Automation AI
Unified Conversation Session
Earlier, integrations with Kore.ai products—Platform, SmartAssist, Search Assist, and Agent Assist—relied on APIs or Webhooks. This resulted in disjointed conversations with no central session to track complete interactions. With AI for Service, conversations are unified across modules, enabling seamless end-to-end tracking within a single session. This improves analytics, simplifies cross-module data retrieval, and enhances workflow efficiency.Improved Dialog Builder Experience
The new Dialog Builder v3 offers an intuitive, collaborative environment with an infinite canvas, multi-user collaboration, visual flow design, and no-code templates. This reduces development time and improves the quality and maintainability of conversational flows. Key capabilities:- Free Flow Designing: Design flows without defining logic upfront—similar to tools like Lucidchart or Miro.
- Easy Transitions: Connect nodes without fully defining transition rules. Start with a description and enforce rules later during testing or publishing.
- Infinite Canvas: Arrange flows logically by purpose or objective for easy review and auditing.
- Commenting: Add comment threads to ask questions and exchange ideas.
- Spacious Nodes: Read and write key information—name, type, prompts/messages, and transition conditions—clearly.
- Annotation Tools / Whiteboarding: Add annotations to explain flows, capture ideas, and support brainstorming alongside the flow.
- Co-Designing: Allow multiple users to collaborate in parallel—essential when conversation designers and AI Agent developers co-build.
- Sharing: Share designs with external stakeholders to collect feedback or get approvals.
- New Node Panel: Simplified panel with four broad categories: Inform, Ask, Action, and AI.
- No-Code Templates: Configure templates in the UI without writing code.
- Node Sequences: Group nodes into sequences for better flow organization.
Advanced AI Capabilities
DialogGPT and enhanced RAG (Retrieval-Augmented Generation) capabilities enable zero-training intent detection, contextual understanding, and intelligent conversation management. The Agent Node with tool calling enables sophisticated LLM-powered interactions that handle complex tasks and integrate with external systems.DialogGPT
DialogGPT introduces a paradigm shift in intent detection and conversation management. It uses embedding-based retrieval combined with generative AI to improve intent detection and conversation handling—without requiring training utterances. Key capabilities:- Contextual intent detection without training utterances
- Identification and fulfillment of ambiguous and multi-intents
- Management of common conversation intents: pausing, repeating, restarting, and transferring to agents
- Granular intent identification using the knowledge graph, with LLM-powered clarifying questions
- Out-of-the-box event handlers to fulfill user requests
Seamless Integration with Search Module (RAG Capabilities)
Powered by LLMs, Search AI provides accurate, context-aware answers. It integrates diverse data sources—web crawlers, structured data, documents, catalogues, and knowledge repositories—into a comprehensive answer index. Key capabilities:- Answer any question from the knowledge base
- Connect to 50+ external connectors for automatic knowledge base sync
- Customize the chunking strategy
- Manage the RAG pipeline
- Test answers using the debug window
Agent Node with Tool Calling
The Agent Node uses LLMs and generative AI with tool calling to create versatile AI agents capable of handling complex, dynamic, data-driven tasks. Key capabilities:- Entity Collection: Simplifies gathering entities within a conversation, reducing the need for multiple entity nodes and making interactions more natural.
- System Context, Business Rules, and Exit Scenarios: Incorporates system context, business rules, and predefined exit scenarios to guide conversations and maintain alignment with enterprise rules.
- Multilingual Support: Supports English and non-English languages for diverse user bases.
- Configuration Flexibility: Configurable like any other node, enabling seamless integration into existing conversational flows.
- Interaction with External Systems: Tool calling enables real-time data retrieval, calculations, and system-specific operations for dynamic, data-driven experiences.
- Dynamic Prompt Enhancement: The prompt dynamically adapts to include tool definitions and contextual information so the model responds or calls a tool as appropriate at each step.
