At Imagicle, we believe Artificial Intelligence (AI) can be leveraged to power better, more inclusive customer conversations for all. We also recognize that by applying this technology, we have a responsibility to mitigate potential harm. That is why Imagicle develops its AI-powered features in accordance with its responsible AI principles [the “Framework”], which are based on the principles of Transparency, Fairness, Accountability, Privacy, Security, and Reliability. Imagicle translates these principles into product development requirements that form part of the product development lifecycle, alongside our Security by Design and Privacy by Design processes.
Accordingly, Imagicle AI Receptionist features that leverage AI are built with transparency, fairness, accountability, privacy, security, and reliability at their core. Each feature powered by AI undergoes an internal AI impact review — an assessment of how the technical underpinnings of the functionality measure against these principles.
Imagicle AI Receptionist AI features are built with these principles at the center of how Imagicle delivers the AI-powered technology. This Technical Note describes more information about Imagicle AI Receptionist AI features, and how Imagicle responsibly leverages AI to deliver the functionality.
Feature Overview
Natural Language Understanding
The Natural Language Understanding feature interprets what customers say or type, identifying their intent and extracting the relevant information from a voice or digital conversation. It allows the AI Receptionist to understand requests expressed in natural language across multiple languages, rather than requiring callers to navigate rigid menus, so each request can be directed to the appropriate automated flow.
Conversational Response Generation
The Conversational Response Generation feature uses large language models to produce natural, human-like replies to customer requests. Drawing on the organization’s configured knowledge base and the context of the conversation, the AI Receptionist answers questions and guides customers through self-service tasks around the clock. Responses are generated within the boundaries of the conversational flows the customer designs in Imagicle Smartflows.
Speech Recognition and Synthesis
For voice interactions, the Speech Recognition and Synthesis feature converts a caller’s speech into text for the AI Receptionist to process and converts the Receptionist’s replies back into natural-sounding speech. This enables real spoken conversations over the phone in place of traditional touch-tone menus, with support for multiple languages. Speech-to-text is provided by Deepgram and text-to-speech by ElevenLabs.
Task Automation and System Integration
The Task Automation and System Integration feature allows the AI Receptionist to perform actions on the customer’s behalf — such as scheduling appointments, creating tickets or orders, and looking up account information — by connecting to the organization’s CRM, ERP, and other back-end systems through Smartflows integrations. This lets the Receptionist resolve requests end to end rather than only providing information.
Human Agent Escalation
The Human Agent Escalation feature transfers a conversation to a live agent when a request is complex, sensitive, or outside the AI Receptionist’s configured scope, or when the customer asks to speak with a person. The conversation and its context are handed over through the Imagicle Omnichannel Agent Console so the human agent can continue the interaction seamlessly.
Model Overview
Introduction
Imagicle AI Receptionist AI features are powered by AI technologies that combine natural language understanding, large language models (LLM), and — for voice interactions — speech recognition and synthesis. Certain functionality relies on third-party AI services — Deepgram for speech-to-text (STT), ElevenLabs for text-to-speech (TTS), and OpenAI (ChatGPT) for large language model capabilities — orchestrated within a cloud environment configured by Imagicle using Imagicle Smartflows. For more information about the underlying third-party services, see the provider transparency documentation referenced below.
Model Architecture
Imagicle AI Receptionist leverages natural language understanding to detect customer intent, large language models provided by OpenAI (ChatGPT) to generate conversational responses grounded in the customer’s configured knowledge base, and — for voice interactions — speech-to-text models provided by Deepgram and text-to-speech models provided by ElevenLabs. These models are accessed through the respective providers’ APIs to generate the output, while the conversational logic is designed and orchestrated by the customer in Imagicle Smartflows. Imagicle does not develop the underlying foundation models; it configures and orchestrates them to deliver the AI Receptionist functionality.
Usage Guidelines
Imagicle AI Receptionist operates on live customer interactions across voice and digital channels and is configured by the customer using Imagicle Smartflows, a no-code visual flow editor. The features are intended to automate first-level customer interactions — answering questions, performing tasks, and routing or escalating to human agents — and are designed to hand a conversation over to a human agent for complex, sensitive, or out-of-scope requests rather than to make automated decisions about individuals without the option of human review.
Model Inputs and Outputs
Natural Language Understanding
A typical input is a customer utterance, spoken or typed. A typical output is a detected intent together with any entities extracted from the request (for example, a date, an account number, or a product name), used to drive the conversational flow.
Conversational Response Generation
A typical input is the customer’s request together with the conversation context and relevant knowledge-base content. A typical output is a generated natural-language response delivered to the customer by voice or text.
