Database performance management has long been the domain of specialists fluent in wait events, snapshot comparisons, and execution plans. Oracle is now betting that AI can widen that circle without sacrificing rigor. In a blog published August 28, Yutaka Takatsu, product manager for Enterprise Manager and Database Server manageability at Oracle, laid out how Oracle Enterprise Manager is evolving into an AI-powered operational intelligence platform.
The problem statement is familiar to anyone running enterprise infrastructure. Modern database estates span Exadata fleets, RAC clusters, Autonomous Databases, multicloud deployments, and Kubernetes platforms, and the volume of operational telemetry keeps growing. As Takatsu puts it, modern database operations are becoming too complex for traditional monitoring alone.
Deterministic Analysis First, AI Explanation Second
What distinguishes Oracle’s approach is its architecture. Rather than pointing a large language model at raw performance data and hoping for accurate answers, Oracle pairs conversational AI with its proven deterministic analysis engines. The planned AWR.ai capability lets administrators ask natural-language questions such as “What changed during the last 30 minutes?” or “Which SQL statements regressed?” Behind the scenes, each request maps to deterministic Automatic Workload Repository analysis rather than relying solely on an LLM to interpret performance data, an approach designed to produce accurate, repeatable, and trustworthy results.
The planned SQLPerf.ai follows the same tuning logic. It begins with deterministic analysis of SQL Monitor execution evidence to find where execution time is spent, then uses AI to deliver a natural-language explanation and remediation guidance for issues such as inefficient nested loops, stale statistics, or missing partition pruning, with planned integration into SQL Tuning Advisor. Both features are planned enhancements covered by Oracle’s Safe Harbor statement, so timing and functionality remain at Oracle’s discretion.
Already shipping are Oracle AI Database Assistant, available from Enterprise Manager 24ai RU10, which lets administrators query Enterprise Manager telemetry conversationally with interactive widgets and zero data exfiltration, and MCP Server support in RU12, which securely exposes operational context to MCP-compatible AI clients. At the same time, Enterprise Manager remains the authoritative platform for authentication, authorization, and auditing.
The Pattern Behind the Product
For the broader industry, the significance is the pattern, not any single feature. Oracle is applying the same governed-AI philosophy to database operations as it is across its application portfolio this year. ERP Today’s coverage of the AI-native builder for Fusion Agentic Applications noted that Oracle insists AI must inherit existing security, governance, and auditability rather than bolting controls on later. Similarly, the NetSuite Next rollout put a conversational assistant at the center of ERP work while respecting existing roles and permissions. Enterprise Manager now extends that doctrine to the infrastructure layer that ERP systems depend on.
That matters because ERP performance problems ultimately surface as database problems. A slow month-end close, a stalled MRP run, or a sluggish reporting cycle often traces back to SQL regressions or workload contention that only a small pool of DBAs can diagnose. If Takatsu’s team delivers on the vision, that diagnostic capability becomes conversational, faster, and accessible to broader operations teams without leaving the governance perimeter.
What This Means for ERP Insiders
Hybrid AI architecture is the credibility benchmark. Oracle’s pairing of deterministic analysis with AI explanation is a template for evaluating any AI operations tool. Teams should ask vendors what grounds the answer, not just what generates the prose.
Governed context is becoming the differentiator at every layer. From Fusion to NetSuite to Enterprise Manager, Oracle’s consistent argument is that AI is only as trustworthy as the security, credentials, and audit controls it inherits. ERP architects should apply that same standard to infrastructure AI purchases.
DBA skills pressure gets partial relief, not replacement. Conversational AWR and SQL analysis can lower the expertise barrier for routine diagnosis, but planned features remain roadmap items under Safe Harbor. Operations leaders should pilot the shipping AI Database Assistant now and treat AWR.ai and SQLPerf.ai as direction, not commitment.
