AI agents are being used to handle healthcare administration, from insurance claims to aged-care staffing, according to a new PYMNTS report. The reported results include faster appeals and less time spent on routine checks, with human approval still in place for some of the work.
Watch Desk analysis
What happened
The report describes several different deployments. At New York’s Hospital for Special Surgery, chief digital and technology officer Ashis Barad says agents now complete 1,100 insurance claims a month. He says the appeals stage fell from 45 minutes to five, while the appeal success rate rose from 65% to 100% over nine months. The hospital also uses an Ema service to help patients find and book clinicians.
In Australia, Wesley Mission is using three agents for aged-care administration. The report says one saves staff 30 to 60 minutes per shift on care-plan assessments; another reduced daily visit verification from six hours to about 15 minutes. A third recommends staffing changes for a supervisor to approve.
Why it matters
These examples put a practical shape on agentic AI in healthcare: software can gather information, handle repetitive steps and flag what needs attention. That may free staff for patient care, but the figures are claims attributed in PYMNTS’ report to the organisations and sources it cites, not a guarantee that every service will deliver the same results.
The human hand-off is part of the story, not a footnote. The report says uncertain or sensitive cases at the hospital are sent to specialists, and Wesley Mission’s agents require approval before final action.
Our read
The most persuasive case here is not that an agent can do everything. It is that a narrow, measurable task can be taken off a busy team’s plate while people retain the decisions that need judgement. Start with the dull paperwork, by all means. Just make sure someone can see what the agent did and step in when it gets stuck.
What to watch
- Whether the reported time savings hold up as these systems are used more widely.
- How providers measure accuracy and handle errors or sensitive cases.
- Whether human review remains meaningful as agents take on more tasks.
Discussion spark: Which healthcare tasks should AI agents be allowed to complete, and which should always require a person’s approval?
Sources and evidence
- Source update (26 October 2026, 08:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.