Omni raised $120M to make agents stop hallucinating your data
Ex-Looker co-founders built something enterprise AI has been begging for: a governed semantic layer that sits between agents and raw data warehouses. ICONIQ led a $120M Series C at a $1.5B valuation, more than 2x from their $650M mark in March 2025. Theory Ventures, First Round, Redpoint, and GV followed on. There's also a $30M employee tender.
The problem is universal. Every company wants to let an LLM agent query the warehouse. Every company finds out the agent hallucinates column names, joins tables incorrectly, and invents KPI definitions that nobody agreed to. The semantic layer solves this by exposing only vetted metrics, dimensions, and relationships. The agent has to speak through the layer, not around it.
Omni connects to Snowflake, BigQuery, Databricks, Redshift, Postgres, ClickHouse. Customers include BambooHR, Checkr, Cribl, dbt Labs, Mercury, Pendo, Synthesia. Growth is 4x year over year, which is why ICONIQ was willing to step up.
The broader read: the people who won the BI era (Looker was a Google acquisition) are now building the BI layer for AI. Tableau, PowerBI, and Metabase all have the same problem, none of them were built for an agent to query. Omni's moat is that the semantic layer is the actual governance surface. If you run it, you control what every agent sees and says. That's going to matter more every quarter.
Link: https://omni.co
← Back to all articles
The problem is universal. Every company wants to let an LLM agent query the warehouse. Every company finds out the agent hallucinates column names, joins tables incorrectly, and invents KPI definitions that nobody agreed to. The semantic layer solves this by exposing only vetted metrics, dimensions, and relationships. The agent has to speak through the layer, not around it.
Omni connects to Snowflake, BigQuery, Databricks, Redshift, Postgres, ClickHouse. Customers include BambooHR, Checkr, Cribl, dbt Labs, Mercury, Pendo, Synthesia. Growth is 4x year over year, which is why ICONIQ was willing to step up.
The broader read: the people who won the BI era (Looker was a Google acquisition) are now building the BI layer for AI. Tableau, PowerBI, and Metabase all have the same problem, none of them were built for an agent to query. Omni's moat is that the semantic layer is the actual governance surface. If you run it, you control what every agent sees and says. That's going to matter more every quarter.
Link: https://omni.co
Comments