Databricks Buys Row Zero for Planned Governed Genie Spreadsheets

Databricks Buys Row Zero to Put a Spreadsheet Inside Genie​

In its announcement, Databricks says the deal will expand the capabilities of Genie, Databricks' AI coworker, which helps business teams turn data into trusted answers and actions. Genie already answers plain-English questions about company data. The company's examples are explaining why margins changed and writing a document about sales-pipeline opportunities. Row Zero adds a grid where people can check those answers, change them and build models on top.

Databricks names three pieces of its platform underneath this. Row Zero will integrate with Databricks' Genie Ontology, Unity Catalog and Unity Gateway. Unity Catalog is Databricks' governance layer for data and permissions. Genie Ontology supplies the shared business context that Genie uses to understand terms. The company says the goal is to let users move between Genie's conversational interface and spreadsheets for tasks such as pivoting, modelling and visualisation.

The planned version will reach more than a browser tab. Databricks will integrate the technology into Genie, its AI assistant for business users, across web, desktop and mobile. Row Zero's staff are moving over as well. Row Zero was founded by former AWS and Tableau engineers. The team will join Databricks and continue working on Genie.

The two products already work together. Techzine Global notes that Row Zero previously integrated with Databricks SQL Warehouses and Genie, allowing users to ask questions in natural language and receive the answer in a linked spreadsheet. Superpower Daily describes the change as a reversal: today's connection starts in Row Zero, and Databricks now wants to reverse that path: start in Genie, then open the answer for editing and collaboration in a spreadsheet. The same outlet states plainly that the native experience is still a plan, not a launched product.

The Spreadmart Problem Databricks Says Row Zero Solves​

Databricks' reasoning targets a habit every data team knows. A dashboard exists, but someone still exports the numbers to a spreadsheet, emails it round and makes changes locally. Databricks calls these files "spreadmarts." The company argues that every export adds to what it calls possibly the biggest set of ungoverned data pipelines in a company. It also warns that AI agents working on those files end up handling sensitive data whose origin and quality nobody can vouch for.

SecurityBrief puts the risk in practical terms. Data often leaves governed systems when employees export it into spreadsheets or CSV files and circulate copies among colleagues. That can make it harder to track how information is used, which version is current and whether access controls remain in place. In other words, once a CSV leaves the warehouse, the warehouse's access rules stop applying to it.

Databricks' answer is to keep the spreadsheet connected to the source. Per Databricks' description, queries follow a user's permissions, data refreshes automatically, and administrators can restrict exports. Users can also write results back to the source instead of saving them to a separate file. Databricks says every interaction is auditable and that security controls stay attached to the data.

The AI-agent angle is newer. Databricks says that because agents work in the spreadsheet through Row Zero's agent tools, their actions stay readable to the business users, since they use ordinary spreadsheet formulas and logic. SecurityBrief puts it this way: Row Zero allows teams and software agents to work through the same spreadsheet structure, making automated actions easier for staff to inspect and review. In practice, an analyst can inspect an agent's pivot or formula the same way they would a colleague's. Databricks hasn't said what fields the audit logs record, how long they're kept or what the review process looks like.

Row Zero Positions Itself Beside Excel, Not in Place of It​

For Microsoft shops, Excel is the obvious comparison. Databricks says Row Zero is designed to complement Microsoft Excel and Google Sheets, with familiar formulas, pivot tables and keyboard shortcuts so users can reuse what they already know. The key word is complement. Nothing in the announcement says Excel workbooks, macros, add-ins or VBA will run unchanged, and Databricks hasn't published a compatibility list.

Scale is the selling point. Databricks says Row Zero's engine lets business users work with billions of rows at interactive speeds. That is a vendor claim, and no benchmark method, test workload or latency figures have been published. According to TechCrunch, Databricks' own finance team adopted Row Zero first, because it was a cloud tool that could scale beyond 1 million live spreadsheet rows. iTWire describes it as much like Microsoft Excel or Google Sheets but at ludicrous scale.

The realistic reading is a split of work. Excel stays where it already is: desktop modeling, macros and files people share with each other. Row Zero goes after the large, live, governed datasets where an export to Excel is exactly what the security team wants to prevent. Databricks CEO Ali Ghodsi summed up the strategy on X, as quoted by TechCrunch, as a plan to combine business intelligence, agents and spreadsheets.

Today's Row Zero Genie Connector Shows How Permissions Actually Flow​

The claim that queries "honor each user's permissions" is true or false depending on how the connection is set up. Row Zero's current Databricks documentation describes two ways to log in, and they behave very differently. The connector works on AWS, Azure and Google Cloud, so organizations running Databricks on Azure are covered by today's version.

