1. Our Approach to AI
We build our business on AI voice, chat, and WhatsApp agents, so how we talk about AI matters as much as how we build it. This policy explains, in plain language, when you're talking to AI rather than a person, what its limits are, and where a human should be involved.
2. When You May Be Interacting With AI
You may be interacting with an AI agent when: you use the live chat assistant or voice demo on our website; you call or message a business that has deployed an AI voice or WhatsApp agent built by us; or you interact with a chatbot we built for a client's website. In each case, the interaction is designed to be identifiable as automated, though the specific disclosure wording is configured per deployment.
3. AI Voice Agents
Our AI voice agents hold real-time spoken conversations to answer questions, qualify leads, and book appointments. They are built on third-party speech and language models and are designed to identify themselves as an automated assistant near the start of a call, though the exact wording is set by the deploying business.
4. AI Chat and WhatsApp Agents
Our chat and WhatsApp agents respond to typed messages using large language models, generating replies based on the business's configured knowledge (FAQs, pricing, policies) plus general model knowledge. They are designed to hand off to a human when a request falls outside what they're configured to handle.
5. Automated Decision Support
Our agents can suggest next steps — such as a recommended appointment slot or a qualification score for a lead — but these are decision-support suggestions for the deploying business to act on, not final decisions made autonomously about a person's rights, eligibility, or access to a service.
6. Human Oversight
Our agents are designed with escalation paths: when a conversation involves a request the agent can't confidently handle, an emergency-sounding situation, or an explicit request to speak to a person, the agent should direct the person to a human contact point configured by the business. We recommend, and where feasible build in, human review of agent conversations, especially in the period after go-live.
7. Accuracy and Hallucinations
Large language models can produce confident-sounding but incorrect answers ("hallucinations"), misunderstand context, or mishandle unusual phrasing — especially in noisy audio or ambiguous typed messages. Treat AI responses as assistive rather than authoritative, and verify anything important through a human channel.
8. AI Disclosure
Where legally required or where we control the deployment, we aim to disclose that a conversation involves AI. For client deployments, the specific disclosure language and placement is a configuration choice made with that client, who is responsible for complying with any AI-disclosure laws applicable to their business and location.
9. Voice Recording and Transcription
Voice conversations may be transcribed or recorded to generate responses and, where enabled, to produce call summaries or logs. Whether a specific deployment stores full recordings, and for how long, is a configuration decision for that business. Call-recording consent laws vary by country and, in some countries, by state — deploying businesses are responsible for configuring appropriate notices.
10. Synthetic Voice Disclosure
Our AI voice agents use synthetic (AI-generated) speech, not a recording of a real person impersonating staff. We do not design our agents to claim to be a specific named human employee when asked directly whether they are AI.
11. Data Used by AI Systems
Our agents are configured using the business information a client provides (FAQs, pricing, policies, scripts) plus the general knowledge of the underlying model. See our Privacy Policy for how conversation data itself is handled.
12. Third-Party AI Providers
Our own website's chatbot and voice demo run on Google Cloud Vertex AI (Gemini models). Client deployments may use different underlying providers depending on the project. Marketing references to providers such as OpenAI or Anthropic describe technologies we can build with for a client engagement, not a claim that every deployment — including this website — uses all of them simultaneously.
13. High-Impact Decisions
Our agents are built for lead qualification, appointment booking, and customer support — not for making final decisions in medical diagnosis, legal advice, credit or lending decisions, insurance underwriting, employment decisions, or similarly high-impact, tightly regulated contexts. Deployments in sensitive sectors (e.g. clinics, insurance) are scoped to administrative and informational tasks (like booking or FAQs), with clinical, legal, or underwriting judgment left to qualified humans.
14. Prohibited AI Uses
Our AI services may not be used to impersonate real individuals without consent, generate deceptive synthetic media, make unlawful automated decisions about people, harass or deceive consumers, or bypass legally required consent for calls or messaging. See our Acceptable Use Policy for the full list.
15. Bias and Fairness
Language and voice models can reflect biases present in their training data or produce uneven performance across accents, dialects, and languages. We test our agents against representative conversation scenarios before go-live and welcome reports of biased or unfair behavior so we can investigate and address them.
16. Security and Abuse Prevention
We apply reasonable safeguards to reduce misuse of our AI agents, such as rate limiting and monitoring for abusive input patterns (e.g. prompt injection attempts). No safeguard is perfect, and we continue to refine these controls as new risks emerge.
17. Reporting an AI Concern
If an AI agent — ours or one built for a client — behaved in a way that seems unsafe, biased, deceptive, or otherwise concerning, please tell us at with as much detail as you can (what was said, when, and where). We take these reports seriously and will investigate.
18. Policy Updates
We'll update this policy as our AI practices evolve and update the "Last updated" date above whenever we do.