Rootlenses Voice combines artificial intelligence, automation, and telephony to execute voice conversations at scale.
This glossary brings together key concepts, both operational and technical, to help users understand how the platform works, how agents are configured, and how call-generated results are interpreted.
- Agent: Artificial intelligence entity configured to automatically manage and conduct phone conversations. An agent can perform tasks such as sales, collections, support, follow-ups, or information validation while following company-defined rules and objectives.
- LLM (Large Language Model): Artificial intelligence model trained to understand natural language and generate conversational responses. It is the engine that enables the agent to understand questions, interpret context, generate dynamic responses, and maintain natural conversations.
- Prompt: Instruction or set of rules sent to the AI model to define how the agent should behave during a call. A prompt may include communication tone, business objectives, restrictions, conversational flows, and objection-handling guidelines.
- System Prompt: Core configuration that defines the overall behavior of the AI agent. It acts as a foundational guide that establishes personality, limitations, priorities, conversational logic, and operational rules.
- CoT (Chain of Thought): Reasoning method used by the agent to structure decisions and respond step by step during a conversation. It allows the AI to analyze context, evaluate information, make more coherent conversational decisions, and maintain logical continuity.
- RAG (Retrieval-Augmented Generation): Architecture that combines artificial intelligence with real-time business information retrieval. Before responding, the agent can consult documents, policies, knowledge bases, or uploaded files to generate more accurate and contextualized answers.
- Knowledge Base: Repository of documents and business information used by the agent to answer questions or validate information during a call.
- Context Window: Maximum amount of information the model can remember and process simultaneously during a conversation. The larger the available context, the more coherent the conversation can remain and the better the agent can recall previous details.
- Token: Minimum processing unit used by AI models to interpret and generate text. Token consumption typically impacts operational costs, response speed, and processing capacity.
- Speech-to-Text (STT): Technology that converts a user’s voice into text so the AI model can interpret what was said during the call.
- Text-to-Speech (TTS): Technology that transforms AI-generated responses into natural audio to maintain fluid phone conversations.
- Latency: The time it takes for the system to process information and respond during a call. Low latency enables more natural conversations and reduces awkward pauses during interactions.
- Endpoint: Connection point used to integrate Rootlenses Voice with external systems, APIs, or enterprise services.
- API (Application Programming Interface): Set of rules and connections that allow Rootlenses Voice to integrate with systems such as CRMs, ERPs, support platforms, databases, or analytics tools.
- Webhook: Mechanism that automatically sends information from Rootlenses Voice to external systems whenever a specific event occurs, such as a completed call, a qualified lead, or a confirmed payment.
- ETL (Extract, Transform, Load): Process used to extract, transform, and load data from external systems into the platform to automate synchronization and information processing.
- CSV: File format used to import contacts, variables, or bulk data into campaigns.
- Campaign: Organized group of calls executed under the same operational configuration and business objective.
- Conversational Flow: Logical structure that defines how a conversation progresses depending on the user’s responses, decisions, or behaviors during the call.
- Intent: Objective or intention detected within a conversation. The system can identify intents such as purchase interest, support requests, payment intent, or cancellation.
- Sentiment Analysis: Automated analysis of the user’s emotional tone during a conversation. It helps detect signals such as satisfaction, frustration, urgency, or commercial interest.
- Lead Scoring: Automated classification system that evaluates the quality or conversion probability of a lead based on company-defined criteria.
- Conversion: Target action achieved during a call. This may represent a sale, a scheduled meeting, a confirmed payment, or any expected campaign outcome.
- Call Transfer: Process in which the AI agent transfers the conversation to a human when specialized attention or manual intervention is required.
- Escalation: Action of redirecting an interaction to another operational or human level due to complexity, risk, or the need for additional support.
- KPI (Key Performance Indicator): Key metric used to measure operational performance and campaign effectiveness. Examples include contact rate, conversion rate, average duration, cost per call, or successful resolution rate.
- Logs: Technical and operational records generated by the platform during call execution, integrations, and internal processes.