Databricks Launches Genie One, LTAP and Panther Deal at Data + AI Summit

Databricks used its Data + AI Summit in San Francisco this week to argue that AI agents need a simpler data estate, not just better models. The company launched Genie One, an agent for business teams with no per-user license fee, introduced a single-copy architecture for transactional and analytical data, and agreed to buy security startup Panther. For CIOs and CDOs, the message was that the platform holding the data now wants to run the agents, the context layer and the security operations too.

What did Databricks announce at Data + AI Summit 2026?

The headline launch was Genie One, which Databricks describes as an agentic coworker for marketing, finance, sales and other business teams. According to the Databricks press release dated June 16, it works across structured and unstructured data inside and outside Databricks, produces documents and reports rather than only answers, and can act in other tools through the Model Context Protocol. It connects to more than 50 business applications, including Slack, Jira, SharePoint and Google Drive, and is generally available alongside Genie Agents and Genie Code. Genie App Builder and Genie ZeroOps, a background agent that watches data and AI assets and proposes fixes, are headed to private preview.

The pricing is the part procurement teams will notice. Databricks said there are no per-user license fees. Organizations get up to $10 of free usage per user each month and then pay only for the AI they consume. That is a direct contrast with seat-based assistants, and it shifts the budgeting question from headcount to consumption.

Underneath Genie One sits Genie Ontology, a context layer that combines manually defined semantics, such as Unity Catalog definitions and glossaries, with a permission-aware ontology that ranks knowledge snippets, as Constellation Research described it. Constellation also reported that Genie Code is already used for about 60% of new data pipelines on Databricks.

What is LTAP, and why does it matter for AI agents?

The infrastructure announcement was Lake Transactional/Analytical Processing, or LTAP. It pairs Lakebase, Databricks' Postgres-based transactional database, with the lakehouse so that transactions and analytics run on one open copy of the data. The goal is to remove the ETL jobs, table syncs and duplicate copies that sit between operational and analytical systems. CEO Ali Ghodsi framed the old approach bluntly, calling complicated data infrastructure "a tax" that agents now turn into a bottleneck, per Constellation.

Databricks also put Lakehouse//RT, a real-time analytics engine that runs directly on Delta Lake and Apache Iceberg tables, into beta, according to RTInsights' weekly roundup. It is aimed at the high-concurrency, unpredictable query patterns agents generate. A Unity AI Gateway adds a single enforcement point for AI usage by people and agents, covering budgets, cost controls, tool governance and model routing.

The logic matches what practitioners have told AI Connect audiences. Data built for human analysts breaks when an agent reads it without a second look. "Agents don't do that. They take the data at the face value and act," Sam Mittal of Snowflake's Partner Solutions Engineering team said during a AI Connect webinar this week on why agents need metadata, not just clean data.

Why is Databricks buying Panther?

Databricks said it agreed to acquire Panther Labs, its third cybersecurity acquisition, as it builds a security lakehouse to compete with CrowdStrike and Cisco's Splunk, Reuters reported on June 16. Terms were not disclosed. Panther was last valued at $1.4 billion after a $120 million Series B in 2021, according to the report. Ghodsi argued that AI has shortened the time attackers need to exploit software flaws and that traditional security information management is finished. "If they're going to attack you with agents, you have to defend with agents," he told Reuters.

For security leaders, the pitch is that security telemetry is just another large data set, and that putting it next to the rest of the enterprise's data makes agent-driven detection cheaper. The move signals that Databricks sees security operations as an early home for agents.

What does this mean for enterprise data and AI leaders?

Constellation noted that the summit came right after Snowflake's annual conference, and that the two companies are competing to be the default data and AI layer. Partners lined up behind Databricks' governance model: Collibra, Immuta and Salesforce all announced expanded Unity Catalog or Databricks integrations, according to RTInsights.

Three practical questions follow. First, consumption pricing for agents like Genie One makes cost visibility a day-one requirement, so leaders should decide who owns AI spend before usage spreads. Second, single-copy architectures such as LTAP are only as useful as the data quality and semantics on top of them, a point AI Connect speakers have made repeatedly in sessions on the context layer for enterprise AI. Third, as more agents act on production data, monitoring matters as much as building; the warning signs are covered in our piece on AI agent observability.

The next thing to watch is how quickly Genie App Builder and ZeroOps move out of preview, and whether consumption pricing holds up once business teams start using agents at scale.

Frequently asked questions

What is Databricks Genie One?

Genie One is an agentic coworker for business teams that works across structured and unstructured data, produces reports and documents, and acts in other apps through MCP. It is generally available.

How is Genie One priced?

Databricks said there are no per-user license fees. Organizations get up to $10 of free usage per user per month and pay only for the AI they use.

What is LTAP?

Lake Transactional/Analytical Processing pairs Databricks' Lakebase transactional database with the lakehouse so transactions and analytics run on one open copy of data, reducing ETL and duplicate copies.

Filed under: Agentic AI, Data & AI Infrastructure, Context, Semantics & RAG