Approvals and control
AI prepares the work. A person decides.
Every outward-facing action in Aiderra OS stops at an approval. Not as a preference you can forget to set, but as part of how each workflow is built. AI should add capability without removing accountability, and in a care organisation accountability is the whole job.
Try it
Three buttons. One rule.
This is the real approval card, with a real draft from the Finance department. Approve sends it. Send back returns it with your note. Reject stops it. Two of those need a reason, and the reason is kept.
You see exactly what will go
The draft, the recipient, the department that raised it, and the urgency it was given.
Nothing waits forever
A decision left too long escalates to you again rather than expiring quietly.
Send “Invoice 0231, a gentle reminder” to the Meadowbrook House account
To accounts@meadowbrookhouse.co.uk · From Elm Row Care · Reply-To you
Hello,
A gentle reminder that invoice 0231, for August’s respite care, fell due on the 19th.
If it has already gone, please ignore this, and thank you.
With thanks, Elm Row Care
The approval queue
One queue for everything outward-facing.
Emails, call lists, invoices, published pages, social posts. Whatever department raises it, it lands in the same place, with the same three buttons, and shows up on your phone.
- Every pending decision in one list, most urgent first
- The department and the workflow step that raised it
- A call list is approved before anyone is dialled, and screened before it reaches you
- Snooze what can wait, without losing it
The audit trail
Who decided, when, and what changed.
Every step of a decision is written to an append-only log: what the AI proposed, what a person changed, who approved it, when, and what was finally sent. Each entry carries a hash of the one before, so the record cannot be quietly edited.
- What the AI proposed, verbatim
- The note a person left when sending it back
- The revised draft, with the difference recorded
- The final outcome, with the provider's delivery confirmation
7f3a…c91e
b02d…41aa
e6c0…9d27
1a9f…70b3
c4e8…2f05
Earning autonomy
Trust is granted, not assumed.
At first, everything comes to you. As a department proves itself, you can let routine work run without a stop. You can take that back at any time, and every action still goes in the trail.
Everything is approved
A new workspace starts here. Every outward action waits for a person. This is the right setting for the first weeks.
Routine work runs, the rest waits
You decide which actions are routine. Those run on their own and are recorded. Anything new, unusual or high-stakes still waits.
You can always pull back
Pause an agent, a department or everything from one screen. The next action stops. The trail shows what happened up to that moment.
Why it matters here
A care organisation cannot outsource accountability.
The registered manager still signs
An AI can draft a reply to a relative or a reminder to a commissioner. The decision to send it stays with the person who is accountable for it.
Evidence, not assurance
When an inspector, a commissioner or a family asks what was said and who decided, the answer is in the trail, with a time and a name.
An off switch that actually stops things
Pausing is immediate and complete. There is no queue of half-sent messages and no "it was already on its way".
Try the queue with your own decisions in it.
Fourteen days free. Nothing is sent without you from the first minute.