Trust · Pre-launch

How we keep your data separate and your actions on record

Tenant isolation, human clearance for critical safety, and a reviewable trail of actions. Website learning reads only a small set of public HTTPS pages, and private or internal addresses are blocked. We do not claim ISO, SOC 2, bank-grade security, or any other certification we have not earned.

What this product is designed to control

Three controls we can describe from product design and source. We describe them as designed and in preview — never as certified or generally available.

Tenant isolation
Boundaries between organisations
Designed as a first-class trust boundary so one organisation's data and actions stay inside its tenant. Multi-tenant operations exist in product source and are in preview — not commercially generally available.
Human clearance
People decide consequential safety
A critical safety latch requires a person before consequential action. Fixed safety and lifecycle rules override AI. Escalation queues are designed for human review — not unsupervised autonomy.
Audit evidence
Actions should be reviewable
Work is designed to leave audit-oriented event trails and action evidence tied to the operation, so supervisors can see what happened — not only what a model suggested.

Fixed rules and an authorised person keep control of safety and money. Read the safety controls

Further design controls

These sit in product source as designed behaviour. Live operational verification remains on the release checklist — we do not treat source presence as a public certification. Local acceptance on 10 September 2026 included a full automated test pass, a closed local run, and a backup and restore check. Recorded build checks and security checks passed on the canonical application code. That is not an uptime percentage, latency guarantee, or SLA.

Fixed safety and lifecycle rules are designed to override AI suggestions.
Telephony request signatures (HTTP/WSS) are verified in source as a design control; live verification is still required before launch.
AI has no payment, refund, compensation, fee-waiver, or client-money execution authority.
The receptionist must not invent bookings, dispatches, refunds, or safety assurances.
Website learning reads only a small, bounded set of public HTTPS pages. DNS and redirects are re-checked, private/internal addresses are blocked, robots.txt is honoured, and missing facts are not invented.

What we do not claim

This page is a trust posture, not a certificate wall. Marketing must not outrun evidence.

No ISO 27001, SOC 2, Cyber Essentials, or other certifications — we have not earned them, and we will not invent seals, badges, or auditor names.
No bank-grade, military-grade, or enterprise-grade marketing fluff.
No fake uptime percentages, customer logos, awards, or named case studies.
Not commercially generally available. Nothing here is a production call-ready or public always-on service claim.
No published internal hosts, IPs, or live product dashboards on this marketing site.

Related: Features · How it works · AI receptionist

Privacy and legal

Marketing-site privacy, cookies, and terms are DRAFT FOR LEGAL REVIEW — scaffolds only, not final policy or legal advice. Product trust (this page) is separate from those drafts.

Privacy · Cookies · Terms(draft)

Talk through the controls

Product interest starts at Request an early demo. An early-access briefing is secondary. We will not present a fake certificate pack.

Request an early demo See the safety rules

What this product is designed to control

Tenant isolation

Human clearance

Audit evidence

Further design controls

What we do not claim

Privacy and legal

Talk through the controls