Security & data handling

What happens to your agency's data, stated plainly.

You are about to put your book, your producers and your leads' phone numbers into someone else's software. Here is how it's separated, who owns what, what the AI is prevented from doing — and, at the bottom, the things we don't claim.

Your data is separated from every other agency's

Isolation is enforced in the database itself, not in application code. Every table carrying agency data has row-level security tied to the agency that owns the row, so a query that forgets to filter returns nothing rather than someone else's book. Platform-level access is a separate tier from agency admin: an agency owner is an administrator of their tenant and has no route to another one.

That distinction matters more than it sounds. The most common way multi-tenant software leaks is a missing filter in one query out of a thousand. Putting the boundary underneath the application means the missing filter fails closed.

You own the parts that are expensive to lose

  • Your carrier account and your numbers. You bring your own Telnyx, SignalHouse or Twilio account. The numbers, the 10DLC brand and the carrier relationship are yours, and they stay yours the day you leave. Most platforms rent you your own phone numbers.
  • Your conversations. Your threads are not pooled into a shared model. Feedback from your team retrains your ranking, not a general one sold back to your competitors.
  • Your records. Export and deletion are available on request through support, and the account-deletion path is documented on the support page and in the privacy policy.

What the AI is not allowed to do

Guardrails matter more than capability when the thing being automated is a regulated sale.

  • It identifies as an assistant on a named producer's team. It does not impersonate the producer, and a check runs before sending to stop it claiming to be a person.
  • Underwriting, coverage and eligibility questions are escalated to a licensed human rather than answered.
  • Messages can be sent under the specific business name your agency is licensed as in a given state.
  • Every outbound message passes a validator and a repetition guard before it leaves.
  • You can pause the AI on one conversation or all of them, and every background agent has its own kill switch and can be run in flag-only mode.
  • New behaviour arrives as a proposal a human approves, not a change that happens to you.

Messaging compliance is enforced, not documented

  • STOP is honoured instantly and everywhere — including cancelling an appointment already on the calendar.
  • Quiet hours are enforced in the lead's own timezone, not your office's.
  • Opted-out and archived records fail closed on every send path, including manual sends.
  • Outbound texting stays hard-blocked until your own 10DLC brand is registered. That is a gate, not a warning.
  • The power dialer brakes itself at a 2.5% abandonment rate — deliberately under the 3% line, so a bad run throttles before it becomes a violation rather than after. The brake is tunable per agency but not removable, because the failure mode there is legal rather than operational.
  • Call recording is opt-in per session, and where the law requires it the lead hears a short notice before the conversation is recorded.

Access and accountability

  • Role-based permissions with per-user overrides, so a producer sees their book and a manager sees their downline.
  • Authentication and session handling are managed by the platform's identity provider; we never store passwords ourselves.
  • Traffic is encrypted in transit.
  • An audit log records administrative actions.

Handling protected health information

If your agency writes health lines and needs a HIPAA-covered configuration, that is available as an add-on rather than the default — which is how this works across the category.

The AI inference layer is already covered: an executed business-associate agreement with AWS, with the region whitelist enforced in code so it hard-fails rather than silently degrades. The database layer is the piece that gets upgraded, and we provision it for the agency that needs it.

It's an add-on because it carries a real recurring infrastructure cost. Folding that into the base price would mean every agency writing final expense and mortgage protection pays for coverage they will never use. If you handle PHI, raise it before you sign and we'll scope it with you — it's a configuration conversation, not a no.

What we do not claim

  • We are not SOC 2 certified. No audit has been performed. If a certificate is a hard requirement for you, we are not the right vendor today and we would rather say so now.
  • The standard configuration is not HIPAA-covered. Inference is; the database layer is not, unless you take the add-on described above. So do not assume coverage you haven't specifically arranged — and if anyone tells you we are "HIPAA compliant" out of the box, they're wrong.
  • No published penetration test. None has been commissioned.
  • No uptime SLA. We don't publish one, so don't assume one.
  • No performance statistics. No booking rate, reply rate or conversion figure appears anywhere on this site, because no defensible public denominator has been agreed.

If you're evaluating us for a business that handles protected health information, or you have a security questionnaire to run us through, send it to support@virtualcloser.com and you'll get a straight answer including where the answer is "not yet."

See the demo See the platform