What is Oracle Fusion Data Intelligence and what does it include?
Oracle Fusion Data Intelligence is Oracle's prebuilt analytics offering for Oracle Fusion Cloud Applications: a managed data pipeline, a data model, a semantic model and packaged content, running on Oracle Analytics Cloud and Oracle Autonomous AI Lakehouse. Oracle describes it as a suite of insight applications with embedded AI for finance, people, supply chain and revenue leaders, delivered as ERP, HCM, SCM and CX analytics. Oracle manages the prebuilt layer, from the Fusion pipeline, data model and semantic model to the key metrics, workbooks and reports, which is why teams choose it over building a Fusion warehouse themselves.
The extensible side is yours. You can load external data sources into the same lakehouse with Oracle Data Integrator or any ETL tool, connect to data through more than 30 self-service connectors instead of loading it, import spreadsheets, extend the semantic model and create your own key metrics, dashboards and reports. One naming note: Oracle renamed Fusion Analytics Warehouse to Fusion Data Intelligence, so older documentation links still carry the earlier name, and current guides say Oracle Autonomous AI Lakehouse, Oracle's next generation of Autonomous Data Warehouse.
Which AI features does Oracle document for Fusion Data Intelligence?
Oracle documents prebuilt machine learning, embedded AI inside each analytics application, and the Oracle Analytics AI Assistant for Fusion Data Intelligence. The product page says prebuilt ML models and embedded AI surface anomalies, predict outcomes and guide decisions across Fusion data without building models or writing queries. The ERP application is described as applying that to cash flow, transactions, spend and profitability. All of it is scoped to the Fusion data model Oracle maintains.
The AI Assistant is the conversational piece. Oracle announced the Oracle Analytics AI Assistant for Fusion Data Intelligence on October 14, 2025, describing it as a way to engage in a dialog with business data rather than being limited to what a dashboard shows. In the documentation, an administrator enables AI Assistant in Workbooks under Generative AI in Oracle Analytics Cloud, chooses which datasets or subject areas consumers may query, and grants the Use Assistant in Workbooks permission. The same announcement covered an Oracle Cloud EPM integration that brings planning, financial close and consolidation data into Fusion Data Intelligence.
Related: See Oracle Fusion AI analytics and data intelligence
How does Fusion Data Intelligence enforce data security and roles?
It enforces access through security assignments that the semantic layer applies as data filters at query time. Oracle's documentation explains that Fusion Data Intelligence maps a security context to a data role, and that users receive data roles through their assigned groups. When a user queries, the semantic layer filters rows to the security assignment values that user holds. A user with a job role but no assignment values sees no datasets for that job role, which is the right default for financial and workforce information.
Any AI access to Fusion data should inherit this model rather than bypass it. That holds for Oracle's own AI Assistant, which is limited to the datasets and subject areas an administrator exposes, and it holds for an external assistant such as Claude. If a connector reaches the warehouse with a service account that has broader rights than the person asking, the security assignments stop meaning anything. The person's identity, roles and assignment values have to travel with the question.
What questions does Fusion Data Intelligence leave open for finance and operations leaders?
It leaves open cross-system answers, access from the assistant your people already use, and one agreed definition for each metric across systems. Fusion Data Intelligence is purpose-built for Fusion data. You can add external sources to its lakehouse, but that pipeline is yours to build, the loaded data arrives without Oracle's prebuilt metrics, and you define and test its security yourself. A CFO who wants Salesforce bookings reconciled against invoiced revenue in Fusion, by region and quarter, still needs someone to build and govern that join.
Conversational access is the next gap. The Oracle Analytics AI Assistant lives inside Oracle Analytics Cloud workbooks. If your finance and operations teams ask questions from Claude, ChatGPT or Copilot during the day, those assistants cannot see Fusion metrics unless something governed sits in between. Definitions are the gap that outlasts any tool. Fusion holds one meaning of gross margin, the CRM holds pipeline stages, a NetSuite subsidiary holds its own chart of accounts, and nobody has written down which definition wins when a leader asks a question that spans all of them.
How does an MCP-connected governed semantic layer complement Fusion Data Intelligence?
It adds the layer Fusion Data Intelligence does not try to be: certified business definitions applied across Fusion and your other systems, exposed to any assistant through the Model Context Protocol. MCP uses a client-host-server design in which a host such as Claude runs clients that each talk to one server; the server exposes tools and resources, and the host enforces consent and security boundaries. A governed semantic layer fits that role well. It answers in business terms and keeps credentials and query logic on its own side of the boundary.
Nexatron, which dotSolved built for this job, has a live Oracle Fusion connector alongside live Salesforce and NetSuite connectors, and exposes governed query tools to any MCP client, including Claude, ChatGPT and Copilot. Every answer carries the query that ran, the certified metric definition and the row-level policies applied, so a controller can check a Salesforce-to-Fusion reconciliation rather than trust it. Oracle Database and Oracle EPM connectors are listed as coming soon, not live, so scope a pilot to Fusion first, with Fusion security assignments as the source of truth for who sees which rows. Write down the ten cross-system questions your leaders ask most, with an agreed definition for each. That list is a pilot scope. Book a demo with dotSolved and bring it.