---
title: What glog.genAI.enabled actually turns on in OTM 26B and 26C, and what to ask before you flip it
description: A read of Oracle's own docs: what the OTM GenAI property enables, which agents read or write your tenant, where the data goes, and what to ask before you flip it.
canonical: https://otmdirectory.com/guides/otm-genai-enabled-what-it-turns-on
updated: 2026-09-11
site: OTM Tools
---


# What glog.genAI.enabled actually turns on in OTM 26B and 26C, and what to ask before you flip it

Oracle put generative AI inside Oracle Transportation Management across the 26B and 26C cloud releases. One property, glog.genAI.enabled, gates it. The What's New banner tells admins to set it to true. What it does not tell them is which surfaces that switch opens, which of the new agents can change data in the tenant, or where the shipment data goes when a summary is generated.

This is a read of Oracle's published documentation as of September 2026: the 26B What's New entries, the 26C Transportation and Global Trade Management AI Guide, and the Fusion AI platform statements on data handling. Nothing here was tested against a live tenant, and the sources are listed at the end so you can check each claim. It is written for the OTM admin or transportation lead who has to answer a security team before enabling anything.

## Key takeaways

- glog.genAI.enabled defaults to false, can only be set by DBA.ADMIN through Property Sets, and is all-or-nothing: one property opens every generative feature in the release.
- Flipping it alone gives you two things inside OTM: the Shipment Summary AI Assistant (Cohere, inside the Enhanced Workbench) and the OTM Assistant on saved queries. The AI Chat agents need a separate Fusion pod with AI Agent Studio and a day of identity setup.
- Six agent families ship for OTM and GTM. Three of them write to your tenant: Shipment Change (change carrier, secure resources, accept or decline tenders), Dock Scheduling (create, modify, cancel appointments), and the two GTM assistants (POST items and documents, PATCH trade parties and locations).
- The agents reach OTM through a confidential application in your OTM identity domain, authenticated with a signing key you generate and paste into AI Agent Studio. The per-user access control list entries are the real blast-radius control.
- Oracle's own wording on data: model input and output is processed within a secure OCI environment, which might be outside of the local OCI region, and is not persisted or used to train models. Both halves of that sentence matter to a residency review.
- Tenants hosted from Asia Pacific or the Middle East must open a service request before the property does anything.

## One property, three surfaces

The 26B What's New entry for the Shipment Summary assistant carries the enabling instruction every later feature points back to. Set glog.genAI.enabled to true under Configuration and Administration, Property Management, Property Sets. The default is false. Only the DBA.ADMIN user can open that page. Oracle's note says the property only needs to be set once to gain access to all the Generative AI features, which is a convenience for the admin and a governance problem for everyone else, because there is no per-feature switch underneath it.

The 26B online help page About Oracle AI for OTM and GTM lists what the property opens. First, the Shipment Summary AI Assistant, launched from a buy-side shipment action, which renders a model-written summary inside a Shipment Summary Workbench. Second, the OTM Assistant on non-finder saved queries, which offers Saved Query Analysis and an Explain Plan for custom SQL. Third, AI Chat, described as agentic AI in OTM to review bulk plan diagnostics and run rate inquiries, and pointed at the Oracle AI Studio agents.

That third surface is where the wording gets ahead of the plumbing. The property lets OTM show the chat entry point. The agents behind it live in Fusion AI Agent Studio, and the 26C AI Guide prerequisites are explicit: any 26C Fusion module which comes with AI Agent Studio, an OTM 26C instance, admin access to both the Fusion and OTM identity domains, admin access to AI Agent Studio, DBA.ADMIN on OTM, and a tool to generate keys and certificates. If your shop runs OTM without a Fusion pod, the property gives you the first two surfaces and a chat button with nothing behind it.

One more gate from the same 26B entry: if you are hosted from the Asia Pacific or Middle East regions, you need to open an SR requesting the GenAI feature before the property takes effect.

## What the Shipment Summary sends to the model

The Shipment Summary is the feature most shops will see first, because it needs nothing beyond the property. Oracle's 26B entry says the connection to the large language model is handled automatically as part of your upgrade or provisioning, and names the model: Cohere.

