en.Wedoany.com Reported - During the second-day keynote of the Databricks Data + AI Summit in San Francisco, company executives reiterated that rising AI costs are a core challenge for enterprises. CEO Ali Ghodsi noted that agentic AI is increasing IT resource consumption, and AI costs "will become extremely expensive, and enterprises have only scratched the surface."
In response to the need for agent oversight, Databricks released the meta-framework Omnigent. Co-founder and CTO Matei Zaharia described it as a "framework of frameworks," arguing that an additional open layer is needed above existing framework layers to manage the interface between language models and the world, as well as the composition, sharing, and control of agents.
Developers are given a key role in AI governance. Enhanced features of Agent Bricks and Unity AI Gateway support context-aware security policies, cost monitoring and budget setting, and intelligent routing to find optimal models. Product Director Kasey Uhlenhuth stated, "Agents are the most privileged roles, and this is what enterprise AI governance should look like."
On the product front, Databricks provided more details on AI Runtime—a serverless graphics processing unit for enterprise deep learning (in preview for two months)—while also releasing two machine learning agents: Genie Code for ML and Genie Zero Ops for ML. Product Management Director Mike Del Balso said, "What we really want to do is automate ML," and believes Databricks has more real-world machine learning experience than any other company.
The marketing field is being reshaped by agentic AI. Databricks launched CustomerLake, an agentic customer data platform (CDP), to help marketers shift from static, one-time campaigns to infinite engagement loops driven by agents, analyzing customer signals and acting across channels. Engineering Vice President Tasso Argyros remarked, "Humans no longer buy alone; commerce is becoming headless," and equipping each customer with an agent is seen as the future direction.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
