Hunch

Hunch is a data layer that sits between a company's existing analytics stack and the AI agents built on top of it. The problem it targets is that raw data warehouses capture numbers but not the tribal knowledge — metric definitions, business logic, and context — that a human analyst would normally supply, which is why AI agents asked to analyze data often give plausible-sounding but wrong answers. Hunch addresses this with a semantic layer that maps data relationships, metrics, and business rules, plus an agent verification step that checks an AI agent's output against those definitions before it reaches a user. The platform supports several agent types out of the box, including conversational chat agents, root-cause analysis agents, proactive monitoring agents, and predictive forecasting agents, all configured declaratively through YAML rather than custom code. It connects to common data warehouses such as BigQuery, Snowflake, Databricks, Redshift, PostgreSQL, and MySQL, to BI tools including Looker, Tableau, and Power BI, to metadata platforms like dbt and DataHub, and to AI models such as Claude and OpenAI's, with a LangChain integration and a HunchClient SDK for developers. It is aimed at data and analytics teams that want to give business users self-serve access to data without every question requiring a human analyst. Pricing is not published on the site.