Speech Recognition and Synthesis
A typical input is the caller’s audio. A typical output is the transcribed text used by the AI Receptionist for processing and the synthesized speech delivered back to the caller.
Task Automation and System Integration
A typical input is a detected intent and the information extracted from the conversation. A typical output is an action executed in an integrated system — such as a created ticket, a placed order, or a booked appointment — and a confirmation returned to the customer.
Human Agent Escalation
A typical input is the conversation context together with the condition that triggers a handover, whether a customer request or flow logic. A typical output is the transfer of the interaction, with its context, to an available human agent.
Data Sources for Training and Evaluation
Imagicle does not use customer content to fine-tune or train the foundation models used to deliver AI Receptionist AI features. Customer conversations are processed solely to generate the requested output (intent recognition, response generation, speech recognition and synthesis, task automation, or escalation) for the customer.
Where AI Receptionist relies on a third-party AI service, each provider — Deepgram for speech-to-text, ElevenLabs for text-to-speech, and OpenAI (ChatGPT) for the large language models — represents that it does not use customer data to train or improve its foundation models, and does not retain customer data passed to the service beyond what is necessary to provide the functionality. For more information about each provider’s data, privacy, and security practices, see the provider documentation referenced below.
The underlying foundation models are trained by their respective providers on large datasets compiled from a wide range of publicly available written and spoken material. This training data encompasses a diverse array of topics, languages, and styles to help the models develop a broad understanding of language and its contexts. For further information, see the provider documentation referenced below.
Model Evaluation and Performance
Imagicle periodically assesses the AI features to maintain and improve both the performance and accuracy of the generated output. Humans participate in the review, testing, and quality assurance associated with the deployment of the underlying models. Intent recognition, generated responses, and other AI outputs are probabilistic and may contain errors; accordingly, the features are designed to support — not replace — human agents, with escalation to a human available whenever a request falls outside the Receptionist’s scope.
Safety and Ethical Considerations
AI models may produce inaccurate, incomplete, or unintended output if such patterns exist in the input or the model. To mitigate this, the third-party AI services used by AI Receptionist include content-filtering and safety controls, and Imagicle constrains how the models are used to the specific tasks of understanding customer requests, generating responses within customer-designed conversational flows, automating tasks, and escalating to human agents.
In addition to these technical mitigations, Imagicle has taken further steps to support safe and ethical use. Imagicle’s third-party AI vendors undergo vendor reviews that include security and safety assessments. Conversations handled by the AI Receptionist take place within the customer’s environment, and the associated transcripts and configuration are accessible only to authorized users within the customer’s organization. Please contact Imagicle if you have concerns or feedback about your experience with AI Receptionist.
Fairness
To understand how the foundation models underlying AI Receptionist are trained and evaluated for fairness, please reference the transparency documentation of the relevant AI service provider, referenced below. Imagicle supports customer interactions across multiple languages and reviews feature output to help identify and reduce systematic disparities in performance.
Privacy and Security
Imagicle AI Receptionist processes customer interactions across voice and digital channels, which may contain personal data. Customers remain the controllers of the interactions handled by the AI Receptionist and are responsible for obtaining any required consents and for configuring the conversational flows, data handling, and retention in line with applicable law and regulation. Where the customer configures AI Receptionist to use its own large language model, conversation content sent to that model is processed under the customer’s agreement with its LLM provider, and Imagicle’s privacy and security commitments do not extend to that processing.
Customer interactions are encrypted in transit and at rest, and access to conversation data and configuration is restricted to authorized users. Imagicle addresses the security of customer data in its security and data-processing documentation, available on request. Where the customer chooses to use their own LLM by providing their own API key, Imagicle is no longer responsible for the security and privacy of that third-party LLM. Responsibility for assessing the provider's security and privacy practices, and for the terms governing the processing of data submitted to that service, rests with the customer. Likewise, where the customer connects remote third-party MCP services to the AI Receptionist, Imagicle is not responsible for the processing of data performed by those services; responsibility for assessing each MCP provider's security and privacy practices, and for the terms governing that processing, rests with the customer.
Data Location
Imagicle AI Receptionist is delivered from the Imagicle Cloud. During onboarding, customers choose the location in which their service is hosted: the European Union or the United States. The EU location serves customers in Europe and the Middle East, and the US location serves customers in the United States.
Customer conversations, configuration, and associated data handled by the AI Receptionist are processed and stored in the selected location.
Where the customer connects remote third-party MCP services, the location in which data is processed by those services is determined by the relevant service provider and is outside Imagicle's control.
Updates and Maintenance
Ongoing Imagicle product updates and release notes describe product changes and improvements to the user experience. Imagicle may also use customer email notifications and its support and community resources to inform customers of changes to AI Receptionist AI features.