Trust & supervision
Supervised, audited, consent-first — by design
Most AI pitches ask for your trust. Ours is built so you don't have to give it. Here is exactly how a Sustainable Agent is kept on a leash — the same guarantees, spelled out.
What runs on its own, and what waits for you
Routine, reversible work — moving data, drafting, reconciling, preparing reports — runs on its own once you switch it on. Anything with consequences does not. We sort every action into three lanes:
- Green — automatic. Reversible, everyday work. It just happens, and it is logged.
- Yellow — approve once, then batched. A repeated action like a message template: you approve the pattern, it is authorised for a set window, and you can revoke it any time.
- Red — always your call. Spending money, sending a contract, provisioning something new, or contacting someone for the first time. Every single time. Un-skippable.
And a single kill switch halts everything at once — every agent, every pending action — the moment you want it stopped.
The guarantees
Four things that are always true
Approval gates
Spending money, sending contracts, and contacting someone new are impossible without your one-tap approval. Built into the system, not left to the model's judgement.
Dry-run first
Every deployment starts in drafts-only mode. You read its work, then you flip the switch — never us.
Tamper-evident log
Every action is written to a record that can't be edited after the fact. You can always answer 'what did it do, and when?'
Consent-first, PDPL
No one is messaged without a recorded lawful basis under Gulf PDPL, and opt-outs are honoured within 24 hours.
Your data, and Gulf PDPL
We keep only what a workflow needs, for as long as the law allows, and we message people consent-first: no contact without a recorded lawful basis, opt-outs honoured within 24 hours, enforced by the software rather than left to good intentions. For data that can't leave your building, the Sovereign option runs the whole worker on hardware you own.
Questions
Questions we get
- Where does our data live?
- In an instance we run for you, or — on the Sovereign tier — entirely on hardware you own, where it never leaves your building. Either way, we keep only what a workflow needs, for as long as PDPL allows.
- What is PDPL, and how do you comply?
- The Gulf's Personal Data Protection Laws govern how businesses handle personal data. We only message people with a recorded lawful basis, honour opt-outs within 24 hours, and enforce it in the software rather than in a policy nobody reads.
- Can the AI go rogue and spend money?
- No. Spending, sending contracts, and contacting new people are impossible without your one-tap approval, and one kill switch halts everything. The limits are structural, not a matter of trusting the model to behave.
- How do I know what it actually did?
- Every action is written to a tamper-evident log you can read at any time. Nothing an agent does is invisible to you.