Friday, April 17, 2026

Ask EM - Chatbot interface: The Next Evolution of Intelligence in Oracle Enterprise Manager (OEM) 24ai



Ask EM is available now in Oracle Enterprise Manager 24ai, starting with Release Update 4. Whether you're a seasoned DBA or new to the platform, it's designed to make your day-to-day operations faster, more intuitive, and measurably smarter.


Example prompts: "How is the performance of my database?" or "How do I troubleshoot buffer busy wait?" — Ask EM handles both with the same natural-language interface.


Managing mission-critical databases and applications across hybrid environments has always demanded deep expertise and fast reflexes. Oracle Enterprise Manager (EM) has long been the backbone for those operations. Now, with the introduction of "Ask EM" in EM 24ai, that backbone just got a whole lot smarter.


Ask EM is a conversational chatbot powered by the Cohere Large Language Model (LLM) on Oracle Cloud Infrastructure. It lets you interact with your operational environment—and its documentation—using natural language, no SQL or navigation required.


What can Ask EM do?


Ask EM operates across two core domains, available as tabs in the chatbot interface:


Telemetry

Query the live health of your monitored targets, diagnose performance issues, and run root cause analyses—all in plain English. Results are rendered as interactive widgets you can pin to build dashboards.


Documentation

Ask any EM documentation question and receive a contextually rich answer using Retrieval Augmented Generation (RAG), complete with clickable links back to the source docs.


How it works under the hood:


When Ask EM launches from your on-premises EM 24ai console, it initializes a chatbot session and opens a secure WebSocket connection directly from the browser to Oracle's Ops Insights service in OCI. Critically, only the text of your question is transmitted—no EM data ever leaves your environment.


Architecture flow

Browser (EM 24ai)
Secure WebSocket
OCI Ops Insights
Oracle DB 23ai
Cohere LLM / GenAI
Widgets / Answer


The GenAI “Ask EM” feature is included in EM 24ai Release Update 4 (RU4). After applying EM 24.1 RU04 or a later patch, the “Ask EM” chatbot icon becomes available in the EM 24ai console, as shown below.


$OMS_HOME/OPatch/opatch lsinventory | grep -i "24.1"

/u01/app/oracle/oemp/middleware/oms_home/OPatch/opatch lsinventory|grep -i "24.1"












For telemetry questions, a similarity search runs against Oracle Database 23ai to surface the most relevant operational data. That data is passed to the Cohere LLM, which augments it into a tailored response and returns interactive widgets to the EM console.


For documentation questions, all public EM docs are pre-ingested into Oracle Database 23ai. RAG retrieval finds the most relevant passages and the GenAI Agent composes a natural-language answer with source links for deeper reading.


No firewall changes are needed on-premises. If your browser can reach the internet, Ask EM can connect to Oracle Cloud.


What you need to get started

OCI AccountSign up or use an existing account subscribed to the us-ashburn-1 region.
Ops InsightsAn active subscription with at least one host or database enabled.
IAM PoliciesTwo policies granting the user group access to manage OPSI GenAI sessions.
EM PatchEM 24.1 Release Update 4 (RU4) or newer applied to your EM environment.


Getting Started with Ask EM


Before you start exploring the power of Ask EM, a few prerequisites need to be in place to ensure a smooth setup and experience.


First, you’ll need an Oracle Cloud Infrastructure (OCI) account. You can either sign up for a new account or use an existing one. Make sure your OCI tenancy is subscribed to the Generative AI service in the US Ashburn region (us-chicago-1), as Ask EM currently relies on this region. Support for additional regions will be introduced in future updates.


Next, ensure you have an active subscription to OCI Ops Insights. At least one resource—such as a host or database—must be enabled within Ops Insights. This service, part of Oracle’s Observability and Management suite, plays a key role in powering Ask EM’s intelligent chatbot interactions.


You’ll also need to configure the appropriate Identity and Access Management (IAM) policies within your OCI tenancy. Add the following policies to grant the necessary permissions:


allow group <em-genai-user> to manage opsi-genai-em-session in tenancy

allow group <em-genai-user> to manage opsi-genai-em-docs-session in tenancy


Once these prerequisites are met, Ask EM requires a one-time initial setup. The first time you access the feature, the Super Administrator will be prompted to provide OCI credential details to establish the connection.


Powered by Generative AI


Ask EM brings the power of Generative AI directly into Oracle Enterprise Manager by leveraging the Cohere Large Language Model (LLM) running on OCI. This enables intelligent, context-aware interactions to help you manage and monitor your enterprise environment more efficiently.