Seamless Integration with Agentic Apps for Multi-agent Orchestration
Fully autonomous Agentic Apps integrate easily with the Platform, simplifying the creation of highly contextual, self-service automation experiences using multi-agent orchestration. Key features:- Choice between Orchestrated Autonomy or Full Autonomy automation modes
- Full support for digital and voice channels
Operational Efficiency
Features like Experience Flows for channel-specific welcome experiences, conditional FAQ answers, and an enhanced Marketplace with ready-to-use templates accelerate time-to-value and reduce the complexity of managing conversational AI applications. Feature Comparison The following table summarizes key differences between XO v10 and XO v11.| Feature | XO v10 | XO v11 |
|---|---|---|
| Channels | 30+ Enterprise and Social digital channels. Third-party IVR/CCaaS integration via VXML. Voice Gateway requires SmartAssist integration; one instance per account. | All XO v10 channels plus a seamless integrated flow for Kore.ai Voice Gateway. Multiple Voice Gateway instances supported via separate XO v11 apps. |
| Welcome and On-Connect Events replaced with Flows | OnConnect Event: Triggers a message, Dialog Task, or script. Welcome Events: Welcome Event, Facebook Welcome Event, Telegram Welcome Event, and Telephony Welcome Event—each triggering a Dialog Task or message. | Experience Flows define the welcome experience when a user connects from a channel. Every channel maps to one Welcome Flow. OOTB Welcome Flows are provided on upgrade. Users can continue with XO v10 events or switch to Welcome Flows. |
| Intelligent User Utterance Handling | Each utterance routes to the NLP engine, which determines whether to start a Dialog Task, answer an FAQ, or handle Small Talk. | Configurable Fallback: Configure the app to prefer Generative Answering and fall back to Automation, or prefer Automation and fall back to Generative Answers. |
| Dialog Creation Options | Predefined basic dialog templates per enabled integration. Automatic Dialog generation from within the Dialog Builder canvas. | All XO v10 options with improved UX, plus:
|
| FAQs | Simple FAQ questions and answers. | Conditional FAQ Answers based on context, enabling more personalized responses. |
| Dialog Builder | Dialog Builder v2. | Dialog Builder v3 with additional features:
|
| Agent Transfer | Prebuilt third-party Agent Transfer options for routing conversations to external agent desktops. | Self-Transfer to CCAI: Transfer to Kore.ai Contact Center AI within the same app. Enhanced Agent Transfer Node: Includes Agent Skills and Agent Groups properties for accurate routing with Kore.ai Contact Center AI. |
| Answer from Documents | Basic support for answering from user-uploaded PDF and Word documents with generative answers. | Upload files, provide a web URL for crawling, or connect to a cloud source. Advanced capabilities include customizable chunking strategies and business rules, powered by the XO GPT model. |
| Display Language | English, Japanese, and Korean. | English only currently. Japanese, Korean, Simplified Chinese, French, and Spanish are in progress. |
| Marketplace | Predefined Bot/process Apps with OOTB dialogs, available on the XO v10 homepage. | Enhanced in-app Marketplace with:
|
| Create New Bot/App | Create an Automation Bot in two steps:
| Guided App creation with any of four modules: Automation AI, Contact Center AI, Search AI, and Agent AI. Includes live testing at the end. Start with one product and enable others later. |
| Talk to Bot | Chat testing and debugging only. | Both voice and chat testing. |
| Guided Onboarding | Simple onboarding. | Automation AI onboarding helps users create an end-user-facing, publishable app. |
| Setup Guide | — | Setup Guide lists important features in a logical order to help users start their learning journey. |
| GenAI and LLM |
| All XO v10 GenAI features, plus GenAI capabilities for Search AI, Agent AI, and Contact Center AI. |
| User Role Management | Managed in Admin Console only. | Available on the homepage. App Owners can create user roles inside the app; roles are reflected across the workspace and other apps. |
| User Roles | Default system roles (Role Type: BOT):
| Additional default roles (Role Type: APP):
|
| Bot/App Name Change | Cannot change the bot name after publishing. | App name can be changed at any time. |
| Deletion of Bot/App | Published bots can only be deleted by suspending tasks in the Admin Console. | App Owners can delete the app from the builder. The app remains recoverable for 30 days. |
| Public APIs | API scope is limited to the workspace and individual bot. | API scopes cover workspace, Automation AI (Bot APIs), Contact Center AI (SmartAssist APIs), Search AI (Search Assist APIs), and Agent AI (Agent Assist APIs). |
Search AI
Search AI is the next generation of Kore.ai’s search and retrieval platform, with improvements across data ingestion, content extraction, retrieval, and answer generation.Intelligent and More Accurate Answer Generation
Agentic RAG Architecture Search AI uses an Agentic RAG (Retrieval-Augmented Generation) architecture where multiple autonomous agents collaborate to optimize query interpretation, retrieval, and response generation. This delivers highly accurate, context-aware answers for complex and multi-faceted queries. Learn more. Advanced Retrieval Technology Search AI offers hybrid retrieval based on RAG Fusion, combining multiple keyword and vector scores to boost accuracy, relevance, and performance. Learn more. Channel-Aware Response Formatting Responses are automatically optimized for voice or digital channels, ensuring the most effective format for each medium. Contextual Answer Generation Search AI improves response accuracy by sending entire documents—rather than fragments—as context to LLMs, enabling the model to grasp structure and intent for more relevant answers.Improved Content Ingestion Capabilities
6x More Connectivity Options- 60+ connectors—a 6x increase over SearchAssist
- Support for both structured and unstructured data from connectors
- Universal RACL support across connectors
- Streamlined configuration and setup
- New, improved crawler with significantly higher success rates
- 80% improvement in web crawling speed
- Better handling of complex website structures, dynamic content, and JavaScript-heavy pages
- Map fields from source applications to the Search AI schema for consistency and relevance
- Use post-processor scripts to update content received from source applications
- Test and save field mappings before applying them for ingestion
Advanced Content Extraction and Processing
Flexible Extraction Framework Search AI replaces SearchAssist’s static approach with a customizable extraction strategy framework that adapts to different document types:- Image-based Document Extraction: Uses VDR-based embeddings to capture semantic meaning from text and images, improving accuracy for complex documents with nontextual layouts (tables, forms).