With OAuth, each user signs in with their own Databricks credentials. Row Zero's documentation says OAuth queries respect the database permissions of the person running the query, including row-level security and role-based access control. Setting up the OAuth app requires Databricks account admin rights and the scopes sql, offline_access and genie. The customer then schedules a session with Row Zero to finish the Row Zero side of the setup. The client secret must be passed over a secure channel, not email.

A personal access token (PAT) is different. Row Zero recommends it for individual accounts or for a dedicated service account in larger organizations. A PAT connection can then be shared with named users or groups inside Row Zero. Everyone using that shared connection queries with the service account's permissions, not their own. An admin who assumes per-user row-level security applies to a shared PAT connection would be wrong.

Prerequisites for Genie in Row Zero today​

Row Zero's documentation lists these requirements for the current Genie integration:

1. An account admin must confirm that Databricks partner-powered AI features are turned on at the account level and not switched off for the workspace, because Genie Spaces depend on them.
2. Space authors need the Databricks SQL workspace entitlement, CAN USE on a pro or serverless SQL warehouse, SELECT on the Unity Catalog data in the space, and CAN EDIT or CAN MANAGE on the Genie Space.
3. Space users need Consumer or Databricks SQL access, SELECT on the underlying Unity Catalog objects, and CAN VIEW or CAN RUN on the Space.
4. The Genie Space must be shared with whichever identity the Row Zero connection uses. That means each user for OAuth, or the service account for a PAT.
5. The connection needs the genie scope. For a PAT, that means creating a new token with Genie permission enabled and updating the connection. For OAuth, add the scope to the app connection and contact Row Zero support to complete the change. Enterprise customers must also set up an AI provider in Row Zero.

When it works, Row Zero finds the Genie Spaces available to the connection and uses the selected Space to generate SQL. The results come back as connected tables. If "no Genie spaces are accessible," the documentation points to three causes: the Space isn't shared with the connection identity, required entitlements are missing, or the PAT was created without Genie permission. For "permission denied," check the Space's access lists, Unity Catalog SELECT grants, workspace entitlements and access to the SQL warehouse.

All of this describes the current connector. Databricks hasn't said whether the built-in Genie version will use the same setup, so treat it as a guide to how things work now rather than a promise about the future.

Row Zero Deal Terms Stay Private as Databricks Keeps Shopping​

TechCrunch reports that the terms weren't disclosed. It adds that Row Zero raised $10 million in May 2025 at an estimated $40 million valuation, citing PitchBook. That figure is an old valuation, not the price Databricks paid.

Databricks has been buying startups steadily this year. TechCrunch lists Quotient AI and SiftD.ai in March, the security-operations startup Panther in June, and Electric, the company behind the PGlite Postgres database, last month. Ghodsi told the outlet that Databricks intends to make many more acquisitions like this one.

Databricks' commitments on availability are broad but not dated. The spreadsheet platform will be available to Databricks customers across major cloud providers and will continue to support data sources outside Databricks. Row Zero's materials currently list connectors for Snowflake, Redshift, Postgres, Athena, BigQuery and S3. The announcement doesn't say whether every one of them will stay on the same terms, and it gives no pricing or rollout schedule.

What this means for you​

Only organizations already on Databricks, or using Row Zero today, have anything to act on now. Everyone else can wait until Databricks ships the Genie version and publishes documentation. Admins at Row Zero shops should check how their connections are authenticated. Anyone assessing the planned Genie spreadsheet should ask how identity, export locks and audit records will work before it replaces any Excel-based process.

• Check whether your Row Zero Databricks connections use OAuth or a shared PAT, because only OAuth applies each user's own permissions, including row-level security.
• Make sure any shared PAT belongs to a dedicated service account with the narrowest Unity Catalog grants that still work, since everyone sharing the connection gets its permissions.
• Confirm that partner-powered AI features, Genie Space permissions and the genie scope are set up correctly before blaming Row Zero for missing Spaces or permission errors.
• Don't plan Excel migrations around Row Zero yet, because Databricks describes it as a complement to Excel and hasn't published which workbook features, macros or add-ins carry over.
• Treat the billions-of-rows performance figure as a vendor claim until you can test it on your own warehouse workloads.
• Ask Databricks for the audit-log contents, export-lock controls and write-back approval process before letting agents change governed data through the spreadsheet.

Databricks has bought a spreadsheet that already worked with Genie and is promising to make it a built-in, governed part of the platform, targeting the export-and-email habit that governance teams have long struggled to stop. The next concrete step is Databricks shipping that built-in Genie experience and documenting how its permissions and audit logs work. Until then, whether any organization gets the governance benefits comes down to how it authenticates Row Zero today.