Data storage giant Databricks Inc. rolled out a suite of artificial intelligence (AI) tools on Tuesday aimed at curbing runaway corporate AI expenses, including Genie One, an agentic co-worker designed to help non-technical business teams seamlessly extract insights from corporate data.
The dual product launch signals Databricks’ strategic push beyond its core data storage offerings and positions the company directly against rivals like Snowflake Inc. in the race to provide foundational enterprise AI infrastructure.
The genesis of Genie One stems from customer feedback on Databricks’ previous natural language interface, Genie Spaces. While originally intended to help data scientists bypass complex programming queries, company executives noticed that customers were frequently sharing the tool with corporate leadership, finance, and marketing departments.
“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?'” Databricks CEO Ali Ghodsi said.
The resulting Genie agents rely on a data context layer called Genie Ontology, a real-time knowledge graph that maps an organization’s internal data, documents, and applications, leading to faster, more accurate responses at a lower token cost.
Databricks also introduced Genie Agents, Genie App Builder, and a developer-centric tool called Genie Code. Early adopters include grocery giant Albertsons Companies Inc., which uses the technology to forecast promotional impacts, and electric-vehicle maker Rivian Automotive Inc., which utilizes the agents to review demand forecasts and financial metrics without writing code.
The strategy appears financially lucrative; Databricks’ AI product line now generates an annual revenue run rate exceeding $1.7 billion, up from $1 billion last September. Despite this momentum and heavy anticipation surrounding a potential initial public offering, Ghodsi indicated that Databricks will likely bypass an IPO this year.
As autonomous AI agents increase software usage across industries, corporate budgets are buckling under unpredictable costs. Databricks co-founder Patrick Wendell revealed that some enterprise clients have seen AI token costs spike from zero to tens of millions of dollars in a single month.
“We’ve definitely seen mistakes that are in the millions,” Wendell said, noting that AI token costs now rank among the top three corporate expenses, trailing only salaries and general IT.
To address this financial friction, Databricks introduced the Unity AI Gateway. The new tool establishes strict spending limits and guards against “runaway spend” by automatically recommending cheaper alternative AI models for simpler tasks.
Further, the gateway monitors individual employee sessions to optimize efficiency. While tracking individual usage can spark privacy concerns, Wendell minimized the potential friction, noting that the data collected is narrow and strictly utilized for cost control rather than model training.
Ultimately, Databricks hopes these guardrails shift corporate behavior from token maxing to value maxing. However, while robust cost controls will protect enterprise bottom lines, they may simultaneously pressure major AI model providers who rely on unchecked enterprise usage for revenue growth.