- Markdown Extraction: Transforms each page into structured Markdown before processing, preserving semantic structure.
- Layout-Aware Extraction: Combines OCR, layout detection models, and layout awareness rules to comprehend tables, graphs, and charts.
- Advanced HTML Extraction: Intelligent processing of images and tables from HTML documents.
- Custom Extraction: Allows enterprises to insert custom extraction logic into the search data pipeline.
- Support for E5 and VDR (Image Embeddings) models for superior semantic understanding across text and image data
- Custom embedding support without requiring middleware
Gen AI Capabilities
- Custom LLM support for answer generation without middleware
- Dynamic Models and Prompts for answers: Runtime selection of prompts and LLM models in the answer generation API, enabling dynamic response tailoring based on query intent, user context, or application requirements
Developer Experience and Tools
Centralized Development Toolkit A comprehensive developer toolkit consolidates all platform resources in one location:- Unified Resource Hub: Centralized access to all development tools and utilities
- Extraction Utilities: Ready-to-use tools for various extraction needs
- Evaluation Frameworks: Ready-to-use tools for testing and validation
- SDKs and APIs: Complete development resources and documentation
- Streamlined Discovery: Easier access to platform utilities and capabilities
Other Enhancements
- Auto Training: Automatically detects new or updated content and retrains the system to keep indexed content current.
- Greater Configuration Flexibility: More configurable answer generation properties than the standalone SearchAssist application.
Integration with AI for Service Platform
Search AI is integrated into the XO platform as part of the AI for Service suite, combining enterprise-grade search, automation, and assistance in a unified ecosystem. Key benefits:- Comprehensive user and role management with precise access controls
- Centralized admin console for configuration and monitoring
- Built-in guardrails for responsible and secure AI deployment at scale
- Curated model and prompt library for faster adoption across diverse use cases
- Gen AI usage logs for enhanced visibility and governance
- Support for multiple JWT app creation for scalable, secure enterprise integrations
Feature Realignment
With the shift from SearchAssist to Search AI on the AI for Service platform, features have been restructured to prioritize performance, user experience, and adaptability. See the detailed feature comparison here.Contact Center AI
- Integrated Experience: Combines Contact Center AI (CCAI), Automation AI, Search AI, and Agent AI in a single app.
- Additional Capabilities: Enhanced Contact Center with Campaign Management, Quality Management, and Case Management as additional top-ups.
- Enhanced Dialog Tasks and Flows building experience.
- Enhanced user experience for agents and supervisors.
- Improved conversational experience using Generative AI capabilities.
- Access to Agent AI and Search AI based on additional licenses.
Agent AI
- Access to Search AI based on additional licenses.
- Enhanced user experience for agents and supervisors.
- Improved conversational experience using Generative AI capabilities:
- Auto Summarization: GenAI-powered summarization with configurable models through the Prompt Library.
- GenAI Node-Powered Dialog Tasks: Uses commercial models to power dialogs and workflows in Agent AI.
- Sentiment Analysis: Choose between XO or commercial LLMs to detect and display customer sentiment to agents in real time.
- Answers through Knowledge AI: Choose commercial models or bring your own through a custom LLM endpoint.
- Playbooks and Coaching Design Time: Generate similar utterances powered by the commercial OpenAI model to improve Playbooks and Realtime Coaching creation.
- GenAI-Powered Runtime Coaching Hints: Uses commercial LLM models for real-time coaching hints.
- Enhanced Collaboration: New experience for building Dialog Tasks and Flows.
- Comprehensive Analytics: Detailed insights into user conversations categorized by each product.
- Integrated Experience: Combines Contact Center AI (CCAI), Automation AI, Search AI, and Agent AI in a single app.