AI Transparency Statement

1. Introduction

Hello publishes this AI Transparency Statement to demonstrate our commitment to responsible AI practices in healthcare communications. Transparency is a foundational principle of how we build, deploy, and operate AI voice agents for healthcare practices.

While this statement draws from principles outlined in the EU AI Act and the NIST AI Risk Management Framework, Hello proactively adopts these transparency measures across all markets we serve. Healthcare AI demands the highest standard of trust, and we believe that clear disclosure of how our systems work is essential to earning and maintaining that trust.

2. Provider Identity

• Organization: Hello, Inc.
• Registered: State of Delaware
• Operations: Denver, CO, United States
• Website: https://usehello.ai
• Contact: hello@usehello.ai

3. Nature and Purpose of the AI System

Hello provides AI-powered voice agents that automate inbound and outbound telephone calls for healthcare practices. The AI system performs the following tasks on behalf of our Clients:

• Appointment scheduling, confirmations, and rescheduling
• Patient intake and information collection
• Deposit and payment collection
• Frequently asked question handling
• Call routing and transfer to appropriate staff members
• Appointment reminders and follow-up calls
• Lead reactivation and patient recall campaigns

Purpose: The system is designed to reduce missed calls, improve patient access to care, free healthcare staff from repetitive phone tasks, and provide 24/7 availability including after-hours and weekends.

4. AI System Functionality

Hello is powered by a multi-provider architecture that combines several AI and telecommunications technologies to deliver natural, reliable voice interactions:

• Real-time speech-to-text (STT): Converts caller speech into text for the AI to understand and process
• Large language models (LLMs): Power natural conversation, intent recognition, and decision-making within the scope of the agent's configured tasks
• Text-to-speech (TTS): Generates natural-sounding voice responses to callers
• Integration layer: Connects the AI agent to EHR/PMS systems, scheduling platforms, and payment processors to take real-time actions
• Sentiment detection and confidence scoring: Monitors caller sentiment and the AI's own confidence level throughout each conversation
• Automatic escalation: Routes calls to human staff when the AI detects confusion, low confidence, or caller distress
• Bilingual support: Supports conversations in English and Spanish
• Multi-provider redundancy: Maintains automatic failover across telephony, AI, and infrastructure providers to ensure continuous availability. See our Business Continuity and SLA page for details.

5. Human Interaction Disclosure

Callers are informed that they are speaking with an AI assistant at the beginning of every interaction. This disclosure is a core design principle of the Hello platform.

This AI disclosure is enabled by default on all deployments. Practices may customize the greeting language and style, but cannot remove the AI disclosure. Callers can request to speak with a human staff member at any point during the conversation.

6. Data Categories and Processing

The following categories of data are processed by Hello AI voice agents during the course of providing the Service:

Data is processed solely for the purpose of providing the Service to the specific Client on whose behalf the AI voice agent operates. Hello does not use individual patient data to train AI models. For full details on data collection, use, and retention, see our Privacy Policy.

7. Sub-processors and AI Providers

Hello uses a multi-provider architecture to deliver the Service. Current provider categories include:

• AI and LLM providers: Large language model providers that power conversation intelligence and natural language understanding
• Speech-to-text and text-to-speech providers: Convert spoken language to text and generate natural voice responses
• Telephony providers: Carrier-grade SIP providers that handle call connectivity and routing
• Cloud infrastructure providers: Enterprise-grade hosting and data storage with HIPAA-compliant environments

The full and current list of sub-processors is maintained at Data Processing Agreement. All sub-processors that handle Protected Health Information (PHI) are required to execute Business Associate Agreements with Hello.

