For two years, the self-serve pitch has been the same.
Bring your business users into the workspace, give them a Genie space, and watch adoption climb.
It works for the people who will open a new tab. It does not work for the largest analytics population in almost every company, the ones who never leave the spreadsheet.
That population just got Genie One, where they already sit. As of last week (5 August 2026 to be exact), Genie One is a natural-language feature in the Databricks Excel Add-in and the Databricks Connector for Google Sheets, both of which are generally available 🚀
You ask a question in plain English, and the answer lands as native rows and columns in the sheet. No SQL, no workspace, no export dance.
I want to be precise about what is new, because it is easy to wave this off as an old feature.
The Genie layer inside the connector is what changed.
Databricks has had spreadsheet connectivity for a while. You could browse Unity Catalog tables, write SQL, and pull results into Excel or Sheets. That is the connector, and it is aimed at people who are comfortable picking a table or writing a query. Useful, but still a technical surface.
What shipped this month is a Genie One button living inside that same connector. A user types "what were the top ten accounts by revenue last quarter in EMEA" and Genie One generates the query, runs it against governed data, and drops the result into the grid. The person asking never sees a table name or a SQL statement. That is a different audience entirely.
Why the spreadsheet is the real total addressable market?
Think about who actually consumes data in a bank, a retailer, or an insurer. Analysts are a small slice. The bulk are finance business partners, commercial managers, ops leads, and controllers. Their working surface is a spreadsheet, and it has been for their whole career. Every prior attempt to move them into a BI tool or a chat window fought that habit and usually lost.
Genie One in the sheet stops fighting the habit. The question gets asked in the tool they already trust, and the answer arrives in the format they already use. For a platform team that has spent a year trying to lift workspace adoption among non-technical staff, this is the shortcut. You are not asking anyone to change where they work.
What actually travels with the answer?
This is the part that matters for anyone responsible for governance. The query runs against governed Databricks data through Unity Catalog, so entitlements are enforced on the underlying tables, not bolted on inside the spreadsheet. A user only ever gets rows they are permitted to see. The sheet is a delivery surface, not a new permission boundary you have to reason about separately.
That is the quiet reason this is safe to roll out. The governance you already defined in Unity Catalog is the governance that applies when the answer lands in Excel. You did not build a second model.
How do you actually turn it on?
There is no heavy setup for the end user.
Once the connector or add-in is installed and connected to a workspace and an SQL warehouse, the Genie experience appears as a button.
In the Google Sheets connector, you open the connector, click the Genie icon, and start asking.
The Excel Add-in follows the same shape. The heavy lifting sits upstream, with the platform team, in the one-time install, the workspace connection, and the SQL warehouse where the queries run.
The person asking the question does none of that. This split is deliberate, and it is why the surface is realistic for a non-technical cohort: the complexity lives with the team that owns the platform, and the business user gets a text box.
The two surfaces are close but not identical.
Both the Excel Add-in and the Sheets connector support Genie One natural-language queries and import results as native rows and columns. There are differences worth knowing before you pick where to start. The Sheets connector supports scheduled refresh on an hourly, daily, or weekly cadence, which suits a recurring report. The Excel Add-in supports write-back to Unity Catalog tables and pivot tables, which is well-suited to heavier local modelling. Check the current docs for the exact per-surface feature list before you commit a team to one path.
For the developers: the same capability as an API.
If you want to embed this in an internal app rather than a spreadsheet, the Genie Conversation API is the building block you need. It lets you start a conversation, ask a natural-language question, retrieve the generated SQL, and pull back the results, with follow-up questions that keep context. It authenticates with OAuth for users, or OAuth for service principals, and a service principal needs permissions on the data and the SQL warehouse it uses. The spreadsheet integrations and a custom app are two front doors to the same governed engine.
Where I would be careful
Meeting users in the sheet is powerful, and it moves the hard questions rather than removing them. Three that I would plan for before a wide rollout:
Cost visibility: When a natural-language query reaches hundreds of casual users, each refreshing a sheet a few times a day, the query volume against your warehouse no longer comes from a handful of analysts. Decide how you monitor and attribute that spend before it surprises you, not after.
Trust in the answer: A number in a cell looks authoritative even when the question was ambiguous. The same semantic grounding that makes Genie reliable in the workspace, curated metrics and clear instructions, matters more here because the person reading the cell is less likely to sanity-check the logic behind it.
Sprawl you cannot see: A shared workbook containing a Genie query can be email to people who were never on your rollout plan. The entitlements still protect the data, but your view of who is asking what gets blurrier. Governed access holds. Observability is the thing to watch.
What good looks like.
The teams that will get value fastest share a few traits. They already have a curated Genie space with certified metrics, so the answers in the sheet are grounded rather than guessed. They pick one surface, Excel or Sheets, for a first cohort rather than opening both at once. They treat the spreadsheet rollout as an extension of an existing governed setup, not a fresh project with its own rules. And the instrument costs from day one.
The Paradigm Shift - Genie and Spreadsheets
Self-serve analytics has quietly assumed that users come to the platform. This flips it. The platform now appears in the tool that the user never leaves. For the finance and ops majority, that is the difference between an interface they were told to adopt and one they already have open. If you have been chasing business-user adoption by asking people to change their habits, this release lets you stop asking.
A few questions for you:
Have you rolled out Genie One in Excel or Sheets to a non-technical team yet, and did adoption actually move where workspace access did not?
How are you handling cost monitoring when natural-language queries reach a large, casual audience, refreshing sheets on a schedule?
Where do you land on the sprawl question: does answering in the spreadsheet grow trusted self-serve, or quietly shrink your visibility into who is asking what?
For more awesome discussion about Genie 🧞👇🏻
Feel free to connect with me directly or drop a DM: linkedin.com/in/lingeshwarankanniappan
References
Connect to Databricks from Microsoft Excel: docs.databricks.com/aws/en/integrations/excel
Databricks Connector for Google Sheets: docs.databricks.com/aws/en/integrations/google-sheets
Genie Conversation API: docs.databricks.com/aws/en/genie/conversation-api
August 2026 platform release notes (Genie One in Excel and Google Sheets, both GA): learn.microsoft.com/en-us/azure/databricks/release-notes/product/2026/august