Databricks' primary product is a technology for holding huge troves of corporate data in a way that makes it easy to access and analyze. But doing so still typically requires a data scientist to write computer code that finds and manipulates the data.
The new system, called LakehouseIQ, will instead let users ask questions naturally, without using computer code. Behind the scenes, an AI system will interpret the question, fetch the needed data, read it and produce an answer. Ali Ghodsi, chief executive of Databricks, hopes that the AI system will be especially useful because it will be trained on a company's own data, rather than generic data from the internet. That should get the AI quickly up to speed on relevant information like the dates of the company's fiscal year or industry-specific jargon, Databricks believes.
"There's lots of three letter acronyms in any organization. They mean lots of different things, and not even everyone in the company knows it," Ghodsi said. By training on the customer's specific data, the new Databricks offering "understands the jargon. It understands the domain you're working in."