Databricks Releases New AI Agents

 

What can they do and how much this costs?

 

Databricks introduced a suite of advanced AI agents designed to automate data analytics, machine learning workflows, and system maintenance directly within enterprise environments.

These tools are built on the "Genie Ontology" data context layer, which grounds AI agents in an organization's real-time data. Key capabilities and pricing details include:

What the New AI Agents Can Do

Databricks offers specialized agents tailored for different technical and business users:

•           Genie Code (for Developers): An autonomous coworker that helps data teams plan, build, and run data engineering, machine learning, and analytics workflows. It automatically explores datasets, iterates on experiments, and develops and tests models within notebooks.

•           Genie Agents & Genie App Builder (for Business Users): Allows non-technical staff to "vibe-code" their own AI agents and applications, enabling them to query data and build custom solutions using natural language.

•           Genie ZeroOps (for IT/Data Assets): A background agent designed to automatically monitor, investigate, and propose fixes for data pipelines, jobs, tables, and ML models.

•           Agent Bricks: A developer platform designed to help teams easily scale, test, and build custom multi-modal agents that can reason over large datasets.

How Much It Costs

Databricks does not publish fixed, flat-rate pricing for these agentic products, as they build upon the underlying Databricks Data Intelligence Platform infrastructure. Instead, costs are calculated dynamically based on three main pillars:

•           Compute Resources: Usage is metered in Databricks Unit (DBU) consumption, which varies based on the virtual machines or serverless compute used to run the agents.

•           Model Serving Costs: Hosting models that power these agents relies on your choice of underlying foundational models (such as those via Databricks Model Serving).

•     Token Usage: Advanced agentic reasoning generates more tokens per request than traditional prompts. Because of this, Databricks introduced the Unity AI Gateway, which features built-in cost controls, per-user alerts, and hard token-level budget limits to prevent runaway AI bills.

________________________________________

To get an exact quote tailored to your organization, it depends heavily on your existing Databricks plan and the scale of your compute needs.

 

 

“Databricks is releasing AI agents that help professionals get answers from their business data, aiming to expand beyond its core data offerings and showcase its staying power in the artificial-intelligence world.

 

The San Francisco-based company, which is valued at $134 billion, calls its new offering Genie One, or an "agentic co-worker" that helps business teams -- from finance to marketing and sales -- get answers and make decisions based on their corporate data.

 

The move comes as data companies, including Databricks rival Snowflake, start to cement their place as providers of enterprise AI. In late May, Snowflake's stock shot up following strong earnings and increased demand for its AI tools.

 

Like Snowflake, Databricks provides the underlying layer for storing and organizing data in the cloud.

 

Databricks was established in 2013, eons before the launch of OpenAI's ChatGPT. But since the start of the AI boom, data companies have said they can help businesses make use of AI by providing the data context the technology needs.

 

When Databricks launched its natural language interface, Genie Spaces, a few years ago, the idea was that data scientists could more easily extract insights from their data without writing complex queries, said co-founder and CEO Ali Ghodsi. But the company quickly discovered that customers were sharing this capability with their marketing departments and with finance and senior executives.

 

"They really pushed us and said, 'Hey, this is really magical, but Databricks is not built for these departments. Can you build something that's completely simpler?'" Ghodsi said.

 

That led to the creation of the new Genie agents, which aim to solve that customer need.

 

Now, there are signs the data and AI message is sticking.

 

The company is generating an over $1.7 billion revenue run rate from its AI products, Ghodsi said. That is an increase from the over $1 billion revenue run rate the company said it was generating from its AI products last September.

 

Ghodsi said the "secret sauce" behind its new AI agents is a data context layer called Genie Ontology -- essentially a graph of all knowledge in an organization, including data, content, apps, documents and people that is updated in real time. Having that context leads to more accurate and faster responses from AI, as well as lower token costs, Ghodsi said.

 

Also on Tuesday, Databricks announced Genie Agents and Genie App Builder, products targeted at business users who want to vibe-code their own AI agents and applications. The company has a separate AI agent for developers called Genie Code.

 

With these releases, Databricks enters a crowded space of providers offering coding agents and general AI agents for knowledge work. Ghodsi said the Databricks agents are meant to be used alongside agents like Anthropic's Claude Code, though he expects all agents to eventually be tailored for specific areas of work.

 

"I actually think we're going to see specialization," Ghodsi said. "This is the specialization that we are good at: data. So we're just focused on that."

 

Albertsons and Rivian have been early users of Databricks' new AI agents.

 

Karthik Iyer, a group vice president and head of merchandising transformation and AI at Albertsons, said the grocery store operator is using the tech to help answer questions based on company data.

 

"A merchant is now able to ask questions like, 'If I promote Sargento cheese, what will it do to my own brands, and how much shelf space do I need to allocate to my own brands to compensate for the promotions that I'm running with Sargento cheese?' " Iyer said.

 

Romit Jadhwani, a senior director of enterprise AI, data and productivity at Rivian, said the electric-vehicle startup uses Databricks agents as a way to "get value from your data" without needing to write programming queries.

 

Leaders at Rivian are using the agents to help understand demand forecasts, how the company is performing in production operations and reviewing financial metrics, Jadhwani added.

 

Databricks has long been considered one of the most highly anticipated potential initial public offerings in the startup market. Ghodsi said Databricks is excited to watch the year's blockbuster IPOs, but likely won't IPO this year.

 

---

 

Belle Lin writes for WSJ Leadership Institute's CIO Journal.” [1]

 

1. Databricks Releases New AI Agents. Lin, Belle.  Wall Street Journal, Eastern edition; New York, N.Y.. 17 June 2026: B4.