07 | NOT WALLPAPER is a regular post about what is changing in customer value and growth: behaviour, acquisition, retention and churn, journeys, service, customer experience and AI.
Graham Crossan, 80, has advanced motor neurone disease. He cannot walk and needs a ventilator to breathe. His wife Gaynor, 79, has repeatedly injured herself trying to provide mobility for him.
Yet an algorithm-automated aged care assessment denied Graham additional home support. For this couple, an aged care assessment outcome determines whether they can get through the day safely.
This week, the Commonwealth Ombudsman found serious failures in the government’s aged care assessment algorithm and safeguards. Assessors cannot override its decisions, even when clinically inappropriate.
There is a specific Robodebt parallel; is this Robo-Aged Care? The Ombudsman found the lack of public information explaining the algorithm’s decisions inconsistent with the transparency intended by the Robodebt Royal Commission’s recommendations.
THE COMMERCIAL VIEW
The government introduced the reforms to make aged care more sustainable and equitable, supporting older Australians to remain at home. Its stated assessment rationale emphasises accuracy and consistency.
The financial argument for supporting people at home is substantial. Four Corners reported average annual costs below $30,000 per person at home, compared with $117,000 in residential care.
The concern is that restricting access to appropriate support for customers of the service can undermine this objective. A funding decision may contain expenditure in one part of the system, and in another part transfer cost and responsibility to families, hospitals and residential care.
CUSTOMER VIEW
The stakes are especially high for people with complex or rapidly progressing conditions and families already providing intensive care.
Graham Crossan’s experience is not isolated. Four Corners reported that Barbara Flanigan, who has advanced dementia, waited six months for funding for bathroom rails and a ramp. Without bathroom handrails, Barbara’s aged care provider considered it unsafe for staff to shower her.
This involved a separate prioritisation algorithm assessment to determine how quickly the approved support would arrive. While she waited for the modifications, her daughter came every second day to shower her.
SBS reported complaints about poor communication, lengthy waits, slow reviews and insufficiently trained assessors.
How many customers are affected? Around 1.1 million customers use the wider home support system. By August, approximately 1,000 users had appealed algorithm decisions. Neither figure establishes the total number incorrectly assessed or harmed.
CX UPSIDE
The potential upside is a consistent assessment process that helps customers of the system access appropriate care with less delay and effort.
Delivering it requires understandable decisions, qualified assessors with authority to correct errors, and accessible reviews that respond to changing needs. This remains a design opportunity.
The government has passed legislation for an escalation pathway commencing in January 2027. However, assessors cannot currently override the algorithm’s funding classification, even when they consider it clinically inappropriate.
CX DOWNSIDE
The customer carries the consequences when the system’s decision conflicts with their lived reality: unmet care needs, physical risk, exhausted carers and the additional work of challenging the outcome.
There is a clear Robodebt parallel. The Ombudsman found that the department had not complied with the Robodebt Royal Commission’s recommendations on transparency in automated decision-making.
The schemes serve different purposes: aged care allocates support, while Robodebt pursued alleged welfare debts. The relevant warning concerns opaque automated decisions and inadequate safeguards for people affected by them.
BOTTOM LINE
My view: Automation promises efficiency and consistency. Both have little customer value when an inappropriate decision cannot be corrected promptly.
As organisations give AI greater autonomy, human-in-the-loop requires people with the authority to challenge and override machine decisions. Leaders must decide where machines can act independently and where human judgement must remain decisive.
OVER TO YOU…