Generative AI Features

Generative AI features improve design-time and runtime capabilities for AI Agent development. LLM-powered tools streamline the development process, reduce time and effort, and improve overall performance.

By default, all features are disabled. To use a feature, select a model and prompt, then toggle it on. You can update the model, prompt, and settings at any time.

---

Enable a GenAI Feature

Steps:

1. Go to Generative AI Tools > GenAI Features.

   

2. Select the Model and Prompt for the feature from the drop-down menus.

3. Toggle Status on. A success message confirms the change.

---

GenAI Features Navigation

Product Filter and Search

Features are filtered by context — the product you’re working in (Automation, Search, Agent, Contact Center). Add or remove this filter as needed. You can also use the search bar to quickly find any GenAI feature across the platform.

---

Change Model Settings for a Feature

Pre-built Model Settings

Default settings work for most cases. Adjust them only if you need to fine-tune the model’s behavior.

Steps:

1. Go to Generative AI Tools > GenAI Features.

2. Click the gear icon for the feature to open Advanced Settings.

   

Setting	Description
Model	The selected model for which settings are displayed.
Instructions or Context	Feature-specific instructions or context to guide the model.
Temperature	Controls output randomness. Higher values (0.8+) produce creative, less focused responses; lower values (0.5 and below) produce focused, deterministic responses.
Max Tokens	Total tokens used per API call. Affects cost and response time.
Number of Previous User Inputs	Number of prior user messages sent as context. For example, 5 sends the last 5 responses.
Additional Instructions	Specific rephrasing instructions — set a persona, tone, and more.
Similarity Threshold	For Answer from Docs: minimum similarity score between the user utterance and document chunks. Higher values restrict results; lower values may dilute responses.

Custom Model Settings

For custom models, you can only modify the post-processor script to map the actual response to the expected format.

Steps:

1. Go to Generative AI Tools > GenAI Features.

2. Click the gear icon for the feature.

3. Click Edit to view the Actual Response.

   

4. Click Configure to open the Post Processor Script.

   

5. Modify the script and click Save & Test to preview the response.

   

6. Click Save.

7. (Agent Node only) Enter the Exit Scenario Key-Value and AI Agent Response Key fields, then click Save.

   • Exit Scenario Key-Value: Defines when to end the GenAI interaction and return to the dialog flow.
   • AI Agent Response Key: Maps the response payload field used to display the agent’s reply to the user.
   

---

Feature Reference

The following tables provide a quick reference of the GenAI features available across the platform modules — Automation AI, Search AI, Agent AI, Contact Center AI, and Quality AI.

Automation AI

GenAI Feature	Description
Agent Node	Builds an AI Agent using LLMs, tool calling, and generative AI. Supports entity collection, context handling, multilingual conversations, and external integrations.
Prompt Node	Defines custom prompts for creative and custom use cases in the dialog flow.
Repeat Responses	Uses LLM to reiterate the last app responses when the Repeat App Response event is triggered.
Rephrase Responses	Enhances end-user experience with empathetic and contextual app responses.
Rephrase User Query	Improves intent detection and entity extraction by enriching the user query with conversation context.
Zero-shot ML Model	Identifies intents at runtime using OpenAI LLM based on semantic similarity, without training data.
Few-shot ML Model	Identifies intents at runtime using platform-hosted embeddings based on semantic similarity.
Automatic Dialog Generation	Builds production-ready dialog tasks from a brief intent description, with preview and regeneration options.
Conversation Test Cases Suggestions	Suggests simulated user inputs covering various scenarios at every test step, used to create test suites.
Conversation Summary	Fetches full conversation summaries using the Conversation Summary API and an open-source LLM.
NLP Batch Test Cases Suggestions	Generates NLP test cases for every intent including entity checks, ready for use in test suites.
Training Utterance Suggestions	Generates high-quality training utterances for each intent; review and add as needed.
Use Case Suggestions	Uses OpenAI LLM to generate use cases during the AI Agent creation journey.

Learn more about Automation AI GenAI Features.

Search AI

GenAI Feature	Description
Answer Generation	Generates answers to user questions based on ingested data.
Enrich Chunks with LLM	Refines or enriches extracted chunks using an external LLM.
Metadata Extractor Agent	Extracts sources and fields from a query, maps to structured data, and applies filters or boosts.
Query Rephrase for Advanced Search API	Adds contextual information to user queries to improve relevance.
Query Transformation	Identifies key terms in a query, removes noise, and prioritizes relevant documents.
Rephrase User Query	Enriches the user query with conversation context to improve intent detection and entity extraction.
Result Type Classification	Determines in Agentic RAG whether the user wants a direct answer or a list of search results.
Transform Documents with LLM	Enhances or updates documents during extraction using an external LLM.
Vector Generation - Image	Generates embeddings for ingested image content to enable visual semantic search.
Vector Generation - Text	Generates embeddings for ingested text content to enable semantic search and accurate responses.

Learn more about Search AI GenAI Features.

Agent AI

GenAI Feature	Description
Agent Coaching	Generates best-response suggestions based on conversation history and the current message.
Agent Playbook	Configure more accurate triggers and step adherence based on conversation history.
Conversation Summary	Creates summaries during chat transfers or as closing notes for conversations between users and agents.
Generating Opposite Utterance Suggestions	Uses LLMs to generate opposite utterances during design time.
Generating Similar Answer Suggestions	Uses LLMs to generate semantically similar answers during design time.
Generating Similar Utterance Suggestions	Uses LLMs to generate semantically similar utterances during design time.
Sentiment Analysis	Identifies customer sentiment across agent conversations using LLMs.

Learn more about Agent AI GenAI Features.

Contact Center AI

GenAI Feature	Description
Agent Response Rephrasing	Rephrases agent responses in formal, friendly, expanded, or neutral tones.
Conversation Summary	Creates summaries during chat transfers or as closing notes for conversations.
Disposition Prediction for Agent Wrap-Up	Auto-suggests disposition codes based on conversation context and disposition metadata.
Sentiment Analysis	Identifies customer sentiment across agent conversations using LLMs.

Learn more about Contact Center AI GenAI Features.

Quality AI

GenAI Feature	Description
Advanced Topic Discovery based on Custom Taxonomy and Resolution Detection	Extracts topics and intents from conversations; detects resolution and surfaces trends.
Agent Empathy Identification	Identifies agent empathy instances where customers expressed negative sentiment.
By Hold Adherence	Detects when an agent places a user on hold and resumes the conversation.
By Transfer Adherence	Identifies whether the agent informs the customer about a transfer.
By Value Adherence Validation	Validates metric values against reference data; returns binary adherence scores.
By Value Metric Extraction	Extracts by-value metrics from user, agent, and bot messages.
Churn & Escalation Identification	Identifies customer churn risk and escalation intents using LLMs.
Conversation Phase Identification	Identifies conversation phases for phase-level sentiment insights.
Crutch Word Usage Detection	Detects agent crutch word usage in customer conversations.
Default Script Adherence	Checks agent adherence to default script steps across key conversation phases.
GenAI-based Agent Answer Adherence and Customer Trigger Detection	Verifies agent adherence and detects customer triggers without utterance configuration or training.
Generating Similar QM Utterance Suggestions	Generates semantically similar phrases during design time.
Post Conversation Sentiment Analysis	Generates post-interaction sentiment and emotion moment insights.
Sentiment Analysis	Identifies customer sentiment across agent conversations using LLMs.
Topic Modelling	Extracts popular topics and intents from agent conversations.

Learn more about Quality AI GenAI Features.

---