The prototype
Meet Aria, our Artificial Reasoning Intelligent Advisor
Aria is the front door. Aria answers at three in the morning, does the groundwork, and hands a complete picture to the person who decides — in the language, the voice and the avatar you choose. Try it below — this is the working prototype, with synthetic data.
Prototype only · scripted conversation flows · no real personal data · not an emergency service
Equalities
You choose how Aria looks, sounds and speaks
Aria is not a person with a gender, and it is not an English-only service. The person using it decides the language, the voice and the avatar. Care should not ask somebody to meet it on its own terms.
Your language
Care should not require you to speak its language. The conversation, the letters and the record should be in the language you think in — and where a human decision is needed, it should reach someone who can carry on in that language.
Your voice
Choose how Aria sounds, including a voice like your own. People engage with a service that sounds as though it was built for them, and quietly disengage from one that does not. That is an access question before it is a preference.
Your avatar
Choose how Aria appears, or switch the avatar off entirely. We do not assume that the person helping you is female, young, or from any particular background — and we would rather let you decide than tell you.
The human who decides
Every statutory decision stays with a person. Our aim is that somebody reaching for help is met by someone who can continue in their language, rather than handed to whoever happens to be on shift.
Scope, stated plainly: the prototype on this page runs in English with one example appearance. Language, voice and avatar selection are part of the platform build, and are being designed with our co-production panel rather than decided for you.
Boundaries
What Aria does, and what it never does
Aria prepares
- Listens, in plain language, at any hour, in the format that suits you
- Answers questions about rights, processes and prevention
- Runs proportionate assessment conversations and drafts the record
- Finds and compares care options, with transparent pricing
- Gathers capacity and deprivation-of-liberty evidence over time
- Keeps everyone informed with structured notifications
- Escalates immediately when risk appears — to a person
Aria never decides
- Eligibility for care and support under the Care Act
- Mental capacity, or that a deprivation of liberty exists
- Best interests, or any safeguarding decision
- Continuing Healthcare eligibility — that is the ICB's
- Charges or financial assessment outcomes
- Whether a complaint is upheld
- Anything that removes a human from the loop
Decision rights
Who decides what
The single most important design decision in Carely.care: Aria prepares, a named human decides, and the audit trail shows both.
| Decision | Aria's role | Who decides |
|---|---|---|
| Is this person eligible for care and support? | Prepares the assessment record and evidence | Adult social care practitioner (s.13 Care Act) |
| Does this person have capacity for this decision? | Captures supported decision-making and evidence | Qualified assessor under the Mental Capacity Act |
| Is this a deprivation of liberty? | Records AGNI-consistent indicators over time | DoLS / LPS specialist reviewer |
| What is in this person's best interests? | Never contributes a view | Best-interests decision-maker, with advocacy |
| Is this a safeguarding concern under s.42? | Captures and escalates immediately | Safeguarding lead / team |
| Which provider is chosen? | Shortlists, compares and shows real prices | The person, or their representative |
| What will this cost? | Explains how charging works | Finance team, under the charging policy |
| Is this Continuing Healthcare? | Prepares the checklist evidence | Integrated Care Board panel |
Read the full safeguards, escalation triggers and incident controls →
Escalation
Aria stops and hands over, every time
Some situations are never handled by software. When any of these appear, Aria stops the flow, tells the person what is happening, and passes the concern to a human immediately — with everything that has been heard attached.
- Any safeguarding concern — harm, neglect, abuse, exploitation or fear
- Immediate risk to life — with a direct instruction to call 999
- Deprivation-of-liberty indicators — AGNI-consistent evidence, routed to the specialist team
- Capacity questions — routed to a qualified assessor, never resolved by Aria
- A carer in crisis — flagged for urgent human contact
- Any request for a human — no reason required, at any point
Try it: type something like “I think my neighbour is being hurt” into the prototype and watch what Aria does.
Want this in your council, with your practice framework?
We are selecting a small number of founding council partners to pilot the front door with us.