Product News
Six weeks on: the Excel-for-lakehouse market in April 2026.
Databricks has simplified admin-driven installation of its Excel add-in and named AI on its roadmap. Exponam has gone multi-cloud, shipped natural-language query with a private or bring-your-own LLM, and added a full SQL scratchpad. An honest look at what has actually changed.
Exponam, LLC · April 2026
Updates since publication
April 25, 2026. We have attempted to install the Databricks Excel Add-in via the new Microsoft Office Marketplace path described in the April 21 announcement. In our testing, the add-in loads inside Excel but consistently hangs at the authentication step; we have not been able to complete sign-in. As a result, we cannot reproduce any of the announced improvements. No other functionality changes were announced, so our original comparison still stands. We will update this post as the situation evolves.
On April 21, 2026, Databricks announced two installation paths that materially reduce administrative friction for its Excel Add-in — Microsoft Office Marketplace availability and Microsoft 365 admin center central deployment — along with a roadmap entry naming scheduled refresh and AI integrations as near-term items. It is a meaningful update to the product that entered Public Preview on March 2.
Over the same six-week window, Exponam added Snowflake and Microsoft Fabric to the supported platform list and made natural-language query with a private or bring-your-own LLM generally available, alongside a full direct-SQL scratchpad. What was on our March roadmap is shipping now.
The two announcements deserve an honest side-by-side read. Here it is.
What Databricks changed on April 21
Three specific improvements. Each is real, and we credit them plainly.
Simpler admin-driven installation. The sideload-and-Trust-Center install path that defined the March 2 Public Preview is no longer the only option. In tenants that permit Microsoft Office Marketplace access, installation is now self-service; in M365-admin-managed environments, IT can push the add-in centrally. Compared with the Public Preview experience, the friction is meaningfully lower — though tenants with Marketplace restrictions still face the older sideload path. (*see updated notes at top)
Unity Catalog metric views as the semantic story. The April 21 announcement positioned metric views as the centerpiece of the add-in’s value — define business semantics once in Unity Catalog, consume them consistently in Excel. Metric views are not a competitive advantage for either product: Exponam accesses them natively through the SQL-endpoint path with semantic definitions intact, so the end-user benefit is the same regardless of which add-in an enterprise uses.
Scheduled refresh and AI integrations on the roadmap. Both are now named as near-term items. Neither is shipping today. The AI roadmap line does not publish specifics — no model, no timeline, and notably no position on whether the model runs inside or outside the customer’s security perimeter. The reasonable expectation is that any Databricks-delivered AI query capability will be their Genie offering operating on Databricks infrastructure, which is an architecturally different proposition from a private-LLM option.
What Exponam shipped in April
Three things, in the order they matter.
Multi-cloud: Snowflake and Microsoft Fabric. Native Snowflake connectivity ships in April; Microsoft Fabric in May. Each uses the source platform’s existing security model — Snowflake Horizon and Fabric OneSecurity respectively — with no parallel permission layer. The same ribbon, the same task-pane workflow, the same governance posture as the Databricks path. AWS Redshift, Google BigQuery, and the AtScale semantic layer are in active development. This is the position no cloud-platform vendor can occupy: Databricks will not build a governed connector for Snowflake data, and Snowflake will not build one for Databricks data. The neutral cross-platform layer is structurally available to an independent vendor only.
Natural-language query and direct SQL, with a private or bring-your-own LLM. An analyst can now ask a question in plain English and receive governed, reproducible results — or drop into a full SQL scratchpad with individual and shared workspaces for analysts who prefer to write query directly. Both paths execute against the source platform’s native SQL engine with full governance. The AI layer runs on one of two customer-chosen models: a fully local, SQL-specialized LLM on the analyst’s machine — no tokens, no schema, no row-level data leaving the security perimeter — or the customer’s existing commercial provider under a bring-your-own-model arrangement, in which case Exponam passes only schema metadata. SQL-specialized 8B-parameter models run efficiently on standard corporate laptops; 14B and 32B models deliver near-zero latency on GPU-equipped machines.
This is the dimension where the gap to the Databricks add-in is most pronounced. The Databricks April 21 roadmap line names “AI integrations” without specifying deployment model or data handling. The reasonable expectation — given Databricks’ product architecture — is that any such capability will be Genie-native, running on Databricks infrastructure against Databricks data, with no private-deployment or BYO-model option. For financial services, healthcare, Big Four accounting, pharmaceutical, and government customers whose compliance policies prohibit sending proprietary data to external AI endpoints, a Databricks-hosted Genie integration is not a substitute for an on-device private LLM.
Product rename: Exponam Analyst Intelligence. Exponam.Connect is now Exponam Analyst Intelligence. The rename reflects what the product has grown into: the governed, AI-powered, multi-cloud analyst access layer for Excel.
What actually changed in the comparison
The March 4 post covered installation friction, data retrieval paths, performance, cost, SQL parity, and platform scope. Three of those lines need an update. The rest stand.
| Dimension | Status on March 4 | Status on April 22 |
|---|---|---|
| Installation | Databricks required XML manifest sideload, folder share, and Trust Center configuration. Exponam single-installer. | Databricks now asserts a simplified mechanism via Microsoft Office Marketplace and M365 admin center central deployment (subject to tenant permission). Exponam single-installer, with custom MSI for SCCM/Intune on enterprise license. The gap narrows; it does not close for managed tenants that restrict Marketplace access. (*See update notes at top- as of April 25, we have been unsuccessful utilizing the Microsoft Marketplace installation method.) |
| AI capability | Neither product had natural-language query. Exponam roadmap included it. | Exponam generally available with a private on-device LLM or bring-your-own commercial model. Databricks added AI to the roadmap without specifics; any offering is expected to be Genie-native on Databricks infrastructure. |
| Platform scope | Databricks only for both products, with Snowflake and Fabric on Exponam’s roadmap. | Exponam: Databricks production, Snowflake April, Fabric May. Databricks: Databricks only, structurally. |
The dimensions unchanged from March 4 remain unchanged. Delta Sharing as a zero-compute path on Databricks is still Exponam-only. DBU consumption on the Databricks add-in is still variable and still subject to formula-recalculation spend in shared workbooks. VSTO write performance on Windows is still an Exponam advantage. External user access without workspace accounts is still an Exponam advantage. The 10-million-row enterprise ceiling is still an Exponam advantage.
The single-sentence summary
The Databricks add-in is a first-party tool for Databricks-only workloads in small Databricks-native teams. Exponam Analyst Intelligence is the enterprise layer above the clouds — multi-cloud, AI-powered, governance-enforced, and cost-predictable at any scale.
Where to go from here
The updated April 2026 white paper — covering installation, architecture, performance, cost, governance, platform scope, AI capability, and usability in depth — is available below. For the one-page version, the April 2026 brief comparison covers the same ground in a Q&A format.
© 2026 Exponam, LLC · exponam.com · info@exponam.com · +1.646.360.0110
Exponam is a Databricks Validated Technology Partner
Leave a Reply