that fit your existing stack. Voice capture uses browser Web Audio or mobile native with on-device voice activity detection. Speech-to-text is handled by OpenAI Whisper, Deepgram, Amazon Transcribe, or Google Speech-to-Text with Indian-English and multi-Indic support. An LLM (Claude or GPT) extracts intent, metrics, dimensions, time windows, and filters -- constrained to what the semantic layer exposes. The structured intent compiles to a query against dbt MetricFlow, Cube, AtScale, or Looker LookML. SQL executes against your warehouse (Snowflake, BigQuery, Databricks, Redshift, or Postgres) with RBAC enforced at the warehouse level. Visualization is auto-selected by an LLM based on result shape and rendered via Plotly, Apache Superset embedding, or your existing BI surface. Optional voice response is available via Amazon Polly, ElevenLabs.

Why the Semantic Layer Matters: Metric consistency ensures "Revenue" is defined once in code with explicit grain. Join correctness means the semantic layer knows which tables join on which keys -- the LLM cannot invent a join. Authorization applies row-level and column-level security at the warehouse governed by user identity. Citations show the underlying tables, columns, and SQL for auditability by default.

What InVocIQ Does Well: Defined metrics over defined dimensions, common time-window comparisons (MoM, YoY), top-N and bottom-N queries, and drill-downs within modelled metrics.

What It Does Not Do: Free-form analysis on un-modelled data, statistical inference or causal claims, forecasts without existing models, or PII access without proper authorization.

Key Differentiators vs. Native BI Tools: BYOM (Bring Your Own Model) -- choose Claude, GPT, or open-weight models per use case including on-prem deployment for regulated workloads. Vendor-agnostic -- works across multiple BI surfaces and warehouses without lock-in. Data-residency control -- ASR, embeddings, and inference can run in customer-controlled regions. Custom governance -- integrates with your identity, audit, and approval flows.

AWS Services: Bedrock, Amazon Transcribe, Amazon Polly, ECS, ALB, RDS, SQS, SES, S3