Databricks' New AI Agent That Tells the CEO the Unvarnished Truth

Databricks has introduced an updated version of its AI agent, Genie, which can autonomously execute multi-step tasks by building context from customer platform data. Company CEO Ali Ghodsi shared how he uses this tool in internal processes such as employee satisfaction surveys and failed sales calls.

In an announcement on Wednesday, Databricks unveiled the new version of Genie, an AI agent capable of completing long-term and multi-step tasks by utilizing data stored by customers on the platform. According to the company's statement, corporate leaders can assign broad tasks to this agent, and the system completes them autonomously.

CEO Ali Ghodsi demonstrated the product's capabilities using his own actual chats and charts. Ghodsi stated that he used Genie to analyze employee satisfaction surveys where the company scored low, evaluate solution proposals submitted by managers, and summarize a negative sales call. He also noted that the tool prepared him for a podcast interview by reviewing his past appearances and preparing potential questions.

Ghodsi stated that this allows leaders in large companies to learn the raw and unvarnished truths that are difficult to hear. Noting that employees in a company with 12,000 employees might hesitate to convey negative information directly to the CEO, Ghodsi referred to Dunbar's number and emphasized that many companies with more than 150 employees could benefit from such agents.

Databricks actively uses this agent, which works continuously for leaders, in its own internal operations as well. On the other hand, Ghodsi drew attention to Databricks' Unity Gateway product against the risk of increased costs that the use of artificial intelligence might bring. It was stated that through this product, companies can set usage limits and transition to cheaper open-weight models.