To use Ask EM, ensure that the browser accessing EM 24ai has internet connectivity to communicate with OCI services. Importantly, the OMS host itself does not require internet access—simplifying deployment in more restricted environments.


Setting it up: three steps:


Once the patch is applied, clicking the Ask EM icon in the EM console launches a guided configuration wizard:


1. Review requirements

Confirm your OCI account, Ops Insights subscription, and IAM policies are all in place.


2. Configure credentials and region

Create or select an OCI named credential, choose your GenAI-supported region, and run a test connection. Optionally enable audit logging to track all Ask EM activity.


3. Set access controls (optional)

Super administrators have access by default. Extend access to additional admins by assigning them the EM_ASKEM_ADMIN role.




The architecture of Ask EM is designed to deliver powerful Generative AI capabilities while maintaining the security and integrity of your enterprise environment. The following diagram illustrates how Ask EM integrates Oracle Enterprise Manager with Oracle Cloud Infrastructure (OCI) services in a seamless and secure manner.


Breaking Down the Architecture


Let’s walk through what’s happening behind the scenes:


1. User Interaction (Enterprise Manager Console):


The journey begins in the EM 24ai console, where the user interacts with the Ask EM chatbot. Queries such as system health checks or performance insights are entered directly into the interface.


2. Secure Communication over the Internet:


The request is transmitted securely over HTTPS via the public internet. Importantly:


The communication originates from the user’s browser, not the OMS host

This design eliminates the need for outbound internet access from the Enterprise Manager infrastructure


3. OCI Ops Insights as the Gateway:


Once the request reaches OCI, it is handled by the Ops Insights Cloud Service, which acts as the orchestration layer. It:


Manages GenAI session requests

Connects enterprise telemetry data with AI processing

Ensures secure and policy-controlled access


4. Generative AI Processing (LLM + Agent)


The request is then processed within OCI’s Generative AI stack:

The GenAI Agent interprets the user’s query

The Cohere LLM generates intelligent, context-aware responses

Relevant enterprise context is fetched when needed


5. Data Sources and Context Enrichment


To provide meaningful insights, the system may leverage:


Oracle 23ai Database for structured data

Documentation and widget metadata for contextual assistance


This ensures responses are not just generic—but tailored to the Enterprise Manager environment.


6. Response Returned to User


Finally, the generated response flows back through the same secure channel and is displayed in the Ask EM chatbot interface—typically within seconds.


How “Ask EM” Works: Behind the Scenes


When Ask EM is launched from the on-premises EM 24ai console, a chatbot session is immediately initialized. At this point, a secure WebSocket connection is established directly from the user’s browser to Oracle’s Ops Insights service running on Oracle Cloud Infrastructure (OCI).


This session is assigned a unique identifier, which is used for all subsequent communication throughout the interaction. This ensures a continuous, secure, and seamless conversational experience between the EM console and OCI services.


Secure, Browser-Based Connectivity:


The connection is established over the public internet, but importantly:


No changes are required to on-premises firewall configurations

No additional ports need to be opened

EM 24ai does not need to operate in any special “online mode”


As long as the user’s browser has internet access, Ask EM can securely connect to OCI services.


A key security principle is maintained throughout the process:


Only the user’s actual question (for example, “How is the performance of my database?”) is transmitted to OCI. No additional EM telemetry or sensitive enterprise data is sent outside your environment.


What Happens in Oracle Cloud


Once a question reaches OCI, Ask EM follows two primary processing workflows depending on the nature of the query:


1. Telemetry-Based Insights Workflow


For operational and performance-related queries, Ask EM leverages enterprise telemetry data through the following steps:


A similarity search is performed using Oracle Database 23ai to identify the most relevant operational data

The selected data is passed to the Generative AI service and LLM for augmentation and contextual reasoning

The LLM enriches the response with domain-specific intelligence tailored to the user’s query

The GenAI service returns contextual widgets back to Enterprise Manager 


Within the EM interface:


Clicking the Findings button opens a dedicated side panel

Relevant widgets are rendered along with smart filters

Users can:

Interact with widgets dynamically

Modify filter values in real time

Pin important widgets for later reference


As users continue asking follow-up questions, new relevant widgets are generated while pinned widgets remain persistent. Over time, these pinned insights can be used to create customized dashboards—built directly from conversational exploration.


2. Documentation Intelligence Workflow


Ask EM also provides powerful AI-driven documentation assistance.


All public Oracle Enterprise Manager documentation is ingested and indexed in Oracle Database 23ai. When a user asks a documentation-related question—such as:


“How do I troubleshoot buffer busy waits?”


the system uses Retrieval-Augmented Generation (RAG) to:


Search across relevant documentation sources

Retrieve the most contextually accurate content

Generate a natural-language, easy-to-understand response


The chatbot also provides:


Clickable reference links

Direct navigation to official documentation in a new browser tab

Support for follow-up questions for deeper exploration


This ensures users get not just answers, but verifiable and traceable guidance from official sources.


