CX Genie

Deploys a pre-trained conversational AI agent that handles customer inquiries across business hours without human intervention. The platform uses a template-based configuration model where businesses define common question-answer pairs and conversation flows through a visual builder or simple JSON schema, then the chatbot automatically routes incoming messages through intent classification and response matching. The system maintains conversation context within a single session to handle multi-turn dialogues without requiring explicit state management from the user.

Analyzes incoming customer messages to identify the underlying intent (e.g., 'order status inquiry', 'refund request', 'product question') and routes them to the appropriate response handler or escalation path. The system uses semantic similarity matching or lightweight NLU models to compare incoming text against a knowledge base of known intents, returning a confidence score that indicates whether the chatbot should respond autonomously or escalate to a human agent. Routing decisions are configurable — businesses can set confidence thresholds to automatically escalate low-confidence matches.

Exposes REST API endpoints that allow developers to send messages to the chatbot, retrieve conversation history, and manage Q&A training data programmatically. The API supports standard HTTP methods (POST for sending messages, GET for retrieving data, PUT for updating) and returns JSON responses with conversation metadata, intent classification results, and generated responses. This enables custom integrations beyond the platform's built-in channels (e.g., embedding the chatbot in a mobile app, integrating with a custom CRM).

Accepts customer-provided documentation, FAQs, or product information in multiple formats (text, PDF, web URLs) and indexes them into a searchable knowledge base that the chatbot queries to generate contextually relevant responses. The system converts documents into embeddings (vector representations) and stores them in a vector database, enabling semantic search — when a customer asks a question, the chatbot retrieves the most relevant knowledge base articles based on semantic similarity rather than keyword matching. Retrieved articles are then used as context for the LLM to generate a natural language response.

Maintains conversation state across multiple message exchanges within a single customer session, allowing the chatbot to reference previous messages and build context-aware responses. The system stores conversation history (messages, intents, responses) in a session store keyed by customer identifier, and passes relevant history to the LLM as context when generating responses. This enables the chatbot to handle follow-up questions like 'Can you tell me more?' or 'What about the other option?' without requiring the customer to repeat themselves.

Detects when a customer inquiry exceeds the chatbot's capabilities (based on confidence thresholds, explicit escalation keywords, or customer request) and seamlessly transfers the conversation to a human agent with full context. The system passes the conversation history, customer information, and detected intent to the agent interface, eliminating the need for customers to repeat themselves. Escalation can be triggered automatically (low confidence) or manually (customer requests to speak with a human).

Tracks and visualizes chatbot performance metrics including conversation volume, resolution rate (conversations resolved without escalation), average response time, customer satisfaction (if feedback is collected), and intent distribution. The platform aggregates conversation logs into a dashboard showing trends over time, identifying which intents the chatbot handles well vs. poorly, and highlighting conversations that failed or were escalated. Metrics are updated in near-real-time and can be exported for further analysis.

Accepts customer messages from multiple communication channels (web chat widget, email, SMS) and routes them through a unified chatbot pipeline, allowing businesses to handle inquiries across channels without deploying separate chatbots. The platform provides channel-specific integrations that normalize messages into a standard format, maintain channel-specific context (e.g., SMS character limits), and route responses back through the appropriate channel. A single conversation may span multiple channels (e.g., customer starts on web chat, continues via email).

Provides a UI for businesses to create, edit, and organize Q&A pairs that train the chatbot's response generation. Users can manually write Q&A pairs, import them from CSV/JSON, or extract them from existing knowledge base documents. The platform stores Q&A pairs in a structured format, allows tagging and categorization, and enables version control (tracking changes over time). When the chatbot receives a customer message, it retrieves the most relevant Q&A pair(s) and uses them as context for response generation.

Allows businesses to define response templates with placeholders for dynamic data (customer name, order number, product name, etc.) that are filled in at response time. Templates support conditional logic (e.g., 'if customer is premium, show premium support hours') and formatting options (bold, links, line breaks). When the chatbot generates a response, it selects the appropriate template and substitutes variables with actual customer data, enabling personalized responses without requiring custom code.

Collects explicit customer feedback (ratings, satisfaction surveys) after conversations and performs sentiment analysis on conversation text to infer customer satisfaction. The platform prompts customers to rate their experience ('Was this helpful?', 'Rate your satisfaction 1-5'), stores feedback alongside conversation data, and analyzes conversation text for sentiment indicators (positive, negative, neutral). Feedback is aggregated into metrics and used to identify problematic conversations for review.