Agents that keep work moving across services.
Amelio agents retrieve information, update records and follow actions through, across the systems involved in someone’s care. Each works within defined permissions and returns to a person whenever judgement is required.
A defined job, with clear boundaries.
Connected systems become tools the agent can use to complete that job. It can check a record, request an update and make an authorised change, with every step recorded. If it reaches something outside its remit, it returns to the responsible person.
One date moves, and everything else has to follow.
When a planned discharge date changes, transport, medication, equipment and the care package may all need to move with it. Today, this depends on somebody noticing every dependency and contacting each service. An agent can follow those dependencies, update systems within its permissions and bring exceptions back to the responsible person while there is still time to act.
- No criteria to reside set to 21 AugAcute trust
- Reablement can start 24 AugReablement
First visit 09:00, answered in eleven minutes
- No transport available on 24 AugBlockedPatient transport
No morning capacity until 25 Aug
- Decided to ask reablement to moveDecision
Only 25 Aug works for both
- Reablement agreed 25 Aug, first visit 11:00Reablement
- Transport booked for the morning of 25 AugPatient transport
- Discharge plan updated and the ward toldCompletedAcute trust
Reason recorded: no transport on 24 Aug
Spot duplicated work before teams set off.
Two services can be scheduled to gather the same information or complete overlapping tasks because neither can see the other’s plan. An agent can flag the overlap, ask whether one visit can meet both needs, and update both schedules only after the teams and the person agree.
- Two teams scheduled to do overlapping workScheduling
Both are due to complete a mobility assessment at the same address on Tuesday
- The assessment only needs doing onceDecision
Neither team can see the other's plan
- Therapy confirmed its visit can cover bothTherapy team
Same clinician, no change to the session
- Reablement accepted the shared assessmentReablement
- Agreed to a single visitScheduling
- Duplicate visit removed, both teams toldCompletedScheduling
One journey saved, one assessment recorded once
Familiar controls, applied to a new kind of user.
An agent has its own identity, scoped permissions and audit trail. The data owner decides what it may read and change. The service defines its job, its stopping points and when it must return to a person. These are familiar controls, but each use case still requires appropriate information governance, operational assurance and clinical safety assessment.
Its own identity
An agent is registered in its own right and holds its own credentials. It never borrows a member of staff's login, so its actions are always distinguishable from theirs.
Permissions set by the data owner
The organisation that holds a record decides what an agent may read and change, down to the field. Not us, and not whoever happens to be running the agent.
Every action recorded
Each read, write and action is captured against the agent that did it, the task it was for and the time it happened, in the same audit trail as a person's activity.
Escalation when judgement is required
The service defines the points at which an agent must stop and hand the task back. Anything needing professional judgement returns to the responsible person.
Low-risk, reversible work first
Agents are pointed at coordination a person can look at, question and undo. Irreversible work, and anything that is properly a clinical decision, stays with people.
Access can be withdrawn immediately
Each agent has a narrow, defined job, and withdrawing its permission takes effect on its next call.
We are starting with narrow, low-risk and reversible coordination tasks. Agents do not make clinical decisions, and any action requiring professional judgement returns to a person. The risks and controls are assessed for each use case rather than assumed from the technology.
Here’s where we’re up to.
Agents are in active development and marked as Beta across the site. We are designing and testing them alongside the services that would use them and the people those services support.
We are starting with narrow, low-risk coordination tasks. The teams involved help define what an agent is responsible for, what it may access and when it must stop and ask. Trust in something new is earned through controlled use, visible evidence and clear accountability.
If there is a piece of coordination costing your team real time, we would like to hear about it.
Talk to us about AgentsFrequently asked questions
- Not as something you can buy yet. Agents are in development and available for early trials and co-development, so a service can point one at a narrow piece of coordination and shape it with us. The useful conversation now is which piece of coordination is costing your team the most time.
What is your team spending its time coordinating?
Tell us where staff are chasing updates, copying information between systems or coordinating dependencies across services. We are looking for narrow, measurable use cases to develop and test with partners.