I think we could help, if you're open to another alternative ;-)

Unpopular opinion: your data warehouse is fine. Agents don't need a new one. They need context: definitions, permissions, what changed since yesterday. Keep it where it lives, or host it with us. Either way, add the context.

Music to our ears

Most AI analytics products focus on query generation. The harder systems problem is context retrieval: before any LLM writes a single line of SQL. If that retrieval layer is fuzzy, the answer is wrong even when the SQL looks perfect. Altertable is built around that boundary.

𝙾𝚁𝙳𝙴𝚁 𝙱𝚈 𝚝𝚜 𝙳𝙴𝚂𝙲 𝙻𝙸𝙼𝙸𝚃 𝙽 is one of the easiest ways to accidentally benchmark your storage engine instead of your product. If metadata can’t prune most files, “recent events” quietly becomes a full-table tax. That’s not a query problem. It’s a layout problem.

There is no single “winning” lakehouse table format in 2026. What has emerged instead is a more interesting split. altertable.ai/blog/2026-04-1…

There probably won’t be one winning lakehouse table format. #Iceberg, #Delta, #Hudi and #DuckLake make different bets on metadata layout, mutation semantics, and engine interoperability. But what are your boundary conditions at 10k files, realtime ingest, and mixed engines?