Human in charge
Defined work. Clear boundaries.
Auxilis workers gather information, follow approved pathways and complete defined operational actions. They do not independently make clinical decisions.
Where the human boundary sits
Some workers do hold clinical conversations — capturing symptoms, medication use and readings as part of an approved protocol. What they never do is interpret them. Anything calling for clinical judgement, along with defined safety indicators, patient distress, uncertainty or out-of-scope requests, is routed to the appropriate human team. Patients are told at the start of the conversation that they are speaking with an AI assistant.
The worker presents the picture. A healthcare professional decides what happens next.
- The worker gathers information
- Anything needing judgement is routed
- A healthcare professional decides
Every worker has
An explicit job
Each worker exists to perform one defined task for one defined patient population — nothing else.
Controlled permitted actions
The worker can only take actions that have been approved for its pathway, inside the systems agreed with the service.
Defined escalation conditions
Clinical judgement, defined safety indicators, patient distress, uncertainty and out-of-scope requests always leave the worker.
A named human route
Escalations go to an agreed person or team, with the context the worker has already gathered attached.
A complete interaction record
Conversations, actions, workflow versions and escalation behaviour are recorded and auditable.
Quality assurance
Every change is tested with real people before it reaches a patient
A worker’s behaviour must remain within its approved boundaries. Before any change goes live it is tested in simulation across thousands of scenarios, and separately with recruited participants who speak with the worker directly — so it is measured against the way people actually talk, not only against scripted cases. Each step is then reconstructed and assessed against defined safety rules.
Tested against real hazards
Our scenario set reflects the ways a patient conversation could go wrong, including missed concerns, inappropriate responses and failures to escalate. A change that weakens a safety boundary does not ship.
Measured, not spot-checked
Conversations can unfold differently each time. We therefore run scenarios repeatedly and assess performance at volume, rather than relying on a small number of successful examples.
Recorded for clinical sign-off
Every release produces a traceable record of what was tested, how it performed and whether it met the agreed safety criteria. That evidence supports review and approval by the Clinical Safety Officer.
Take routine patient contact off your team’s day.
See how an Auxilis worker could handle a defined patient workflow, complete approved actions and bring in your team only when they are needed.