The same entry lists the shipment data elements the summary can draw on. Shipment ID, service provider ID, transport mode, start and end times, transit time, source and destination location details, total actual cost, total weighted cost, volumes, weights, en-route status, number of stops, loaded distance, chargeable volume and weight, planned cost, preferred carrier, rate offering ID, planned rate offering ID, and stop-level detail. Read that list as a security reviewer would: carrier assignment, the rate records behind the cost, and customer locations leave the OTM application boundary and go to a model call every time a planner clicks the action.

You can control the layout but not the payload. The action parameter SHIPMENT_SUMMARY_LAYOUT_ID points the action at a custom Enhanced Workbench, and the workbench table parameter defaults carry an AI Summary option. Nothing in the documentation describes a field-level allow list for what the summary reads.

## The six agent families, and which ones write

The 26C AI Guide lists nine agents staged in AI Agent Studio for OTM and GTM. Grouped by what they touch, they are six families, and the guide's access control page gives the exact ACL entries a user needs for each one. That page is the most useful governance document Oracle has published on this, because ACLs are the one lever you hold that the agents cannot argue with.

Bulk Plan Diagnostic ships in four variants: Full Summary, Open Ended, Order Errors, and Summary. They take a diagnostic log file ID or a bulk plan GID and explain what the planner did and why, with a retrieval tool loaded from the OTM Bulk Plan Tuning Guide PDF that you upload yourself. Required ACL: REST - Planning Diagnostics. Read-only.

Rate Inquiry takes an origin, a destination, an item, a quantity, and a volume, and runs one. Required ACLs: REST - Item - View, REST - Location - View, REST - RIQ Actions. Read-only, though it exposes your rate structures to whoever can chat.

Shipment Change is the one to slow down on. Oracle's guide lists four prompts: change service provider, secure resources, accept tender, and decline tender, each keyed on a shipment GID. Required ACLs: REST - Shipment - View, REST - Shipment Actions, REST - Tender Actions, REST - Tracking Event Configuration - View. Those are tender-lifecycle writes against live shipments. The guide's test page does not describe a confirmation step between the prompt and the action, so treat the prompt as the confirmation until you have watched it behave in a non-production tenant.

Dock Scheduling can schedule, modify, cancel, and look up appointments, and it asks a clarifying question when a location has more than one stop. Required ACLs: REST - Appointment - View, REST - Appointment - Update, REST - Location - View, REST - Shipment Actions. Appointment - Update is a write to a resource your carriers and your dock staff both depend on.

The two GTM assistants are writes as well. Product Classification Assistant needs GET and POST on items and documents plus GET on item structures. Restricted Party Preprocessing Assistant needs GET and PATCH on trade parties and PATCH on locations. Both create or amend compliance records.

Read the ACL names once more. Every entry begins with REST. The agents reach OTM only through its REST resources and the actions those resources expose. None of the other ways into OTM are in play: GLogXML transmissions over HTTP or SOAP, CSVDataLoad, the DBXML servlet, Integration Manager uploads, or the Power Data screens where configuration lives. That is good news for the security review, because the blast radius is exactly the REST surface and the ACLs on it, and there is no path from a chat prompt to a property set or an agent definition. It is also the honest limit of what these agents are: they act on transactional objects, orders, shipments, appointments, tenders, and they cannot configure the system they run in. Anything you do to OTM through the other integration methods stays outside their reach, in both directions.

The Logistics Execution Command Center, which Oracle released in 26B as a Fusion agentic app launched from the WMS Redwood desktop rather than as an OTM chat agent, sits alongside these. It joins WMS execution data with OTM order and shipment data, and on the OTM side it needs REST - Order - View and REST - Shipment - View. Oracle's blog post announcing it describes an Ask Oracle chat over both systems and says that accepting a recommendation triggers the warehouse allocation workflow, so it acts on the WMS side as well as reading from OTM. Neither the post nor the AI Guide says what a session logs or where.

## How the agents get in: the identity chain

The AI Guide spends most of its length on identity, and that is where a governance review should spend its time too. Oracle splits the setup across thirteen how-to pages; they are listed in the order you run them, as setup steps 1 to 13, in the sources at the end of this article. The chain has five links.

