# Oracle Digital Assistant, OCI AI and JD Edwards Orchestrator

> Source: https://netray.co/stack/oracle-digital-assistant-oci-ai-jd-edwards-orchestrator
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: ERP Stack

## Short answer

Oracle describes giving JD Edwards users conversational access to data and workflows, document understanding and generative AI use cases such as supplier spend and quality inspection. The mechanism is JDE Orchestrator calling OCI AI services over REST, and it needs a current Tools Release. The OCI services run in Oracle's cloud, which matters for air-gapped sites.

## Erp

JD Edwards EnterpriseOne

## Erp Key

jd-edwards

## On Prem

true

## Product

Oracle Digital Assistant and OCI AI services via JDE Orchestrator

## Vendor

Oracle

## Category

AI assistants and copilots

## Integration

**Method:** JDE Orchestrator calling OCI REST APIs

**Detail:** Orchestrator can call external REST services, so an orchestration invokes an OCI AI endpoint and returns the result to a JDE flow or to a conversational front end. Oracle's white paper states a current Tools Release is required, so check your release first.

**Method:** Orchestrator as the JDE-facing API

**Detail:** In the other direction, Orchestrator publishes JDE business functions and data as REST services that a digital assistant can call. Security follows the JDE user and role used by the orchestration.

## Data Flows

**From:** User

**To:** Digital assistant

**Data:** Natural-language question or request

**Trigger:** User message in the chat channel

**From:** Digital assistant

**To:** JDE Orchestrator

**Data:** Structured request such as a supplier or order lookup

**Trigger:** Intent recognized by the assistant

**From:** JDE Orchestrator

**To:** OCI AI services

**Data:** Text or document content for summarization or extraction

**Trigger:** Orchestration step that calls the OCI REST endpoint

**From:** OCI AI services

**To:** JDE

**Data:** Extracted fields, summaries or classifications

**Trigger:** Response returned to the orchestration

## Setup

### Check Tools Release and AIS

**Detail:** Confirm the JDE Tools Release meets Oracle's requirement and that the AIS server hosting Orchestrator is available and secured.

### Pick one use case

**Detail:** Start with a single scenario, such as supplier spend questions or document extraction, instead of a general assistant.

### Build the orchestrations

**Detail:** Create orchestrations in Orchestrator Studio for the data the assistant needs and for the OCI REST calls.

### Configure OCI access

**Detail:** Set up OCI credentials, network access from the AIS server and the AI service endpoints, with least-privilege policies.

### Decide data handling

**Detail:** Decide what JDE data may leave your network for OCI, and get that approved by security and compliance before building.

### Test with real roles

**Detail:** Test the assistant under real JDE user roles so it never returns data the user could not see in JDE.

## Pitfalls

- Data sent to a cloud AI service leaves your network, which conflicts with export-controlled or air-gapped policies.
- An orchestration running under a powerful service account bypasses row-level restrictions the end user would have in JDE.
- Tools Release or AIS version gaps cause orchestrations to fail in ways that look like network problems.
- Unbounded generative output used in a business decision without review introduces errors that nobody checks.
- Cloud AI usage costs and quotas are easy to overlook until a high-volume process hits a limit.

## Ai Layer

### On-prem model behind the same Orchestrator pattern

**Description:** For sites that cannot send data to a cloud, the same Orchestrator-calls-REST pattern can point to an open-weight model served on your own GPUs, with retrieval over JDE data and documents.

### Grounded questions with the query shown

**Description:** A private assistant answers questions about orders, suppliers and inventory from live JDE data, shows the query used, and respects JDE roles.

### Document extraction for approval

**Description:** It reads supplier documents or inspection reports and drafts structured JDE entries for a person to approve before anything is written.

## Alternatives

- If the cloud is acceptable and your use case is narrow, Oracle's own OCI services through Orchestrator are the simplest route.
- Plain JDE One View reports or dashboards may answer the question without any AI at all.
- A private on-prem assistant is the better choice when data must not leave the site.

## Faq

### How does JD Edwards connect to OCI AI services?

**Answer:** Through JDE Orchestrator, which calls OCI REST APIs, per Oracle's white paper on AI with JDE. It requires a current Tools Release. The orchestration passes data to the OCI service and returns the result to a JDE flow.

### Can this run fully on-prem?

**Answer:** The OCI services themselves run in Oracle's cloud, so it is not air-gapped. The Orchestrator pattern can call a model hosted inside your network instead, which is the usual approach for restricted environments.

### Does the assistant respect JDE security?

**Answer:** Only if you design for it. Run orchestrations under the end user's role where possible, avoid broad service accounts, and test with real roles. Writes should require human approval.

## Source

https://www.oracle.com/webfolder/technetwork/tutorials/jdedwards/White%20Papers/Oracle-AI-with-JDE.pdf

## Search terms

jd edwards orchestrator oci generative ai digital assistant, jde orchestrator rest api, jd edwards ai assistant, oracle digital assistant jd edwards, jde orchestrator oci, jd edwards on-prem ai

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json