AI Policy

How HostTalk Uses Artificial Intelligence

HostTalk uses AI to answer guest calls on behalf of property managers. This page explains how that AI works, what it can and cannot do, and how we protect everyone involved.

Informational

Purpose of AI outputs

Human-in-Loop

You stay in control

PIPEDA + GDPR

Privacy compliance

The Technology

What HostTalk's AI Actually Does

HostTalk uses large language model AI (provided by Anthropic) and voice synthesis technology (provided by ElevenLabs) to generate spoken responses to inbound guest calls. When a guest calls a HostTalk property number, the AI retrieves the property-specific information entered by the property manager and formulates a natural-language response based solely on that information. HostTalk's AI does not browse the internet, access external databases, or make assumptions beyond the information your property manager has provided. It is a bounded, information-retrieval system — not a general-purpose AI assistant.

Informational Only

All AI-generated responses are informational in nature. HostTalk does not provide legal advice, safety assessments, medical guidance, or emergency services. For emergencies, guests are directed to contact the property manager or emergency services directly.

Bounded by Your Data

The AI only answers based on the property information you provide. It does not guess, fabricate, or source answers from outside your knowledge base. If information is missing, it tells the guest and offers to take a message.

Human Override Always Available

Property managers retain full control at all times. You can update, restrict, or pause AI responses at any moment from your dashboard. HostTalk supplements your hospitality — it does not replace your judgment.

No Training on Your Data

Your property information and guest call transcripts are never used to train AI models. Your data is yours. We process it to deliver the service and for nothing else.

Limitations

What HostTalk AI Cannot Do

HostTalk's AI is designed for a specific, bounded use case. It is not suitable for and will not attempt to: provide emergency response guidance beyond directing callers to appropriate services; make representations about safety, legal compliance, or contractual obligations; handle financial transactions or collect payment information; operate as a general-purpose assistant outside of the property context it has been configured for. Property managers are solely responsible for the accuracy and completeness of the information they provide to HostTalk. The quality of AI responses depends directly on the quality of your property knowledge base.

Compliance

Our Responsible AI Commitments

OECD AI Principles

We align with OECD principles of transparency, human oversight, and accountability. AI systems are designed to augment property manager capabilities without replacing their judgment.

NIST AI Risk Framework

We follow NIST AI RMF concepts including risk identification, guardrail implementation, and continuous review of AI behavior and outputs.

EU AI Act (Limited Risk)

HostTalk is designed to operate within limited-risk and minimal-risk categories. We provide transparency notices and avoid automated decision-making in high-risk contexts.

PIPEDA Compliance

All call data, transcripts, and property information processed by HostTalk's AI is handled in compliance with Canada's Personal Information Protection and Electronic Documents Act.

Subprocessor Transparency

We use Anthropic (AI), ElevenLabs (voice synthesis), Twilio (telephony), and Supabase (data storage) as subprocessors. Each is bound by data processing agreements aligned with our privacy standards.

Continuous Monitoring

Our AI prompts, guardrails, and response quality are reviewed on an ongoing basis. We update our systems as regulations evolve and as we learn from real-world usage.

Questions About Our AI?

HostTalk

Handle guest calls automatically

Get product updates, tips for property managers, and HostTalk news — no spam.