Every aged care facility in Australia is already generating the intelligence it needs to transform care. The problem is that most of it is locked away, labelled incorrectly, or simply never connected to anything meaningful.
Agentic AI - the next generation of intelligent systems capable of acting autonomously on behalf of staff - is arriving fast. It can monitor residents, surface early warning signs, and reduce the administrative load on frontline workers.
But, and this is a BIG but, without a deliberate data foundation, even the most advanced AI will fail.
The conversation in aged care needs to shift from what can AI do to what data does AI need — and the context for that discussion comes down to how we store, catalogue, and structure the underlying data it to make that possible.
The Scale of the Challenge
Australia's aged care sector is under pressure from every direction. Demand for residential care services is growing steadily, with 185,127 permanent residents recorded in 2023 — up 40% from 1999. The workforce cannot keep up. Research from CEDA projects a shortfall of at least 110,000 direct care workers by 2030, growing to 400,000 by 2050 unless dramatic action is taken. By 2035, aged care alone faces a shortage of nearly 80,000 nurses.
Technology is not, nor will it ever be, a replacement for human carers - but it is a proven force multiplier!
The Aged Care Act 2024, which commenced on 1 November 2025, now requires providers to meet stronger accountability and quality standards. The Australian Government's Aged Care Data and Digital Strategy 2024–2029 explicitly targets the secure sharing and reuse of data to deliver a sustainable, continually improving care system.
The changed regulatory environment and the operational reality are pointing in the same direction:
organisations that invest in their data infrastructure now will be better equipped for everything that follows.
Every System in A Facility Is a Data Source
A modern aged care facility generates a surprising volume and variety of data every single day. The challenge is that most of it lives in separate systems, owned by different vendors, in different formats, with no common language.
Consider what a typical resident interaction generates across a 24-hour period:
- Access control and visitor logs - record who enters and leaves the facility, at what time, and through which areas — data that speaks to both security and the social engagement of residents.
- CCTV and activity monitoring systems - capture movement patterns, time spent in communal areas, and changes in gait or mobility — all of which are early indicators of clinical deterioration.
- Nurse call systems - log the frequency, time, and type of alert — data that, over weeks, reveals patterns in pain, anxiety, or changing care needs.
- Meal ordering and consumption records - track appetite, dietary preferences, and changes in eating behaviour — a well-established early signal of depression, infection, and cognitive decline.
- Clinical care documentation and medical notes - hold the formal record of health status, but in most facilities this remains largely unstructured — locked in PDFs or free-text fields that AI cannot easily process.
- Medication management systems - track administration, refusals, and timing — data critical to identifying adverse events or emerging non-compliance.
- Excursion logs - reflect social connectedness and quality of life, which research consistently links to cognitive and emotional wellbeing.
- TV and streaming media consumption - can indicate changes in alertness, interest, and engagement — a resident who stops watching their favourite programme is telling you something.
- Phone call activity - reflects social connection and, when significantly reduced, can be an early indicator of withdrawal or low mood.
- IoT sensors and wearable devices - provide continuous data on movement, sleep quality, falls risk, and physiological signals.
Each of these data streams is valuable in isolation. Together, they create something qualitatively different: a continuous, multi-dimensional picture of the whole person.
The Problem With Silos
Right now, most of that data never talks to anything else. Up to 80% of healthcare data globally exists in unstructured formats — clinician notes, PDFs, scanned documents — that are largely unusable without significant preprocessing. Data scientists in healthcare settings spend 50 to 80% of their time just cleaning and preparing data before analysis can begin.
This is the silent cost that never appears in an IT budget. Every hour a system is not connected, data is being generated and discarded. Every week that passes without a structured data layer, the gap between what is captured and what is useful grows wider.
The CSIRO and the Digital Health Cooperative Research Centre have identified connected data and coordinated care as vital to quality outcomes in aged care.
The National Aged Care Data Asset now draws on over 80 data tables spanning 27 years of program data. But at the facility level, providers are often still operating with fragmented systems that cannot communicate with each other, let alone with national infrastructure.
How & where to Store, Catalogue, and Structure Data for Agentic AI
Getting data ready for AI is not a single project, nor a single decision. It is an ongoing practice that acknowledges that the approach must be designed for the full range of data types that aged care generates. And no, "the file server" is not the answer to the "where do we store this data"!
The above data sources, common to most facilities, can be loosely categorised into three types:
- Structured data — nurse call logs, meal orders, medication records, access events — can be stored in standardised formats using agreed coding systems. In healthcare, this is achieved by adopting standards such as HL7 FHIR, SNOMED CT, and LOINC - aged care doesn't have the maturity of data standards, however, the storing structured data is essential as it is relatively easy for AI systems to query and analyse at scale.
- Unstructured data — clinical notes, voice recordings, scanned documents — requires transformation before it can be used. Natural language processing tools can extract meaning from free-text clinical documentation. AI-powered data catalogues can classify and tag this content automatically, dramatically reducing the manual effort required. A well-designed metadata catalogue ensures every piece of data is findable, accessible, interoperable, and reusable — the FAIR principles that govern modern health data ecosystems.
