Home

System Overview

From questions to trusted answers - powered by autonomous semantics.

Connecty AI System Overview

Connecty AI is a semantic AI system that transforms natural language questions into trustworthy, explainable insights across your modern data stack. It bridges user questions to client data warehouses and delivers answers via Connecty's own user interface, or optionally via various data consumption tools.

1. User Questions

Users ask analytical questions in natural language, with access managed via role-based controls.

2. Client’s Data Warehouse

Connecty AI integrates with leading cloud data warehouses (e.g. Snowflake, Databricks, BigQuery, Redshift, Azure Synapse, PostgreSQL). It syncs metadata and results, executes queries, and materializes responses.

3. Agentic AI

A multi-agent framework handles:

  • Conversational interface (Chat)

  • Intent understanding (Intent graphs)

  • Multi-step reasoning

  • Coordination among agents

4. Context Engine

This component ensures context fidelity through:

  • Syncing and reconciling metadata

  • Schema evolution handling

  • CMDC integration

  • Versioning and mapping

5. Autonomous Semantic Graph (ASG)

The core intelligence layer, ASG includes:

  • The Metricverse: A graph of metrics, filters, dimensions, measures, relations, and business domains

  • Rich metadata: Catalogs, schemas, objects, columns, and data definitions

  • Semantic consistency

6. AI Use-Cases

The system supports workflows to:

  • Analyze data

  • Discover insights

  • Refine and validate metrics

  • Collaborate across teams

7. Trusted Answers with Expert-in-the-Loop

Every answer is enriched through:

  • Inference: Logical reasoning over the semantic graph and metadata

  • Actionable Explainability: Transparent insight into how answers were generated with expert actions e.g. controlled follow-up, verification, unverification

  • Expert-in-the-loop review: Human feedback and validation when needed

8. Data Consumption

Answers are delivered across tools like Power BI, Looker, Tableau, dbt, Google Docs, GitHub, TensorFlow, and more.

Last updated