Built so you're always in command.
The approval gate.
Oprativ's AI can read, prepare, and draft — but it cannot send an email, make a payment-affecting change, or touch the outside world on its own. Consequential actions stop at a human gate: you see exactly what's about to happen, and it happens only when you approve. This is enforced by the software's architecture, not by a policy we hope the AI follows.
The permanent record.
Every action — yours or an assistant's — lands in an append-only ledger: who, what, when, what it cost, whether you approved it. Printable. Exportable. Your audit trail and your operations cost-book in one.
Your data, your machine.
The product's knowledge of your business lives in files on your computer (or your own private hosted space) — not in a shared cloud. We built the company so we can't read your business data: our servers handle accounts, licences and payments, and nothing else.
Poison-proof by design.
Content the AI merely reads (a stranger's email, a web page) can never itself trigger an action — those paths are separated in the architecture, so a malicious message can't puppet your workforce. Outgoing messages that touch payment details are checked against your own records before anything sends.
Backups that answer to you.
Encrypted backup to a destination you choose — your drive, your cloud, your repository. A lost laptop is not a lost business.
Honest limits.
Oprativ tells you when it isn't sure instead of guessing; it stops at your spending allowance instead of billing you; and it shows its sources so you can check its work.
The boring essentials.
UK company, registered with the ICO as a data controller [ICO registration no. — pending]. Payments processed by Paddle (our merchant of record) — we never see or store your card details. Software updates are signed. Questions: hello@oprativ.ai — a human reads it.