---
schema_version: "1"
id: 0cf7a2f5edc31e08bf2189c1ba50177fb2c374f334590c7bf5ef3a3171b67438
title: Explain Order Routing Decisions with AI Tools
collection: admin
version: current
locale: en-us
portal_url: https://docs.salesforce.com/admin/current/en-us/commerce/264-0-0/om_steelthread_routing_explainer.md
canonical_url: https://help.salesforce.com/s/articleView?id=commerce.om_steelthread_routing_explainer.htm&release=264.0.0&type=5
source_site: help.salesforce.com
content_type: documentation
content_hash: eacbeab14f6b17fa97331cdf0dcae77f3c4b50033b9379bc154a7e8bf69bf68f
description: Ask why an order was routed to a specific location, and why it split, without reading logs or filing a ticket. With headless Order Management, the AI client reads the recorded routing explanation and returns the winning rule, the competing scores, and a per-shipment reason.
products:
- Commerce
- Salesforce Order Management
release: "264"
---
# Explain Order Routing Decisions with AI Tools
Ask why an order was routed to a specific location, and why it split, without reading logs or filing a ticket. With headless Order Management, the AI client reads the recorded routing explanation and returns the winning rule, the competing scores, and a per-shipment reason.
### Required Editions
Understanding a routing outcome normally means reading logs or escalating to someone who can. With headless Order Management, an analyst or a service agent can ask why an order was routed the way it was. The answer is an authoritative, plain-language reading of the recorded routing decision.
Note Because an AI client uses large language models, which are probabilistic, results can vary from one run to the next. If you get an unexpected result, describe to the agent what happened so that it can help you diagnose and adjust. You don't need a specific model.
## Scenario
Northern Trail Outfitters ships from a network of stores and distribution centers. Sam, a supply chain analyst, gets a question about why a single order was shipped from two locations instead of one. Rather than pull logs, Sam asks the AI client why the order was routed and split the way it was. In seconds, the agent returns the routing explanation. It shows which rule won, the scores of the competing locations, and how an override rule outranked a higher proximity score for one shipment. A service agent later uses that explanation to give the customer a plain-language summary of why the order arrived in two packages. The summary leaves out internal rule details.
## Requirements
Complete the shared setup in [AI Agent Use Cases for Order Management](https://help.salesforce.com/s/articleView?id=commerce.om_steelthread_overview.htm&release=264.0.0&type=5 "Use an AI agent to set up Order Management and complete common Order Management tasks conversationally, such as finding and canceling order items.").
This use case also needs:
- Order Management configured with locations, location groups, and a live routing rule group for at least one region.
- A rule group that includes proximity, overstock, and override rules, so the agent can explain both consolidated and split outcomes.
- The Order Management Agent or Order Management Operations Manager permission set. If your org doesn't include Omnichannel Inventory, use Order Management: Agent without Omnichannel Inventory or Order Management: Operations Manager without Omnichannel Inventory.
The routing-explanation functionality comes from the **Salesforce Order Management APIs** (Generally Available), which read the recorded routing explanation for an order. Recommended for supply chain analysts who validate routing behavior and for service agents who explain outcomes to customers. Assumes familiarity with fulfillment routing rules and location scoring.
## Agentic Tips
- Describe your goal in your own words. For example:
- "Why was order 00035107 fulfilled from the Dallas location?"
- "Why did this order split into two shipments?"
- "Explain the routing decision for order 00035107 in plain language I can share with the customer."
- Include the order number. For a customer-facing summary, tell the agent to leave out internal rule details.
- Have the agent explain the routing decision. Ask for the explanation, and then ask the agent to rephrase it as a customer-safe summary when you need one.
- Don't approve a change. The routing explanation is read-only.
- Avoid pasting credentials into a prompt. The agent acts as your authenticated user through the connected org.
## Design Considerations
- The routing explanation is read-only. The agent explains a past routing decision. It doesn't re-run or change routing.
- The recorded routing decision is the same for the analyst and the service agent. The agent reads that decision back, so both always see the same authoritative result.
## Explain a Routing Decision
Ask why an order routed to a location, and why it split.
### Required Editions
1. Confirm that your AI client is connected to Order Management as an authenticated user with read access to the routing explanation.
2. Ask the agent why an order was fulfilled from a particular location and shipped as one package.
The agent retrieves the routing explanation. It shows which rule won, the scores of the competing locations, and the inventory context that led to a consolidated shipment.
For an order that split into multiple shipments, ask why it split instead. The agent explains each shipment separately: why each item routed to its location, and how an override rule outranked a higher proximity score.
3. To brief a customer, ask the agent for a plain-language summary that leaves out internal rule details.
The agent returns the routing explanation for the order you asked about. The explanation includes the winning rule, the per-rule contributions, and, for split orders, a per-shipment reason.