You generate a public and private key pair with a certificate, and the guide names the files: jwt-signing.key and jwt-signing.crt. You create a confidential application in your OTM identity domain and import the certificate against it. In AI Agent Studio you create a data source named OTMENV holding your OTM base URL, your OTM identity token endpoint, the client ID and scope of that confidential application, and the contents of both key files, the private key included. Oracle notes one data source per OTM instance per release. Every call an agent makes into OTM is made as that confidential application.

The reverse direction is configured in OTM under Configuration and Administration, Power Data, AI. An AI Agentic Trust record named AI FUSION holds the Fusion identity domain's token URL, the RS256 signing algorithm, the Fusion confidential application's scope, client ID and certificate alias, and a reference to the signing key. Two AI Agentic Link records, AI FUSION - V1 and AI FUSION - V2, hold the Fusion orchestrator endpoints at /api/fusion-ai/orchestrator/agent/v1 and v2.

Users are matched by name. Each OTM user who will chat needs a user in AI Agent Studio, and the OTM user Nickname must be the same as the user name in AI Agent Studio. The guide suggests copying the Fusion user from SCMOPERATIONS, SCMIMPL, or an admin user. That is a sensible shortcut for a test tenant and a poor default for production, because the Fusion side of the identity inherits whatever the copied user had.

Three consequences follow. Your OTM private key sits inside the Fusion pod's Agent Studio configuration, so key rotation is now a two-system procedure. The confidential application is the identity OTM sees on every agent call, so the audit trail in OTM attributes agent writes to it unless the orchestrator passes the end user through, which the guide does not describe. And the per-user ACL grants from the previous section are the only place where you decide that a given planner can ask about rates but cannot accept a tender.

## Where the data goes, in Oracle's words

Oracle answers the residency question in the Fusion AI platform documentation, not in the OTM guide, and it is worth quoting exactly. Each organization gets their own securely partitioned instance of the platform within their local Oracle Cloud Infrastructure region. Then: the platform accesses the pretrained generative AI models securely, and the input and output data from the models for the given use case is processed within a secure OCI environment, which might be outside of the local OCI region. And: there's no sharing or mixing of data between organizations, and your information won't be persisted or used to train models or to generate responses for others.

The OCI Generative AI data handling page adds that the service does not retain inference inputs or outputs, and that prompts and responses are not shared with third-party model providers such as Cohere.

So three statements are true at once. The data is not stored by the model service. It is not used for training. And it may be processed outside the region your tenant lives in. A team with a data residency commitment to a customer or a regulator needs the third statement in writing before the property is set, and the 26B note about Asia Pacific and Middle East tenants needing a service request is the visible edge of that same constraint.

Which model, and whose cloud, depends on the surface. The Shipment Summary is fixed to Cohere, running on Oracle's infrastructure. The chat agents run in AI Agent Studio, where the model is picked per agent from the LLMs Oracle provides, or from a model you add yourself under Bring Your Own LLM: Oracle's 26C page lists OpenAI through Microsoft Azure, Google Gemini through Vertex AI or the direct API, Anthropic through Vertex AI, and additional models on Oracle Cloud Infrastructure. Adding one of those moves the model call to a third party's cloud under your own credentials and contract, and Oracle's data statements above stop at that boundary. Whether the pre-seeded OTM agents can be switched to an added model is not stated in the OTM guide; ask before assuming either answer.

What the documentation does not give you: a list of which OCI region processes a given tenant's model calls, a statement of what the Fusion orchestrator logs from a chat session, or a retention period for those logs. Those are the questions to put to your Oracle account team, and to write down the answers.

## What has to be true in your configuration first

The agents are only as good as the objects they can find, and they find everything by GID. Every example prompt in the guide is written as DOMAIN_NAME.OBJECT_ID: a shipment, a location, a bulk plan, a diagnostic log file. A planner who does not know the location GIDs for a lane cannot ask the rate agent about it. Shops with clean, human-readable XIDs will get a chat experience; shops with system-generated identifiers will get a lookup exercise.

Bulk Plan Diagnostic needs diagnostic log files to exist, which means diagnostic logging turned on for the bulk plans you want explained, and it needs the tuning guide PDF uploaded and processed through the Fusion scheduled process Process Agent Documents before its retrieval tool returns anything.

Rate Inquiry needs rate records, lanes, and items that resolve from the identifiers a user will type. Dock Scheduling needs locations with appointment configuration and calendars. Shipment Change needs the tender workflow to be one you are willing to drive from a chat window.