Key Takeaways:


  • Secure WebSocket connection is established directly from the browser to OCI
  • No inbound or outbound changes are required on the EM infrastructure
  • Only user queries are transmitted—no additional EM data is shared
  • Two core workflows power Ask EM:
  • Telemetry insights via GenAI + Ops Insights + 23ai
  • Documentation Q&A via RAG over indexed EM docs
  • Interactive widgets and dashboards transform insights into actionable monitoring views


Ask EM Setup Wizard:


Step 1: Review the required OCI account, Ops Insights subscription, and IAM policies requirements.




Step 2:  Create a new OCI named credential or select an existing OCI credential. If you are unsure where to locate these values, helpful links are provided for each field to guide you through the process. Choose your GenAI-supported region, then click Test Connection and Save to verify. Optionally, enable Audit to track all “Ask EM” activities.



Step 3 (Optional):  By default, super administrators have access to use “Ask EM”. To extend access to additional administrators, simply click on ‘Select Users’. The chosen administrators will be assigned the ‘EM_ASKEM_ADMIN’ role, allowing them to access “Ask EM”.




Finalizing the “Ask EM” Setup


Once you have completed the three key steps—subscribing to Oracle Cloud and OCI Ops Insights, configuring credentials and selecting the appropriate region, and setting up the required access control—your Ask EM interface is fully ready for use.


At this stage, Ask EM is successfully integrated into your Oracle Enterprise Manager environment, enabling a seamless GenAI-powered experience.







The benefits in practice:


  • Faster issue resolution—diagnose problems through conversation instead of navigating menus
  • Reduced reliance on support teams for common operational and documentation queries
  • Data-driven decisions backed by live telemetry surfaced as interactive, pinnable widgets
  • Full auditability with built-in activity logs for every Ask EM interaction
  • Privacy-first design—only your typed question is ever sent to Oracle Cloud












Tuesday, March 31, 2026

Oracle 26ai ATPS — When Your Database Starts Thinking for You

Oracle 26ai ATPS — When Your Database Starts Thinking for You:


Imagine a database that provisions itself, patches itself, scales itself, and now — agents itself. Oracle's Autonomous AI Transaction Processing (ATPS), running on the new 26ai engine, is no longer just a cloud database. It is the intelligent core of enterprise AI.

For over two decades, database administrators spent considerable energy on provisioning, tuning, patching, and scaling — the repetitive "plumbing" work that consumed talent better spent on innovation. Oracle's Autonomous Database disrupted that model. Today, with Oracle AI Database 26ai, the disruption reaches further: AI agents are first-class citizens inside the database engine itself.

This blog explores what Oracle Autonomous AI Transaction Processing (ATPS) is, how it has evolved through 26ai, what new capabilities matter most to enterprise teams, and why this architecture shift is significant for organizations running critical OLTP workloads — including government systems, financial platforms, and ERP environments.

Highest Gartner OLTP score in a row
96%
Of Fortune 100 on Oracle DB
99.995%
SLA with Active Data Guard
$0
Extra charge for AI Vector Search

What is Oracle Autonomous AI Transaction Processing?

Oracle Autonomous AI Transaction Processing (ATPS) is a fully managed cloud database service optimized for high-volume, high-concurrency transactional workloads — the kind that power e-commerce checkouts, banking ledgers, ERP systems like Oracle EBS, and government financial platforms.

Unlike traditional managed databases that still require significant human intervention for performance tuning, patching, and capacity planning, ATPS automates all of that using machine learning built into the database kernel. The service runs on Oracle's Exadata hardware, which provides the I/O and compute density needed for demanding OLTP at scale.

ATPS — Autonomous Self-Driving Pillars
ORACLE EXADATA INFRASTRUCTUREPROVISIONDeploy inminutesNo sizingrequiredAuto-configon launchTUNEIndex &SQL planStats &memoryauto-tunedcontinuouslySCALECompute &storage growZero-downtimeelastic scalePay-per-useOCPUsSECUREAuto-patchzero-downtimeData Safe &maskingTDE always-onencryptionBACKUPWeekly full+ daily incr.60-day PITRto ObjectStorageinstantlyREPAIRPredicts &self-healsHW + SWfailuresContinuousmonitoringOracle Autonomous AI Transaction Processing

ATPS stores data in row format — optimal for transactional access patterns where all columns of a record must be retrieved or updated together. It supports the full spectrum of data types: relational tables, JSON documents, graph data, XML, spatial objects — all within a single converged engine, eliminating the fragmentation cost of running multiple purpose-built datastores.


