Configuring ICM Intents and Fulfillment
In a traditional Aisera Workflow, Intents invoke the workflows. If you are using RAG indexing, Conversation AI 2.0, or Hyperflows, you may be using the Title or Descriptions instead of Intents in your workflows. Choose Intents from the Aisera Admin UI to see all of the Intents that are visible to your instance of the Aisera platform (your company's Aisera tenant).
This page lists all the Intents that are available to your Aisera applications. Use the Search option on this page to search for existing Intents, and filter by Category or Status. If you don't see an Intent that you want to use to launch your workflow, select the Browse Intent Library to add an existing Intent to your Aisera platform, or choose the + New Intent button to create a new Intent.
Adding Intents to Your Bot
Click the Browse Intents Library button.
Select the Intents you want to use with your bot and click the Add Selected Intents button.
The Intent Details window allows you to edit individual Intents. You can also create Escalation Flows, add Variables, and create Phrases that will trigger an Intent.
Click on an individual intent to edit the Intent.
Choose Add Fulfillment to associate a Fulfillment with the Intent.
Select the Type of Fulfillment
Set any Fulfillment Bindings.
Create a New Intent
This section describes the available options when you create a new intent.
To create a new Intent:
Click the + New Intent button to start a new Intent. You can either use an existing flow or create a new flow to fulfill that intent.
As you create the Intent, enter a Name and Category for the Intent.
Choose the type of Fulfillment Flow that you want to associate with your Intent. The options for each flow are described in the section below. On the main Intents window, you can filter a search by when you choose one of the flow names at the top of the window.
Most of the Fulfillment Flow options allow you to choose an existing flow or create a new flow.
After you choose the Fulfillment Flow Type, your new Intent will show up in the view that you use to customize that flow type.
Action Flow:
The Action Flow is the part of the Workflow that performs tasks behind-the-scenes, like searching for answers or making a decision.
Conversation Flow:
The Conversation Flow consists of the Request that a user makes to the application or bot (application/bot UI) and the answer or Fulfillment that the application returns to the user (inside the application/bot UI).
Limited Response Includes Clarification Message
If a query matches only one KB document it is not necessarily unambiguous. For example, a prompt about “software installation” matching only one document related to printer installation.
In this case, the agent acknowledges the fact that its context is limited relative to the scope of the query and serves the matched document along with a disclaimer, such as, “I was only able to find information about installing a printer”.
Hyperflow:
The Hyperflow is a type of workflow that is associated with a specific resource. See Hyperflows for more information.
Knowledge Article:
A Knowledge Article flow includes a search of one or more knowledge articles that your Aisera application or bot has been associated with.
Service Catalog Flow:
A Service Catalog flow includes a search of one or more service catalogs that your Aisera application or bot has been associated with.
Custom Message:
The Custom Message Flow is a message that you add to a flow that is not part of the default conversation or action.
Intent Details
After you choose or create an Intent, the Intent Details window is visible. Note that the details for each Intent are grouped in multiple tabs based on their Fulfillment, Escalation Flow, Variables, and Phrases associated with this Intent.
Click on + Add Fulfillment to add an action/knowledge flow, or both. One or more flows can be added; this is called multi-fulfillment.
Add an Intent Fulfillment
Click + Add Fulfillment to add an action/knowledge flow, or both.
You can add one or more flows. Using multiple flows with a single intent is called Multi-Fulfillment.
After you provide a Fulfillment, you can see all of the additional Intents.
Add Intent Phrases
Intent phrases identify the actions that a user wants to take. For example, "get tickets", or "get ticket details" will be some of the phrases used to get ticket-related details.
Phrases with variables can be used to match several generic query phrases.
Every time you add or delete phrases, the model needs to be retrained and published. The changes in the workflow will take effect only after the model is retrained and published.
You can add, delete, modify, activate/inactivate the phrases at any time.
Phrases are the training examples that an ICM bot tries to match with Intents. They can contain variables written with the convention: ${varName}.
Single-word phrases
Single-word phrases are fine. However, they can only be matched against single-word phrases (after stop word removal) as input by the user. This is to limit the False Positive (FP) rate.
Single-word phrases can ONLY be found via an exact match (after punctuation and stop word removal)
Using variables inside phrases
Variables should be used whenever possible. Instead of writing out a training phrase of “book conference room”, it’s better to add a variable (varConferenceSystems) for the appropriate entity category (such as, Conference System) and write out a training phrase like: “book ${varConferenceSystems}”.
If the training phrase of “book conference room” was used instead, if someone were to type “I want to book a conference room” in Webchat, this phrase might get transformed to “I want to book a conference-room” as part of the NER step of ICM scoring. Hence, ICM may encounter a False Negative (FN).
How to Add Variables
Choose the + Add Variable button to include variables in your Intent, that you can use in your phrases. This is where Ontology plays a role.
You can use values and variables as part of expected phrases to invoke and use in a Fulfillment. The variables can be edited or deleted at any time.
Variables are used for slot-filling within a phrase.
The variable Name can be anything. Each variable has an associated Entity Category (Entity Class). This variable can thus take on any entity within that Aisera Entity class.
Activate Intents
To activate an Intent, choose the Activate Intent button at the top of the Intent Details window.
Choose OK when you see the confirmation question.
Last updated