Databricks Mosaic AI

Databricks Mosaic AI is a toolkit for building machine learning and generative AI on top of company data. Data teams use it to train and fine-tune models and build AI agents that answer questions using that data. It is the AI layer of the Databricks data platform. The built-in MLflow tool tracks experiments, and Unity Catalog gives admins one place to set access rules, track data lineage, and audit model use across teams. It includes AI Search, a vector database for retrieval-augmented generation, and a document intelligence tool that reads and classifies files such as contracts or invoices. It runs on AWS, Azure, or Google Cloud, so companies keep existing cloud contracts. Databricks bills usage by Databricks Units, a consumption-based unit tied to compute use.

This fits a company that already stores data in Databricks, or is choosing a data platform and wants built-in machine learning and AI agent tools. A good starting point is one team piloting a single use case, such as a support-ticket summarizer, before rolling out further. Databricks offers a 14-day trial with about $400 in credit. Ongoing costs scale with compute and data volume and are hard to predict without a sample workload, so budget for both Databricks Unit charges and the cloud compute they run on. The main risks are cost creep from unmonitored jobs and a learning curve for staff used to simpler, spreadsheet-based tools. Unity Catalog helps control access, but someone still needs to own governance rules day to day.