The 26ai Engine — What Changed and Why It Matters

Oracle AI Database 26ai, released in October 2025 as a long-term support release, replaces Oracle Database 23ai. The branding shift from a version number to a year-based release name ("26ai") signals Oracle's intent: this is no longer a product with version increments — it is a platform with a living AI architecture baked in at every layer.

The transition from 23ai to 26ai requires no full database upgrade and no application re-certification. Organizations simply apply the October 2025 Release Update (RU) and gain access to 26ai's capabilities as they become available — a deliberate low-friction migration path designed for enterprises running regulated or critical workloads.

Oracle Database 23c — 2023

Introduced AI Vector Search, JSON Relational Duality. The first step toward AI-native capabilities.

Rebranded to 23ai — 2024

Database identity aligned with AI-first strategy. Vector Search and Natural Language SQL queries via Select AI arrived.

Oracle AI Database 26ai — October 2025

Long-term support release. AI architected into the full data and development stack. Agentic AI, Unified Hybrid Vector Search, Autonomous AI Lakehouse, NVIDIA NIM integration, Private Agent Factory.

26ai — March 2026 updates

Oracle Unified Memory Core, Deep Data Security, Autonomous AI Vector Database Limited Availability, Agent ROI Dashboard. Private Agent Factory with pre-built agents for enterprise workflows.

"By architecting AI and data together, Oracle AI Database makes 'AI for Data' simple to learn and simple to use."

— Juan Loaiza, EVP Oracle Database Technologies

Key Capabilities of Oracle 26ai ATPS

Agentic AI — Agents as First-Class Citizens

The headline capability of 26ai is the elevation of AI agents to first-class citizens inside the database engine. Two primary mechanisms enable this:

Agentic AI Architecture Inside Oracle 26ai ATPS
Oracle Autonomous AI Transaction Processing (26ai)Select AI AgentDefine · Run · Governagents in-databaseIn-database tools (built-in)External tools via REST APIMCP Server integrationMulti-step agenticworkflow automationPrivate Agent FactoryNo-code builder &deployment platformVisual no-code canvasPre-built enterprise agentsPrivate container deployData never leavesyour infrastructureUnified Memory CoreStateful · PersistentAgent memory inside DBVector + Relational +JSON + Graph + SpatialSingle ACID transactionONNX + NVIDIA NIMOCI GenAI / OpenAIAll three layers share Exadata compute · 99.995% SLA · Deep Data Security (row & column level)

Select AI Agent provides a developer-facing framework to define, run, and govern AI agents directly inside the Autonomous AI Database. It supports in-database tools, external REST integrations, and MCP (Model Context Protocol) servers — enabling iterative, multi-step agentic workflows where the agent can explore different solution paths and fetch additional context during analysis.

Private Agent Factory targets enterprise teams who need agentic capability without writing code. It provides a visual, no-code canvas to design agent workflows, connect to data sources (Oracle databases, SharePoint, REST APIs, object storage), and deploy agents as portable containers in environments the customer controls — so sensitive data never leaves their infrastructure.

Unified Hybrid Vector Search

26ai introduces Unified Hybrid Vector Search, which blends AI vector embeddings with relational predicates, full-text search, JSON filters, graph traversals, and spatial queries — all in a single SQL statement. This means an agent or application can retrieve related documents, images, audio, video, and structured rows together without any cross-system stitching. Vector Search capabilities are included at no additional charge.

Practical example for EBS / UFMS environments: A compliance agent can simultaneously query structured transactional records, PDF policy documents via vector search, and graph-linked entity relationships — all in one query, inside the database, under full ACID guarantees.

Autonomous AI Lakehouse

26ai introduces the Autonomous AI Lakehouse, which supports Apache Iceberg open table format. This allows ATPS to read and write Iceberg-format data in OCI Object Storage, AWS S3, Azure Blob, and Google Cloud Storage — and remain interoperable with Databricks, Snowflake, and other Iceberg-compliant systems on the same clouds. Enterprise data lake workloads gain Exadata-powered performance and serverless scaling without data movement.

Deep Data Security

Announced in March 2026, Deep Data Security introduces context-aware data access control specifically designed for agentic AI. AI agents operating on enterprise data must be governed at the row and column level — not just the table level. Deep Data Security enforces these controls natively in the database engine, so agents cannot inadvertently expose sensitive data even when executing complex multi-step retrieval chains.


