Freedom is earned, and lost after any slip
Four levels of trust
- 1
Observe
The AI watches the work and records what it would have done. It changes nothing.
- 2
Recommend
The AI proposes each step. A person decides and acts.
- 3
Supervised
The AI acts within set limits. A person approves every action before it takes effect.
- 4
Independent
The AI acts alone inside a defined scope, with results checked after the fact.
Who approves what
- Advice is not authority
- AI specialists can recommend. Only a person with the right role can approve.
- One approval, one action
- An approval covers one exact action, amount and deadline. It never stretches to cover the next one.
- Separate hands
- Whoever proposes work never approves it, and whoever does the work never accepts it.
- An emergency stop
- Any authorized person can stop all work in progress at once. The stop is rehearsed, not assumed.
Safe when things go wrong
- Unknown is not failed
- If we are unsure whether an action happened, we check before retrying, or stop for review.
- Rehearsed, not assumed
- Data separation, outages, recovery and the emergency stop are tested and rehearsed.
- Authority does not drift
- If plans change, we replan the remaining steps. We never quietly lower the bar or widen anyone's authority.
- Rules stay put
- Changes go through test, trial and rollout, with rollback. AI cannot rewrite its own rules.
Your data, under the rules your industry follows
- It stays in your workspace
- Job leads work inside the workspace on managed office machines. Files are not downloaded, printed or copied to personal devices.
- Health data
- For medical billing, coding and prior authorization, we sign a HIPAA business associate agreement before we see any patient record.
- Tax return data
- In the US, the taxpayer's written consent is required before return information is disclosed outside the country, and we do not start tax work without it. Other countries have their own rules, and we follow them.
- Financial data
- Bookkeeping, lending and compliance work follows your safeguards program, and access is limited to the people on your jobs.
We are pre-launch and hold no security certifications yet. Before you send us data, we will tell you which audits we have completed and which we have not.
Tested before release
Expert services
An AI-backed rep resolves a medical billing dispute, a licensed reviewer signs off and the payer confirms.
Logistics
Resolve a delivery delay within policy, update the transport system and confirm the delivery.
Fintech
Reconcile a payment exception, get the required approval and confirm the ledger.
Creative delivery
Build a portal and campaign assets, pass acceptance checks, approve and go live.
Have questions about controls?
We will walk your risk and compliance team through approvals, evidence and the emergency stop.