TL;DR
Home-based care providers are exploring agentic AI for tasks such as caregiver scheduling and call-queue management, but adoption remains cautious. Legal and industry leaders warn that autonomous tools could create serious operational or patient-safety risks if they make mistakes, especially in clinical settings.
Home-based care providers are beginning to test agentic AI for scheduling and other operational tasks, even as legal and industry leaders caution that errors by autonomous or semi-autonomous systems could carry serious consequences for patients. The emerging debate, reported by Home Health Care News, is over how quickly providers should adopt tools that can act on tasks rather than only generate responses.
Agentic AI can interpret goals, make decisions, plan and carry out actions, and use digital tools, according to Stanford University’s description cited by Home Health Care News. That distinguishes it from generative AI systems that produce content in response to prompts but do not independently complete tasks. Polsinelli attorney Greg Leighton told the publication that giving an agent control over operational tasks also means placing some operational risk in the tool’s hands.
Leighton warned that the potential effect of an error depends on what a system is assigned to do. A hypothetical patient-monitoring agent that missed a serious health issue or judged its urgency incorrectly could put a patient at risk, he said. He advised providers to first adapt existing processes using generative AI and address compliance questions before moving to agentic workflows.
Some companies are testing administrative uses. TheKey is rolling out Caribou’s Conductor agentic scheduling product, which it expects to help find caregiver coverage when someone calls out. Chief People Officer Emily Kraeter said the company is considering agents to support employees in different roles. Modivcare CEO of personal care services Marshalina Ramos said agents could help streamline call queues, while the company is focusing on more accessible operational improvements.
Why Scheduling Is an Early Test
Scheduling and other back-office tasks offer providers a way to test whether agents can reduce administrative work without giving them direct responsibility for care decisions. At TheKey, the proposed benefit is not only faster coverage coordination: Kraeter said freeing caregiver managers from some tasks could give them more time for training, relationships and staff support.
But the stakes differ sharply by use case. An error in arranging a shift may disrupt operations; a mistake in monitoring or escalating a patient’s condition could have a direct health impact. That distinction makes human oversight, clear limits on what an agent can do and a sound business case central to decisions about deployment. In a highly regulated industry, providers also have to weigh efficiency against compliance and the possibility of adding risk.
From Scheduling to Wider Adoption
The providers described in the Home Health Care News report are at different stages of adoption. Modivcare, a Denver-based provider of nonemergency medical transportation, personal care, remote monitoring and supportive care in 35 states, is prioritizing lower-barrier ways to improve efficiency. Ramos said providers are wary of being the first to use unproven technology amid scrutiny over fraud, waste and abuse.
TheKey, which provides home care and memory care in 29 states and Canada and is expanding home health services, began looking at automated scheduling. Chief Operating Officer Greg Meyers said that work led the company to consider how agents might support other roles across the organization. Leighton, however, recommended that providers build experience with generative AI and compliance before assigning more independent tasks to agents.
“When you go to agentic, you’re really putting your operational risk in the hands of the tool.”
— Greg Leighton, privacy and incident response vice chair at Polsinelli, speaking to Home Health Care News
Safety Limits Remain Unspecified
The report does not provide performance data showing whether the tools being tested save time, reduce missed shifts or produce errors. It also does not specify what human review or safeguards will govern TheKey’s scheduling rollout, or whether any providers are using agents to make or act on clinical decisions. The patient-monitoring example raised by Leighton was hypothetical, not a reported incident.
Providers’ plans and risk assessments may differ by product and task. The report does not give a timeline for broader deployment or describe a common industry standard for evaluating agentic AI. How much authority agents will receive, and how responsibility will be assigned when they make mistakes, remain open questions.
Provider Trials Will Shape Adoption
The next evidence to watch is how providers evaluate early administrative deployments, including TheKey’s scheduling product: whether it performs reliably, what human checks remain in place and whether it measurably reduces staff workload. The companies have not announced results or a schedule for wider rollout in the source report.
Leighton’s advice is for providers to identify a clear business case, understand possible downsides and begin with bounded administrative tasks such as appointment coordination, billing or reimbursement management. Any move into more consequential work will depend on whether providers can show that the technology supports operations without disrupting care.
Key Questions
What is agentic AI?
It refers to AI systems that can interpret goals, make decisions, plan and carry out actions, and use tools. Unlike generative AI that responds to prompts with content, an agent can act on tasks with some degree of autonomy.
How are home care companies using or testing it?
TheKey is rolling out Caribou’s Conductor product for agentic scheduling, which it expects to help find caregiver coverage when a worker calls out. Modivcare has discussed potential use in streamlining call queues.
What risks have experts identified?
Polsinelli attorney Greg Leighton warned that an agent could make an operational mistake. In a hypothetical clinical use, such as patient monitoring, missing or misjudging a serious issue could have severe consequences. The report does not describe a real patient-safety incident involving these tools.
Are providers adopting agentic AI across patient care?
The report describes interest in administrative tasks and a longer-term idea of agents supporting employees in different roles. It does not establish that providers are broadly using agents to make clinical decisions or manage patient care.
Source: homehealthcarenews.com