8. Model Limitations and Risk Information

Hello is committed to transparency about the limitations of AI technology. Users and callers should be aware of the following:

8.1 Known Limitations

• AI responses may occasionally be inaccurate or incomplete
• The system cannot provide medical advice, diagnoses, or treatment recommendations
• Unexpected behavior may occur with significant background noise, heavy accents, or ambiguous requests
• The AI may misinterpret caller intent in rare cases
• Voice synthesis may not perfectly replicate human speech patterns

8.2 Mitigations

Hello implements multiple layers of mitigation to reduce the impact of these limitations:

• Multi-model architecture with fallback: If the primary AI model fails or returns low-confidence results, the system automatically routes to a secondary model or escalates to human staff
• Confidence scoring: Every AI response is assigned a confidence score. Low-confidence responses trigger review or escalation
• Automatic escalation: Calls are transferred to human staff when the AI detects confusion, repeated misunderstanding, or caller frustration
• Human handoff: Callers can request a human at any point during a conversation
• Regular quality audits: Hello conducts ongoing reviews of call recordings and transcripts to identify and correct issues
• Configurable action restrictions: Practices define exactly what the AI agent can and cannot do, limiting exposure to unintended actions

9. AI Safety Guardrails

Hello AI voice agents operate within strict safety boundaries. For detailed technical information, see AI Safety and Guardrails on our Security page.

9.1 Conversation Controls

• Max call duration limits: Configurable per practice to prevent calls from running indefinitely
• Turn limits: Maximum conversation turns prevent infinite loops
• Confusion detection: Low-confidence streaks across consecutive turns trigger automatic transfer to human staff
• Topic boundaries: Agents are restricted to the topics and tasks configured for each practice and cannot deviate outside those boundaries

9.2 Action Restrictions

Hello AI agents are configured with explicit action policies. The AI cannot:

• Modify billing records or financial information
• Access clinical notes or medical records beyond what is required for scheduling
• Make medical recommendations, provide diagnoses, or suggest treatments
• Share PHI with unauthorized parties
• Make autonomous decisions on clinical matters

9.3 Emergency Handling

If a caller indicates a medical emergency, the AI agent immediately routes the call to human staff or directs the caller to call 911. The system is trained to recognize emergency keywords and phrases and prioritize caller safety above all other tasks.

10. Consent and Compliance

Hello operates in accordance with the following regulations and frameworks:

• HIPAA (Health Insurance Portability and Accountability Act): Hello acts as a Business Associate and executes BAAs with all healthcare practice Clients
• Colorado Privacy Act (CPA): Hello operates from Denver, Colorado, and complies with CPA requirements for personal data processing
• California Consumer Privacy Act (CCPA/CPRA): Hello honors the data rights of California residents as described in our Privacy Policy
• NIST AI Risk Management Framework: Hello voluntarily adopts principles from the NIST AI RMF to guide responsible AI development and deployment

10.1 Outbound Call Consent

Outbound calls (appointment reminders, recall campaigns, reactivation) are placed only when the Client confirms that appropriate patient consent has been collected in accordance with applicable telecom regulations, including the Telephone Consumer Protection Act (TCPA) and state-specific consent laws.

10.2 Right to Human Contact

Callers can request to speak with a human staff member at any time during an AI-assisted call. The system is designed to honor this request immediately and transfer the call with full conversation context preserved.

11. Rights of Individuals

Individuals who interact with Hello AI systems have the following rights:

• Request a human: Speak with a human staff member at any time during a call
• Access recordings: Request access to their call recordings and transcripts through the healthcare practice
• Request deletion: Request deletion of their data through the healthcare practice, subject to legal retention requirements
• Opt out of AI: Request human-only communication from the healthcare practice for future interactions

For data-related inquiries, individuals may contact privacy@usehello.ai or reach out to the healthcare practice directly. The healthcare practice is the data controller for patient data and is the primary point of contact for individual rights requests.

12. Updates and Versioning

This AI Transparency Statement is reviewed and updated at least annually, or whenever material changes are made to our AI systems, data processing practices, or applicable regulations. Material changes will be published at this URL with an updated version number and effective date.

For questions about this statement or our AI practices, contact us at hello@usehello.ai.