In Short: What Is Rayfin?
Rayfin is Microsoft's AI-powered intelligence capability built natively into Microsoft Fabric, announced at Microsoft Build 2025. It provides natural language querying over your Fabric semantic models and OneLake data, proactive insight generation and anomaly detection, and the intelligence foundation for building autonomous Fabric Data Agents.
For organisations running Microsoft Fabric, Rayfin is the activation layer - the capability that makes your existing, well-governed data estate accessible to every user in the organisation, not just the analysts and engineers who know how to write DAX or SQL.
The Problem Rayfin Solves
Well-implemented Fabric estates consistently run into the same structural problem: data literacy is unevenly distributed. Data engineers, BI developers, and analysts work directly with the data. Business users - the majority of the organisation - access it through dashboards and reports that someone else built for them.
This creates a bottleneck. Ad-hoc questions that do not map to an existing report require an analyst to run a manual query. Decisions are made on approximate or stale information when the exact answer is sitting in the data, accessible to those with the skills to retrieve it.
Rayfin eliminates this bottleneck by giving every user a governed, AI-mediated interface to the data - in natural language, through Microsoft Teams, Outlook, and the Fabric interface itself.
What Rayfin Does
Natural Language Data Querying
Users type questions in plain business language - "What were our top five products by revenue last quarter?" or "Show me customer churn by region for the last six months" - and Rayfin translates those questions into precise DAX or SQL queries against your published semantic models and Gold-layer Lakehouse tables.
The accuracy comes from where Rayfin looks: not at raw table schemas, but at your defined semantic model - the measures, hierarchies, and relationships your BI team has already built and governed. Answers are consistent with what Power BI reports would show because they are drawn from the same logic.
Proactive Insights and Anomaly Detection
Rayfin monitors patterns in your Fabric data continuously and surfaces anomalies, threshold breaches, and trend changes without users needing to know to look for them. These insights appear through Microsoft 365 Copilot in Teams and Outlook, putting data intelligence where business decisions are actually made.
Fabric Data Agents
Rayfin provides the intelligence substrate for Fabric Data Agents - autonomous AI workflows that query Fabric data, reason over results, and trigger downstream actions through Power Automate, APIs, or webhooks. A Data Agent can detect an operational anomaly, cross-reference it against financial and supply chain data, determine severity, and notify the right team or trigger a corrective workflow - all without human initiation.
Reports show you what happened. Data Agents built on Rayfin detect, reason, and respond.
How Rayfin Integrates with Your Existing Fabric Investment
Rayfin layers on top of what you have already built rather than replacing it:
- Semantic models: Rayfin reads your published Power BI semantic models. The governed business logic your BI team has already defined becomes the layer Rayfin queries against. No retraining required.
- OneLake tables: Gold-layer Lakehouse tables can be queried directly for datasets not yet covered by a semantic model.
- Microsoft Purview governance: Row-level security, column-level security, and sensitivity labels apply to all Rayfin interactions. Users see only data they are authorised to access.
- Microsoft 365 Copilot: Rayfin surfaces through Copilot in Teams and Outlook, placing data intelligence where business users spend most of their time.
What Rayfin Means for Your Teams
Business Users
The immediate impact is direct access to governed data through plain language questions. Sales managers, finance leads, and operations directors can ask the questions they need answered without waiting for a report to be built or a query to be run. This is what AI-powered self-service analytics actually looks like in practice.
BI and Data Teams
Ad-hoc query requests decrease substantially, freeing BI developers and analysts to focus on building new capability rather than answering one-off questions. Rayfin also creates a strong incentive to maintain clean, well-structured semantic models - since answer quality is directly proportional to model quality.
Data Governance Teams
All Rayfin interactions operate within Fabric's existing Microsoft Purview governance framework. Enabling Rayfin does not introduce new data access vectors outside the established security perimeter. All interactions are logged and auditable.
What You Need to Get Value from Rayfin
Rayfin's quality is directly proportional to what sits beneath it:
- Active Fabric F-SKU capacity
- Well-structured semantic models with measures named in business language and relationships properly defined
- Clean Gold-layer Lakehouse tables containing business-ready data
- Microsoft 365 Copilot licences for Teams and Outlook surface access
Organisations with mature Fabric estates can activate Rayfin and see value immediately. Organisations with poorly governed implementations will find Rayfin amplifies their data quality problems rather than solving them. In this sense, Rayfin is a forcing function: it rewards good governance and punishes bad governance in ways that static dashboards never did.
Rayfin in the Broader Microsoft AI Stack
Rayfin sits at a specific layer in Microsoft's emerging enterprise AI architecture: OneLake provides unified storage → semantic models provide governed business logic → Rayfin provides natural language access and Data Agent foundations → Microsoft Foundry provides multi-agent orchestration → Microsoft Copilot provides the end-user interface across Microsoft 365.
Every layer depends on the one beneath it. The organisations that deploy Rayfin and the full AI stack most effectively are the ones with solid, well-governed Fabric foundations already in place.
If you are building or scaling a Fabric implementation and want to ensure it is structured to support Rayfin and Data Agents, our Microsoft Fabric team works with organisations across the UK, US, and South Africa to deliver governance-first implementations that AI capabilities can operate over.
