Introduction to the Supply Chain AI Assistant

Feature Summary

The Supply Chain AI Assistant provides help in three general areas:

• Accessing supply chain data via natural language. You can directly query either specific data entries (such as a specific customer order) or data in aggregate (such as revenue by region and timeline). Note the following user benefits:

  • The query does not have to contain technical detail such as the names of data schemas or SQL commands.

  • The AI Assistant can find supporting data for a specific issue resolution.

  • The user can verify the response and can provide additional input or information if the response is not accurate.

  • The AI Assistant is aware of any data in the Supply Chain Orchestrator data model and any custom data objects.

• Supporting decisions. The AI Assistant can help users understand the business logic and rules used in issue analysis and action recommendations.

• Memory management. There are many reasons memory is important for user experience with AI solutions, such as to help AI understand a specific industry better, to minimize user prompts by saving user-specific preferences or settings, and so on. The AI Assistant supports:

  • Short term and long term memory management

  • Global and user level memory management

Technical Notes

• The AI Assistant is installed separately from the Supply Chain Framework. You use IPM to install it on top of a Supply Chain Orchestrator instance.

• The AI Assistant is exposed through a set of APIs.

• The current release does not support any custom development, such as adding a custom agent. However, through memory management, the AI Assistant can be trained for different verticals/terminology/use cases.

• The AI Assistant requires specific LLM access through Azure. Initial release supports only bring-your-own-license for LLM.

  For information on supported models, see Posting and Updating Credentials.

• The AI Assistant implements name entity resolution to automatically generate the related context bases on business data, such as product information or other master data, which can significantly improve the user experience. For example. a user may ask “show me all apple inventory”; based on the master data in the Supply Chain Orchestrator instance, the AI Assistant will generate different queries for electronic retailer versus a grocery chain. 

Sample Questions

The following list represents the kinds of questions that the AI Assistant can answer:

• Show me monthly revenue year to date.

• Find me the latest order from Customer_XYZ.

• Where do we have inventory for Product_XYZ?

• Who are my top five customers?

• Show me top three high performing suppliers. For this, you may want to add this memory to get a more specific response: “I define supplier performance by their on-time shipping rate”)

• Why is option A recommended?

• What shipments do we expect today? For this, you may want to add a memory that defines a default location for the user: “My default warehouse location is Location_XYZ.”

• When do we expect the next shipment for Product_XYZ?

• We are running low on XYZ. Do we have a substitute part for that?

• Get me a list of all customers who orders Product_XYZ.

• Show me all suppliers for Product_XYZ.

In these questions, any referenced data such as Customer_XYZ does not have to be the exact value as saved in the database (thanks to the name entity recognition feature).

See Also

• Setting Up the Supply Chain AI Assistant

• Using the Supply Chain AI Assistant