Since the Data and AI Summit 2026, most of the Genie conversation has centred on the big headline themes: Genie One, Genie Ontology, Agents, Teams, Copilot, and broader agentic workflows. Those topics deserve the attention they are getting. But one part of the announcement feels underexplored so far, and it matters for very practical reasons. That is Genie in Excel.
Here is what stands out to me. Databricks is not only growing Genie as a conversational experience inside Databricks. It is pushing Genie into the places where business users already spend their time. The summit material described the Databricks Excel Add-in as a Public Preview capability that brings the lakehouse directly into spreadsheets, with no SQL to write and no per-user ODBC drivers to configure. The same announcement says support for Unity Catalog metric views lets data teams define business logic once and make it available in Excel and beyond, in a governed and consistent way. The June 2026 release notes separate add that users can write data from Excel back to a Unity Catalog table, which closes the loop between analysis and action.
That may read like a product detail, but I think it is bigger than that. If chat-based Genie is about meeting users in conversation, Genie in Excel is about meeting users inside the working surface, where many real decisions still happen. For finance, planning, commercial, and operations teams, Excel is not an edge case. It is often the last mile of business work. Budgets get adjusted there. Forecasts get challenged there. Numbers get checked and annotated there. If Genie can show up in that environment with governed business logic, that is not just another integration. It changes how self-service analytics lands with the people who own recurring business processes.
Why does Excel specifically change the conversation
This topic deserves its own thread because it forces a more grounded discussion than a generic "Can Genie work in other tools?"
We already know the answer is YES!
The more interesting question is what changes when the tool is Excel. Spreadsheet work has its own habits, risks, and strengths. People trust cells they can inspect. They reuse tabs for monthly processes. They copy formulas forward. They annotate numbers before a meeting and compare versions side by side. A Genie plus Excel workflow is therefore not only about access. It is about trust, repeatability, and the handoff between governed data and familiar business practice.
Why metric definitions are the real test
Many Genie discussions rightly focus on semantic accuracy and the explicit articulation of business logic. Excel is exactly where weak definitions get exposed. If "net revenue," "pipeline," or "active customer" means one thing in a dashboard and another in a spreadsheet pack, users notice immediately. The summit positioning around Unity Catalog metric views points to a cleaner model. Define business logic once, govern it centrally, and let users consume it in the tools they already use, including Excel. That is a more concrete and testable value proposition than talking about AI in the abstract.
The workflow, in concrete building blocks
The building blocks are now visible in product documentation. The AI/BI release notes state that Unity Catalog metric views can be imported into Excel as pivot tables through the Databricks Excel Add-in. The Excel setup documentation states that the add-in uses single sign-on, requires Unity Catalog, and connects through a Databricks workspace and a SQL warehouse. The June 2026 release notes state that users can create a new table or overwrite an existing table in Databricks without leaving Excel. Chat in Genie One is documented as a unified interface that answers natural-language questions across your existing Genie Spaces, dashboards, and queries. The Excel Add-in announcement also lists AI integrations among the capabilities coming next, which points toward Genie reasoning landing inside the add-in itself.
Put together, a finance workflow could be composed of layers rather than shipped as a single monolithic feature:
Metric views define the certified business logic for measures such as revenue, pipeline, forecast variance, or working capital.
The Excel Add-in surfaces those governed metrics inside the spreadsheet interface that finance users already trust.
Genie provides the natural-language layer so that a user can ask for a bridge, a variance explanation, or a filtered slice of the same governed metrics.
Write-back sends approved adjustments or planning inputs back into Unity Catalog tables for users with the right permissions.
Downstream workflows pick up those writes for refreshes, reconciliation, notifications, or posting into the next finance step.
The broader point is that better-governed ingestion and orchestration make Genie more accurate and more useful, because the agent works from a more complete enterprise context.
What is real today, and what is not
This is why I think the idea is technically credible right now. The individual mechanics exist. Governed metric views exist. Excel connectivity exists. Table-oriented write-back exists. Genie can already reason across metric views and other governed assets. What the public evidence does not yet show is a single shrink-wrapped financial planning product with Excel-native approvals, locking, commentary, exception routing, and built-in end-to-end transactional posting at the cell level.
That caveat matters, and the wording in the release notes is precise. Write-back refers to creating a new table or overwriting an existing one. That is enough to support serious workflows, but it is not the same as a full finance planning control plane inside Excel. So the best way to frame this topic is not as a finished SKU. It is a well-composed Databricks pattern that is now close enough to real product capabilities to warrant serious discussion.
The honest test of "meet users where they work"
Genie in Excel is one of the clearest tests of whether Databricks is serious about that phrase. The broader Genie One story is explicitly about working across everyday business tools, not just inside a single product surface. Excel is probably the hardest honest test of that idea, because it is where business users are most comfortable, most opinionated, and least willing to change their habits for the sake of platform purity. If Genie works well there, the adoption story gets much stronger. If it does not, that gap will show up quickly.
For that reason, I think this is a strong next topic for the community. It sits at the intersection of adoption, governance, semantics, and workflow design. It is close enough to the summit announcements to be timely, but practical enough to invite real implementation stories instead of abstract speculation. It also gives both technical and non-technical members something to contribute. Data teams can talk about metric views, permissions, and source design. Business users can talk about how spreadsheet workflows actually work in the real world. That is usually where the most valuable discussion starts.
If this became a dedicated thread, I would be especially interested in a few concrete questions:
Are people planning to use Genie in Excel mainly for retrieval, or for two-way workflows as well?
Which personas benefit most first: finance, sales operations, procurement, or something else?
How should teams decide when a process belongs in Dashboards, when it belongs in Genie chat, and when Excel is the right interface?
If metric views become the bridge into Excel, which definition gaps do they expose fastest?
What guardrails do people want before they trust write-back in a production business process?
Genie in Excel deserves more than a passing mention inside a larger launch summary. It is not the flashiest announcement, but it may be one of the most consequential for day-to-day adoption.
A lot of enterprise analytics does not fail because users dislike insight. It fails because the insight arrives in the wrong place, at the wrong time, in the wrong format. Excel remains one of the places where business work turns into decisions. If Genie can operate there with governed logic and low friction, this is not a side integration. It is a serious step toward making Databricks useful in the actual flow of work.
If you want to read my previous articles in this space, feel free to follow me on LinkedIn