With the latest capabilities added to its Genie family of agentic AI-powered features, Databricks is trying to alter how enterprise work gets done.
On Wednesday, Databricks unveiled always-on Genie, which turns Genie agents from assistants that users call on when needed into autonomous workers that can be given an assignment or goal and take ownership of the work; the always-on agents can even assign tasks to subagents when needed.
In addition, Databricks made Genie Ontology, a knowledge layer first introduced in June, generally available. Genie Ontology captures a company's context, including definitions and metrics across both structured and unstructured data no matter where it lives, and make it available to agents. Finally, the vendor unveiled Genie Search, a governed way -- grounded in Genie Ontology -- for humans and agents to search for information across their entire data estate.
"Taken together, I'd call [the new capabilities] significant because Genie stops waiting to be asked," Mike Leone, an analyst at Moor Insights & Strategy, told TechTarget. "Up to now, getting value out of Genie meant knowing the right question and when to ask it. Handing it a goal instead puts answers in front of folks who would never open a dashboard in the first place."
With the general availability of Genie Ontology, Databricks is among the first to deliver an ontology layer for its users. However, most Databricks competitors -- including rival Snowflake and hyperscalers AWS, Google Cloud and Microsoft -- are similarly developing tools that empower agents to become autonomous coworkers, Leone continued.
"Databricks is in step with the market on the overall idea, but it pulls ahead on the ontology," Leone said. "It picks up a company's definitions from how people actually query the data, so a team doesn't have to write them all out first. The entire data platform market is chasing ontologies right now, including ones AI builds and manages with people signing off on changes, so Databricks has plenty of company."
From assistance to autonomy
Since its introduction in June 2024, Databricks has continuously expanded Genie to include Genie One for natural language-based data analysis, Genie Agents to help users develop domain-specific agents built on reasoning from curated datasets and business logic, Genie Code to help data engineers and scientists write and debug code for building pipelines, and Genie Ontology.
Now, Databricks is expanding Genie to include Search and make Genie agents more autonomous.
"Many of these features were always in the plan, but customer feedback definitely influenced our thinking," Ken Wong, senior director of product management at Databricks, told TechTarget.
For example, Ontology Manager, a feature in Genie Ontology that enables data teams to govern business context and includes recommendations for refining ontologies that are most frequently used to inform agents' actions, was developed in response to requests from customers. Expanding the sources Genie Ontology pulls business logic from to include external business intelligence platforms so agents can better understand analytical data was also a direct result of customer feedback.
"Most significantly, customers increasingly asked Genie questions that depended on information and context outside Databricks and traditional data systems, like context captured in documents, tickets and chats," Wong said. "And while we had federated connectors at launch, we wanted to boost performance, so we built Genie Search to index those systems as well."
"Taken together, I'd call [the new capabilities] significant because Genie stops waiting to be asked. Up to now, getting value out of Genie meant knowing the right question and when to ask it. Handing it a goal instead puts answers in front of folks who would never open a dashboard in the first place."
Mike Leone
Analyst, Moor Insights & Strategy
Perhaps the most valuable of the new Genie capabilities is Genie Ontology, according to Sanjeev Mohan, founder and principal of analyst firm SanjMo.
He noted that making Genie an always-on coworker is the most eye-catching new feature. Genie Ontology, however, is what makes all other agentic AI capabilities possible.
"Always-on Genie will get the headlines, but it only works if the agent understands the business," Mohan told TechTarget. "By now, we understand that agents fail less from bad models than from missing context."
For example, if agents don't know what the term "active customer" means at an individual enterprise, or which revenue number the finance department trusts, they lack the context to deliver accurate responses.
"Ontology learns those definitions, metrics and relationships from the data and from usage," Mohan said. "Pulling in business logic from external BI tools stands out, because much of that logic lives outside Databricks today."
In addition to Ontology Manager and expanding the sources Genie Ontology pulls from, new capabilities that have been added to the knowledge layer since it was introduced in preview include the ability to generate live dashboards and interactive reports rooted in governed data, agents that can work across data sources alongside governed tables, and expanding users' access to Genie so they can interact with agents in any workspace, including web browsers and mobile applications.
"Until now, most enterprise AI has been reactive: a user types a prompt and waits," Mohan said. "Genie Search and Ontology give … access to all relevant data, plus an understanding of what that data means. Users can now hand off recurring analytical work instead of doing it by hand."
However, like Leone, he noted that competing vendors are providing similar capabilities.
"Every major data platform and hyperscaler is converging on the same pattern: a semantic or context layer, unified search across structured and unstructured data, and agents that act on both," Mohan said.
While Mohan highlighted Genie Ontology, Leone tabbed Genie Search as perhaps the most valuable of Databricks' new features because it unifies searches across entire data estates so that agents no longer have to run individual searches for each data source.
"Genie Search stands out because it changes how agents reach information spread across a company," he said. "Right now, an agent hopping from app to app needs its own access to each one, and every extra stop adds to the bill. One governed search gives you one place to set permissions and a lot fewer calls to pay for."
Greater governance still needed
The new Genie capabilities from Databricks are appropriately designed to make agents more autonomous, according to Mohan. But before Databricks customers can truly assign tasks to agents and empower agents to do complex work on their own, Databricks needs to add capabilities that not only govern access to data but also govern what agents do in production.
"They make real progress on governed access, [but] enterprises need to govern what agents do, not just what they read," Mohan said. "That means a clear identity for each agent, limits on which actions it can take, human approval for consequential steps and a full audit trail of its reasoning and actions."
Leone similarly noted that Databricks is headed in the right direction as it attempts to turn Genie agents into autonomous coworkers. However, like Mohan, he suggested that more transparency is needed.
"I think Databricks gets most of the way on keeping roaming agents in bounds, [but] I'd add a way for always-on Genie to say when it isn't sure of an answer," Leone said. "A wrong answer that shows up unprompted can reach a lot of people before anyone thinks to question it."
Eric Avidon is a senior reporter for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.
