Over the past few months, we’ve seen every major data platform invest in semantic context.
Semantic layers, governed metrics, and shared business definitions are quickly becoming the key infrastructure piece that determines whether AI is trusted.
Inside this issue, we look more into why semantic layers are emerging as critical AI infrastructure, architecture decisions that influence AI ROI, the latest platform updates, and how one client modernized their data foundation to enable trusted AI.
We’re also introducing Databricks, Applied, a new section focused on getting the most from the platform that does the most.
Below you'll find:
🔥 Hot Topic: Semantic layers are having a moment
⚙️ Databricks, Applied: The early decisions that determine ROI
🛠️ Other Tech Insights: Genie in Teams, Spotter-first ThoughtSpot, dbt 2.0, and Power BI web mirroring
📋 From the Field: The modern data foundation that unlocked AI at Whetstone
As agents become responsible for more business tasks, they'll increasingly rely on these shared definitions to understand what “revenue,” “customer,” or “margin” actually mean before they act.
These are the factors that influence whether AI becomes a competitive advantage — or an expensive experiment:
Clear ownership
Governed data
Shared business definitions
Trust
Performance standards
Our latest guide shares five implementation practices we've seen consistently improve ROI across Databricks environments, with lessons that apply to any organization building a modern data platform.
Top updates for Databricks, ThoughtSpot, dbt, and Power BI
Databricks brings Genie into Microsoft Teams
The Databricks Genie app is now available in Microsoft Teams in Public Preview. Users can ask data questions through direct messages, group chats, or channels, while channel owners can connect a specific Genie Agent to provide answers for a defined use case.
The practical implication: Before enabling Genie in Teams, organizations should define guardrails: which agents users can access, who can see responses, and how users will verify answers before acting on them.
ThoughtSpot previews a Spotter-first home page
ThoughtSpot is introducing a focused home page that places Spotter at the center of the user experience. The new option will enter beta in the next ThoughtSpot Cloud release, and administrators will be able to manage it across the organization or for individual Orgs.
The practical implication: ThoughtSpot is positioning conversational analytics as the primary way users interact with data. Teams should test the new experience with users and address unclear terminology, incomplete metadata, and weak content governance before making it the default.
dbt modernizes the transformation layer for speed and AI
dbt Core v2.0 will move onto the faster, Rust-based engine developed for Fusion, bringing Core and Fusion onto one shared foundation. dbt State will reduce unnecessary processing by skipping or reusing models when code and upstream data have not changed, while dbt Wizard introduces an AI agent grounded in project lineage, tests, contracts, and metric definitions.
The practical implication: These updates could reduce development time and warehouse compute while making AI-assisted analytics engineering more reliable.
Power BI brings TMDL View to the web
Microsoft has added TMDL View to web modeling in Power BI Service. Developers can now inspect, script, and modify semantic model metadata, including tables, measures, and relationships, without downloading model files or switching to Power BI Desktop. The web editor also supports bulk changes and reusable model definitions.
The practical implication: More semantic modeling work can move into browser-based workflows, making repetitive changes easier to automate and apply consistently. Teams should review who has permission to modify models, since code-level access makes it possible to change many objects at once.
From the Field
From fragmented systems to a modern data foundation that supports AI initiatives
After a merger left Whetstone operating across four ERP environments, they were spending 15 hours each week locating data and manually building reports. To modernize their environment, they centralized on Databricks, built a shared semantic model, and utilized Power BI for reporting.
The result: they eliminated ~1,000 hours of manual work and got a single trusted source of sales, operations inventory, and weekly KPIs.
