About this blog

This is a design diary from an Azure Master Architect (AMA) capstone, Agentic Care Operations for an Asian aged-care group. The brief describes a regional operator with 180 facilities and home-care stations and about 9,000 caregivers, and asks for five AI agents that give time back to care without taking decisions away from people.

I've framed the design around Australian regulation and built it on Azure in Australia East. The customer in these posts, Zava Care Group, is fictional. So are its staff, residents and families.

How to read the status labels

Design blogs tend to blur what's drawn and what's running. Every component in this series carries one of these labels:

• Target Designed or required, but not built.
• As-built Deployed in a sandbox Azure environment with synthetic data only.
• Demo Shown working end to end with synthetic data, with its simplifications stated.

Figures from the case-study brief, such as "documentation time cut by 60%", are targets set by the brief. None of them has been measured.

The five agents

• Care Record Turns spoken or typed notes into draft entries that a caregiver confirms.
• Billing Prepares claims from confirmed care. A person certifies each one.
• Shift Planning Recommends rosters. The facility manager approves them.
• Night Watch Raises falls and wandering alerts on site. Staff respond.
• Family Communication Answers families within a fixed disclosure policy and passes sensitive topics to staff.

Disclaimer

This is learning and portfolio material. Nothing here is clinical, legal or regulatory advice. Views are my own.