DecisionBox is an autonomous AI platform that connects directly to data warehouses such as BigQuery, Redshift, Snowflake, PostgreSQL, and Databricks, where an AI agent generates, runs, and validates 50–100+ SQL queries per execution to surface ranked, severity‑graded insights and
Funding
Funding not disclosed
Founders
Product
Problem
Organizations with large data warehouses often struggle to extract actionable insights because analysts must manually write and run many SQL queries, validate results, and synthesize recommendations, which is time‑consuming and prone to error.
Solution
DecisionBox provides an autonomous AI discovery platform that connects directly to data warehouses such as BigQuery, Redshift, Snowflake, PostgreSQL, Databricks, and others. An AI agent iteratively forms hypotheses, generates and executes 50–100+ SQL queries per run, and validates each finding with independent verification queries to eliminate hallucinations. The system surfaces ranked, severity‑graded insights and concrete, numbered recommendations, all fully transparent with logs of every query and reasoning step. Domain packs tailor the analysis to specific industries (e.g., gaming, social, e‑commerce) and users can author custom packs without altering the core engine. The platform supports multiple LLM providers, self‑heals failed queries, and continuously improves through cumulative learning from prior runs and user feedback.
Target Audience
Primary users are product managers, data analysts, and developers who need rapid, validated insights from their data warehouses, as well as enterprises in gaming, social media, and e‑commerce seeking industry‑focused analysis.
Features
- Autonomous discovery loop that writes, runs, and iterates over SQL queries without user prompting
- Multi‑warehouse support with automatic schema discovery for BigQuery, Redshift, Snowflake, PostgreSQL, Databricks, etc.
- Insight validation via independent verification queries to ensure all numbers are accurate
- Industry‑specific domain packs (gaming, social, e‑commerce) and a framework for creating custom packs
- Full transparency: every SQL statement, reasoning step, and decision is logged and viewable
- Self‑healing SQL that diagnoses and rewrites failed queries using warehouse‑specific fix strategies
- Cumulative learning that refines hypotheses based on previous discoveries and user feedback
- Multi‑LLM integration (Claude, OpenAI, Gemini, Vertex AI, Bedrock, Ollama) with no vendor lock‑in