And the nickname match is a real prerequisite, not a footnote. If an OTM user's Nickname field does not equal their Agent Studio user name, the agents do not know who is asking.

## Eight questions to answer before you flip it

Which of the three surfaces do you want? The property opens all of them. If the answer is only Shipment Summary, plan to leave the chat entry point unwired rather than expecting a switch that does not exist.

Is shipment cost and carrier data acceptable input to a model call? The Shipment Summary sends it. If the answer is no for some domains, there is no field-level control documented; the control is who holds the action.

Which users get AI Chat, and what does their Agent Studio identity inherit? Copying from an admin user is the guide's shortcut. Decide what a production copy should be based on.

Which ACLs do you grant, and in what order? Start with the read-only agents: Planning Diagnostics, Item View, Location View, RIQ Actions. Hold Shipment Actions, Tender Actions, and Appointment Update until you have watched the read-only agents behave in a non-production tenant.

Where is your OTM pod hosted, and does processing outside the local region matter to any customer contract or regulator you answer to? Get Oracle's answer in writing.

Who holds jwt-signing.key, where is it stored on your side, and what is the rotation procedure now that a copy lives in AI Agent Studio?

What gets logged, where, and who reviews it? Nothing in the AI Guide or the Command Center announcement documents what a chat session or an agent write leaves behind. Find out what that covers for chat sessions and agent writes, and how you would reconstruct who asked an agent to decline a tender.

Have you set the property in a non-production property set first? Property sets are per environment. There is no reason for the first tenant to see glog.genAI.enabled to be production.

## What this read did not cover

Oracle revises these pages each quarter, and the 26C What's New adds conversational access to shipment and order information and self-service dock appointment scheduling on top of what is described here.

If you want the same questions answered against your own configuration, with the ACLs, property sets, and identity records read directly rather than from the guide, that is a read-only engagement. Nothing is written to the tenant and nothing leaves it. Use the contact form on this site.

## Frequently asked questions

### Does setting glog.genAI.enabled send my OTM data to Oracle's cloud AI?

Setting the property alone sends nothing. Data leaves when a user runs a feature it enables. The Shipment Summary action sends the shipment's carrier, cost, timing, and location fields to a Cohere model call. The chat agents send whatever the prompt and the agent's REST calls return. Oracle states this data is processed in a secure OCI environment that might be outside your local region, and is not persisted or used for training.

### Can I enable the Shipment Summary without the AI Agent Studio agents?

Yes. The Shipment Summary and the saved query assistant need only the property. The agents need a Fusion pod with AI Agent Studio, identity setup in both directions, and matching users. Without that setup the agents are not reachable, but there is no property-level way to turn the chat entry point off on its own.

### Do any of the OTM AI agents write to my tenant?

Three families do. Shipment Change can change the service provider, secure resources, and accept or decline tenders. Dock Scheduling can create, modify, and cancel appointments. The GTM Product Classification and Restricted Party assistants create and update items, documents, trade parties, and locations. Each is gated by specific REST ACL entries on the user, listed in Oracle's 26C AI Guide.

### Is my data used to train Oracle's models?

Oracle's Fusion AI platform documentation says your information won't be persisted or used to train models or to generate responses for others, and the OCI Generative AI data handling page says prompts and responses are not shared with third-party model providers. The separate statement that processing might be outside of the local OCI region is the one residency reviews should read carefully.

