AI in Home Care: Aides, Agencies and Unions Warn of Promise and Peril

Artificial intelligence is moving rapidly into American health care, and one of the places it is arriving most quietly is also one of the most intimate: the private homes of older adults who depend on paid aides to eat, bathe, move through their days and stay out of hospitals. A new qualitative study published in the Journal of General Internal Medicine offers one of the first systematic looks at how the people who will actually live with these technologies—home health aides and attendants, agency leaders, labor advocates, clinicians and technology company staff—believe AI will reshape home care. Their verdict is strikingly two-sided. The same tools that promise to streamline scheduling and sharpen documentation could, if deployed carelessly, deepen the very workforce crisis they are marketed to solve.

The research arrives at a moment of sharp inflection. According to data cited by the study’s authors, AI adoption across the health care sector rose from 5.9 percent in 2023 to 8.3 percent in 2025, with a 481.5 percent acceleration in the rate of adoption after late 2024. Home care is following the broader curve. AI-enabled wearable devices and ambient in-home sensor systems now track movement, sleep and vital signs to flag early clinical changes. Software platforms can automatically adjust caregiver schedules in real time based on patient acuity and geographic proximity, generate individualized care plans from aggregated patient data, and act as virtual assistants guiding patients through symptom reporting and communication with care teams. Yet the study’s authors note that almost nothing has been known about how the workforce at the center of this transformation perceives these systems.

The context makes that gap consequential. The United States faces a growing home care crisis as its population ages and more older adults experience complex medical needs and cognitive decline while wishing to remain at home. The home health aides and attendants who make aging in place possible are predominantly women from racial and ethnic minority backgrounds, earn low wages, receive insufficient training and recognition, are often undervalued by the health care system, and frequently work in isolation in patients’ homes with limited oversight or peer connection. Persistent workforce shortages and rising demand have made AI an attractive proposition for agencies and investors, but the study’s authors argue that integration is poised to carry both intended and unintended consequences for the relational, hands-on core of the work.

To capture stakeholder perspectives, researchers from Cornell Tech, Weill Cornell Medicine and Cornell University conducted semi-structured interviews between July 22, 2024 and April 18, 2025. Using purposive and snowball sampling, they recruited 43 participants across five groups: 11 home health aides, 10 home care agency leaders and staff, 14 worker advocates from unions and labor organizations, 5 clinicians with home care experience, and 3 technology company leaders and staff. Among the 36 participants with demographic data, the mean age was 44.6 years, 63.9 percent identified as women, 97.2 percent had completed at least some college, and 44.4 percent reported no or low knowledge of AI. Interviews lasted roughly an hour and were conducted over video conferencing, with all procedures approved by the Cornell University Institutional Review Board.

A distinctive methodological feature was the use of fictional vignettes grounded in real, commercially available AI products for home care. Each participant was assigned two of three scenarios: an AI tool to facilitate matching between aides and patients, a system to monitor aide-patient interactions in the home, and a tool to improve care coordination tasks. The vignettes were designed to give participants—who varied widely in technical familiarity—a consistent, realistic basis for discussion, and to encourage balanced reflection on benefits, risks and tensions. Technology company participants, given their deep familiarity with AI, instead answered questions about their own products. Interviews concluded with broader questions about governance and control.

The analysis followed a rigorous three-stage qualitative process: structural coding of high-level topics, inductive generation of sub-codes—223 in total—and thematic analysis, with transcripts coded in ATLAS.ti and transcription performed using a locally run open-source AI tool. Four major themes emerged. The first concerned benefits: participants saw AI improving documentation and symptom capture in the home, breaking down information barriers between different caregivers so that families and clinicians stay on the same page, strengthening aide engagement and retention through better job matches, and delivering efficiency gains for agencies and even for advocacy organizations answering routine member questions about contracts and benefits.

The second theme was darker. Participants, particularly clinicians, worried that AI could dehumanize care and corrode trust between providers and care recipients, with one physician imagining a world in which people are supposed to trust AI and therefore distrust each other. Many feared that AI would shift administrative labor onto aides, who would spend more time filling out forms on tablets and less time in direct human contact with clients. Agency staff and worker advocates emphasized that AI systems require large volumes of manually entered and maintained data, and that aides and agency staff would bear primary responsibility for keeping that information accurate, current and eventually removed—a burden one agency participant, drawing on a prior non-AI technology rollout, described as requiring hundreds of hours of manual input.

Participants also warned that these added burdens could worsen workforce shortages in an already strained labor market by stripping out the relational aspects of the job that keep people in it. Some noted that AI investment would flow most powerfully where incentives favor maximizing profits and minimizing wages, raising the specter of wage suppression. A third theme centered on data responsibilities, privacy and AI literacy: technology participants conceded that a system is only as good as the information put into it, clinicians questioned surveillance in the home and called for opt-out options, and aides themselves described a stark lack of agency—if the agency decides to put AI in a patient’s home, one aide said, the aide cannot say no. Limited AI understanding among aides, advocates added, could hinder their ability to advocate for themselves in contract negotiations.

The fourth theme carried a note of urgency and opportunity. Because AI adoption in home care is still nascent, participants saw a closing window to establish operational guidelines and labor protections before the technologies become entrenched. Labor organizations could embed AI provisions in contracts and educate their members, though several acknowledged their organizations lack internal technical expertise and may need to retool. Nearly all participants wanted clear governance and regulatory frameworks ensuring that care recipients, family members and the full care team have input into design and oversight. The authors note that while home care is regulated under HIPAA and Electronic Visit Verification requirements, AI-specific legislation has yet to emerge despite recommendations from the World Health Organization.

The study’s limitations are acknowledged by its authors: it was a small-scale qualitative study concentrated in urban, largely northeastern U.S. settings, so generalizability to rural or non-U.S. contexts remains unknown; the vignettes may have shaped participants’ views; and the perspectives of older adults and family caregivers themselves were not captured, representing a key direction for future research alongside ethnographic studies of real deployments. Still, the conclusions are pointed. AI tools may improve patient care, documentation and workforce engagement, but they risk eroding patient-provider relationships, increasing burdens on a strained workforce, and raising unresolved questions about data quality and privacy. Because adoption is early, the authors argue, stakeholders have a critical window to build inclusive governance frameworks centered on the aides and patients most affected—so that AI supports, rather than undermines, care quality and workforce sustainability in one of health care’s highest-stakes settings.

Subject of Research: Stakeholder perspectives on the benefits, risks and governance of artificial intelligence in home care work for older adults

Article Title: Understanding Key Stakeholders’ Perspectives Towards Artificial Intelligence in Home Care Work

Article References: Solano-Kamaiko, I. R., Dicpinigaitis, M., Tan, M., Avgar, A., Vashistha, A., Dell, N., & Sterling, M. R. (2026). Understanding Key Stakeholders’ Perspectives Towards Artificial Intelligence in Home Care Work. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10782-z

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10782-z

Keywords: artificial intelligence, home care, home health aides, qualitative research, health care workforce, AI governance, data privacy, labor unions, elder care, health services research, surveillance, workforce shortages