Multicloud Reach — ATPS Everywhere

One of Oracle's most significant strategic moves with 26ai is making ATPS available across all four major hyperscalers. Organizations are no longer required to migrate their entire workload to OCI to benefit from Oracle's autonomous database capabilities.

ATPS Deployment Models — Where You Can Run It
OCINative homeFull feature setExascale optionAWS+ SageMakerintegrationS3 lakehouseAzureInterconnectBlob lakehouseATP-Azure IOGoogle CloudGCS lakehouseIceberg tablesIceberg compat.🏢On-PremExadata @CustomerDedicated Reg.

For federal and regulated-industry customers running HHS or EBS workloads, this multicloud reach is particularly significant: it allows Autonomous AI Database to operate on dedicated regions or Exadata Cloud@Customer on-premises deployments, maintaining full data sovereignty and compliance posture while still gaining 26ai's AI-native capabilities.


23ai vs 26ai — What Actually Upgraded

CapabilityOracle DB 23aiOracle AI DB 26ai
AI Vector Search Available + Hybrid (vector + SQL + graph + JSON)
Natural Language SQL Select AI (basic) Select AI Agent (multi-step agentic)
AI Agent support Not available First-class citizen in DB engine
No-code agent builder Not available Private Agent Factory
Persistent agent memory Not available Oracle Unified Memory Core (Mar 2026)
Lakehouse / Iceberg Not available Autonomous AI Lakehouse (all 4 clouds)
NVIDIA GPU acceleration Not available NIM containers, CAGRA/cuVS planned
Row/column-level AI securityTable-level Deep Data Security (context-aware)
Upgrade path Apply Oct 2025 RU — no upgrade needed
Self-driving (provision/tune/scale/patch/backup) Available Continued + AI-enhanced management

Why This Matters for Enterprise OLTP Teams

🏛

Federal / Regulated Workloads

ATPS 26ai automates patching at zero downtime, critical for systems requiring 99.95%+ uptime SLAs and audit trails like HHS UFMS and FBIS.

📊

Oracle EBS on OCI

EBS R12.2 workloads running on ATPS benefit from ML-driven SQL plan tuning, auto-index management, and Data Safe integration for separation of duties.

🤖

AI-Augmented Operations

DBAs can deploy in-database agents for ETL validation, anomaly detection, and compliance checks — dramatically reducing manual triage cycles.

💡

Cost Reduction

Eliminating tuning overhead, manual patching, and multi-database fragmentation significantly reduces TCO. AI Vector Search is included at no extra charge.

🔒

Zero-Trust Security

Auto-patching, TDE encryption-always, Data Safe masking, and the new Deep Data Security (row/column level for agents) align with modern zero-trust frameworks.

APEX + Low-Code Apps

APEX AI Application Generator uses natural language to build trusted apps against live ATPS data — cutting application development cycles from weeks to hours.


Getting Started with ATPS 26ai

Oracle has made multiple free entry points available for evaluation and development:

Always Free Autonomous AI Database — available on OCI with no time limit, suitable for development and proof-of-concept workloads. Visit FreeSQL.com for instant browser-based access to Oracle AI Database 26ai without any provisioning.

For existing Oracle Database 23ai customers, the upgrade path is frictionless: apply the October 2025 Quarterly Release Update. No application re-certification is required. Existing SQL, PL/SQL, and application code continue to run as-is while the 26ai capabilities become progressively available.

For teams running Oracle EBS R12.2 on OCI, the Autonomous Transaction Processing service on Dedicated Exadata Infrastructure provides the highest performance tier — with full ATPS autonomy layered on the Exadata hardware that EBS is already optimized for. This combination is particularly compelling for Wave 8 and post-wave deployments targeting long-term operational sustainability.

"Oracle's rationale reflects its broader strategy of making the database a central pillar of enterprise AI, given that execution ultimately depends on where the data resides."

— Noel Yuhanna, Principal Analyst, Forrester Research (2026)

Oracle AI Database 26ai and its Autonomous AI Transaction Processing service represent a fundamental architectural shift: the database is no longer just a storage and retrieval system. With 26ai, it is the trusted, ACID-governed, AI-native core of enterprise automation — where agents live, reason, and act on the data that matters most.

For enterprise teams managing regulated OLTP environments, the case is straightforward: the operational overhead reduction is measurable, the AI capabilities are included, and the upgrade path requires a single patch. The question is no longer whether to modernize — it is how fast to capture the advantage.

Ask EM - Chatbot interface: The Next Evolution of Intelligence in Oracle Enterprise Manager (OEM) 24ai

Ask EM is available now in Oracle Enterprise Manager 24ai, starting with Release Update 4. Whether you're a seasoned DBA or new to the p...