Kate is the engine. The personas are the voice.
Every persona in the careplans AI family runs on Kate: an agentic voice engine that calls, listens with emotional intelligence, and turns every conversation into care records, insights, and actions.
Kate is the agentic voice engine behind every persona.
Not a persona. Not a chatbot. Kate is the intelligence layer that makes every call purposeful, every insight actionable, and every outcome actionable. She is the reason this is more than an AI that makes phone calls.
What Kate does on every call
- Schedules the call at the right time for each person
- Selects the right persona (Mary, Jennie, Bloom, or any other)
- Loads full context from every previous conversation
- Configures the conversational pathway with branching logic, conditions, and guard rails
- Orchestrates the conversation in real time
- Receives emotional signals so the conversation adapts to mood in real time
- Triggers mid-call API actions: check a database, book an appointment, update a CRM record, send an SMS
- All while the person is still on the line
- Transcription and recording
- Routine analysis on cost-optimised local models
- Advanced reasoning through frontier models
- Emotional analysis (tone, sentiment, mood)
- Generates WHO-5 wellbeing scores, structured JSON data, and compliance evidence
What makes Kate different
Escalation and action
Escalation alerts to the care team. Family notifications through nonni.ai. CRM updates. Webhook pushes. Compliance evidence packs. Follow-up call scheduling.
Longitudinal intelligence
Kate maintains memory across calls. She tracks wellbeing trends over weeks and months. She detects a gradual mood decline over three weeks, increasing mentions of pain, growing isolation.
Multi-model architecture
The right model for each task. Routine analysis on cost-optimised local models. Advanced reasoning through frontier models. Emotional analysis through specialised voice models. Cost-optimised without compromising accuracy.
Cross-call memory
Every persona remembers. Margaret mentioned her daughter last Tuesday. John has been sleeping poorly for two weeks. Context carries forward automatically.
Conversational pathways
Node-based conversation flows with branching logic, conditions, and guard rails. The AI follows the pathway. Hallucination-proof. Safe by design.
Continuous compliance
Every call maps to the relevant quality standards. Evidence generates as a byproduct of caring, not as an admin task. Always audit-ready.
114+ voice tools. 19 capability areas.
Every capability is available to every persona. Hover or tap any area to see what is inside.
View the full capabilities listVoice to insight to action
One sentence. Ten actions. Zero data entry.
Margaret mentions she needs a shower chair during a regular check-in. Watch what Kate does next. (Example only, subject to internal processes and procedures. Full support for human-in-the-loop via voice, SMS, email or existing authorisation processes.)
This is what "voice to insight to action" means. Mary cares. Kate acts. Margaret gets her shower chair.
Want to see it for yourself?
Live demos are by appointment, walked through with Andrew. Ten minutes, real conversation, real questions.