- Time based (Streaming) data — from IoT devices, CCTV analytics, and real-time sensor feeds — requires a different architecture: an event-driven data mesh that ingests, processes, and routes data in near real-time. Technologies such as Apache Kafka for ingestion and Apache Spark for processing have proven effective in most IoT environments.
While this covers the categorisation of the data types, the where and how to store is a little more challenging. Structured data will typically be stored in a database. Unstructured data in a file repository and time based data in an object store. Each of these is unique in its own way and requires specific design characteristics to enable AI access.
The practical goal is a resident-centric data layer — a unified, longitudinal record that pulls from every system in the facility and presents a coherent view of each individual. This is not just a technology decision. It requires clear data ownership, role-based access controls, consent frameworks, and audit trails that meet the requirements of the Privacy Act 1988 and the Australian Privacy Principles.
Data sovereignty matters too. Sensitive resident data must be stored on Australian soil and governed by Australian law. Any cloud infrastructure used to support AI workloads must comply with this requirement. This is a design decision, not an afterthought!
What Agentic AI Can Do With Well-Structured Data
Once the data foundation is in place, agentic AI systems can begin delivering outcomes that matter to staff, residents, and families.
Below are ideas once thought impossible to achieve, but now well within reach once a centrallised, consistent data structure is implemented:
- Resident-centric causality analysis. When a resident's condition changes, clinical staff spend time reconstructing what happened. Agentic AI can do this automatically. Correlating a recent change in nurse call frequency with a dip in meal consumption, a reduction in movement captured by IoT sensors, and a note from the previous shift. It surfaces the likely cause before it becomes a crisis. Multi-agent systems (where specialised AI agents collaborate to analyse different data streams simultaneously) have demonstrated meaningful improvements in diagnostic accuracy and clinical traceability.
- Longitudinal wellbeing trend monitoring. The most important changes in a resident's wellbeing often emerge over weeks, not hours. Agentic AI can monitor long-term patterns across all data streams: visitor frequency, activity levels, sleep quality, appetite, social engagement. Identifying deviations before they become clinically significant. Research using unsupervised learning on household movement data has successfully detected urinary tract infections and impending hospitalisations in people living with dementia days before clinical presentation. The same principle applies in residential care.
- Anomaly detection and alerting. Rather than requiring staff to review dashboards or reports, agentic AI can proactively alert frontline workers when something unusual is detected. A resident who has not been near the dining room for 36 hours. A nurse call pattern at 3 AM that has not occurred before. An access log showing no family visits in three weeks. Each of these is a signal and the system can be configured to bring them to a staff member's attention in the context of everything else it knows about that resident.
- Population-based insights. The value of well-structured data extends beyond the individual. When data is catalogued consistently across a resident cohort (or even across multiple facilities) organisations can identify trends that are invisible at the individual level. Which dietary changes correlate with improved mobility outcomes in residents over 85? Which nurse call response time thresholds are associated with reduced hospital admissions? Population analytics turns operational data into quality improvement intelligence.
Where Do I Start?
Here are some practical steps to improve data storage, reporting and readiness for Agentic AI solutions:
- Begin by auditing what systems you have and what data they generate.
Identify the three or four systems that are currently disconnected and prioritise integration work based on clinical value. Even basic nurse call data connected to a care management platform creates immediate value.
Work with, or migrate to, vendors who support open APIs and data export in standard formats.
- Invest in a data governance framework before you invest in AI tooling.
Define data ownership roles, establish consent processes for data use, and implement a metadata catalogue that spans all facilities.
Standardise clinical documentation templates to reduce the proportion of unstructured data in your records.
- Design for a federated data architecture that supports both facility-level and population-level analytics.
Implement streaming data infrastructure for real-time IoT and security system integration.
Commission an AI readiness assessment that maps your current data maturity to the capabilities you want to deploy.
Engage with the National Aged Care Data Asset to understand how your data connects to the national picture.
The Investment That Enables Everything Else
Aged care organisations are being asked to absorb significant change — new legislation, new quality standards, workforce pressure, and the arrival of genuinely transformative technology. While it is tempting to focus on the AI itself (which platform, which vendor, which use case), the evidence is clear.
AI is only as powerful as the quality of data it learns from.
Without clean, connected, well-catalogued data, even the most capable agentic system will produce outputs that are unreliable, incomplete, or clinically unsafe. The data is not the means to an end. It is the asset.
Investing in how, where, and why your data is stored (and who can access it and under what conditions) is the decision that will determine whether AI delivers genuine value for your frontline staff and your residents. Get the foundation right, and the technology will follow.
Organisations that treat their data infrastructure as a strategic investment today will be the ones that can deploy agentic AI with confidence tomorrow — delivering the resident-centred, proactive, and evidence-driven care that every older Australian deserves.