## Sources

- [Oracle AI Guide 26C, setup step 1: Prerequisites](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/important-prerequisites.html)
- [Oracle AI Guide 26C, setup step 2: Generating public and private keys with a certificate](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/generating-public-private-keys-with-a-certificate.html)
- [Oracle AI Guide 26C, setup step 3: Creating a confidential application in OTM IDCS](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/creating-a-confidential-application-in-otm-idcs.html)
- [Oracle AI Guide 26C, setup step 4: Importing the certificate into the IDCS of OTM](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/importing-a-certificate-into-idcs-of-otm.html)
- [Oracle AI Guide 26C, setup step 5: Adding the users in OTM IDCS, Fusion IDCS, and OTM](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/adding-a-user-in-otm-idcs-fusion-idcs-and-otm.html)
- [Oracle AI Guide 26C, setup step 6: Creating OTM users in AI Agent Studio](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/creating-otm-users-in-ai-agent-studio.html)
- [Oracle AI Guide 26C, setup step 7: Creating a confidential application in Fusion IDCS](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/creating-confidential-application-in-fusion-idcs.html)
- [Oracle AI Guide 26C, setup step 8: Importing the public key on the IDCS of the Fusion pod](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/importing-public-key-on-idcs-of-fusion-pod.html)
- [Oracle AI Guide 26C, setup step 9: Creating the AI Agent Studio data source (OTMENV)](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/creating-the-ai-agent-studio-data-source.html)
- [Oracle AI Guide 26C, setup step 10: Adding the Bulk Plan Tuning Guide to the RAG tool](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/adding-the-bulk-plan-tuning-guide-to-the-rag-tool.html)
- [Oracle AI Guide 26C, setup step 11: Setting up AI Agentic Trusts (the AI FUSION record)](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/setting-up-ai-agentic-trusts.html)
- [Oracle AI Guide 26C, setup step 12: Setting up AI Agentic Links (AI FUSION - V1 and V2)](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/setting-up-ai-agentic-links.html)
- [Oracle AI Guide 26C, setup step 13: Adding access control lists to users (ACLs per agent)](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/adding-access-control-lists-to-users.html)
- [Oracle: About Oracle AI for OTM and GTM (26B online help)](https://docs.oracle.com/en/cloud/saas/transportation/26b/otmol/ai/about_ai_in_otm_gtm.htm)
- [Oracle: OTM 26B What's New, AI Assistant Shipment Summary (enabling steps, Cohere, data elements)](https://docs.oracle.com/en/cloud/saas/readiness/logistics/26b/otm26b/26B-otm-wn-f43428.htm)
- [Oracle: OTM 26B What's New, AI Agentic App Logistics Execution Command Center](https://docs.oracle.com/en/cloud/saas/readiness/logistics/26b/otm26b/26B-otm-wn-f43787.htm)
- [Oracle blog: Logistics Execution Command Center, smarter, faster logistics with agentic AI (Ask Oracle chat; accepted recommendations trigger WMS allocation)](https://blogs.oracle.com/scm/logistics-execution-command-center-agentic-ai)
- [Oracle: AI Guide 26C, OTM AI Chat Agents](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/otm-ai-chat-agents.html)
- [Oracle: AI Guide 26C, Testing the SHIPMENT CHANGE Agent](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/using-the-shipment-change-agent.html)
- [Oracle: AI Guide 26C, Testing the DOCK SCHEDULING Agent](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/using-the-dock-scheduling-agent-scheduling-an-apppointment.html)
- [Oracle: AI Guide 26C, Testing the DOCK SCHEDULING Agent: Modifying an Appointment](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/using-the-dock-scheduling-agent-modifying-an-appointment.html)
- [Oracle: AI Guide 26C, The Logistics Execution Command Center Agentic App](https://docs.oracle.com/en/cloud/saas/transportation/26c/otmai/using-the-logistics-execution-command-center-agentic-app.html)
- [Oracle: What's the Oracle AI Platform for Fusion Applications? (data handling statements)](https://docs.oracle.com/en/cloud/saas/fusion-ai/aiaqa/what-is-the-oracle-fusion-ai-platform.html)
- [Oracle: Data Handling in OCI Generative AI](https://docs.oracle.com/en-us/iaas/Content/generative-ai/data-handling.htm)
- [Oracle blog: Oracle Fusion Cloud SCM 26C, practical innovation across the supply chain](https://blogs.oracle.com/scm/oracle-fusion-cloud-scm-26c-practical-innovation-across-the-supply-chain)
- [Oracle: AI Agent Studio 26C, Add your LLM (Bring Your Own LLM providers)](https://docs.oracle.com/en/cloud/saas/fusion-ai/26c/aiaas/add-your-llm.html)
- [OTM Tools: How systems integrate with Oracle Transportation Management (the integration methods the agents do not use)](https://otmdirectory.com/guides/otm-integration-methods)

---

Canonical: https://otmdirectory.com/guides/otm-genai-enabled-what-it-turns-on
HTML version: https://otmdirectory.com/guides/otm-genai-enabled-what-it-turns-on