Trust & safety
Plain answers about the AI, your data, and how to reach a human. No fine print tricks - the fine print says the same thing this page does.
The messages are AI-drafted. Never AI-sent.
After a pilot business and its live-model gate are activated, bounded conversation context and approved operating-pack facts may be sent to the configured model to prepare a suggested reply. A person at the business reads the draft, edits it if needed, and decides whether to approve it. The model does not set prices or create commitments.
This is the Release 1 product rule. There is no setting that turns on autonomous business sending. Narrow carrier-required consent or opt-out confirmations may be automatic and are recorded.
What the AI is allowed to do
- Draft replies for the business owner to review.
- Summarize long supported message threads. Voicemail recording and transcription remain off until their separate gate is approved.
- Use bounded, approved conversation and operating-pack context to propose a customer-care reply.
- Flag urgent-sounding messages - anything that reads like an electrical hazard gets pushed to the top of the owner's list with a prompt to call you.
What the AI is not allowed to do
- Send you anything without a person's approval.
- Set or change prices, or promise dates, warranties, or permit approvals.
- Direct the business to deny AI assistance. The business and pilot support must answer AI-use questions accurately.
- Give electrical safety advice. Urgent language is routed for human review; messaging is not an emergency service.
If a message ever seems dangerous or wrong
If you smell burning, see sparks, or hear buzzing at your panel - stop texting and call 911 or your utility's emergency line. A text thread is never the right tool for an emergency, with or without AI.
For anything less urgent that still looks wrong - a price that doesn't match what you agreed, a wrong address, or a message that seems out of character - reply in the thread or call the business. If you think the problem is on our side, email support@ownerswitch.com. Monitoring, escalation, and the tenant-scoped draft/approval audit path must be verified before a pilot business is activated.
Your data, shortly
- Your messages belong to your conversation with the business. We process them to provide the service to that business - not to build advertising profiles, and never to sell.
- Tenant isolation is release-blocking. Conversation access is tenant-scoped, cross-tenant negative tests are required, and any known exposure stops the pilot.
- Your mobile opt-in is never shared or sold to third parties for their marketing. Full detail in the privacy policy.
- Payments are outside Release 1. The pilot does not create invoices, payment links, or payment reminders.
- You can ask what we have and ask us to delete it - see "Your rights" in the privacy policy or email privacy@ownerswitch.com.
Escalation, in order
- The business itself - reply in the thread or call the same number. Fastest for anything about your project.
- OwnerSwitch pilot support - support@ownerswitch.com or the contact form. Monitoring and escalation are verified before any business is activated.
- Privacy or consent issues - privacy@ownerswitch.com. Opt-out problems get fixed first and investigated second.
- Security reports - security@ownerswitch.com. Don't email passwords, API keys, or live customer data.