Workers are gaining a voice in health system AI decisions, from technology investments at Kaiser Permanente to tool design at HCA Healthcare and new AI concepts at Mayo Clinic.
The approaches range widely: Kaiser bargained with its unions to give labor half the seats on a task force with authority over technology investments, HCA involved front-line nurses in designing a handoff tool, and Mayo asked nurses to propose and help select AI agent concepts.
The efforts come as unions across sectors, including hospitality and entertainment, increasingly seek AI agreements with employers, according to MIT Sloan School of Management researchers.
At Oakland, Calif.-based Kaiser, that participation extends to systemwide technology decisions. An AI task force includes five labor representatives and five Kaiser leaders, with direct decision-making authority over technology investments.
The arrangement stems from a 2025 agreement with the Alliance of Health Care Unions, which represents 62,000 nurses, technicians and other healthcare workers, according to a Sept. 28 article from Cambridge, Mass.-based MIT Sloan School of Management. The researchers called it the first comprehensive agreement by a major U.S. healthcare employer and its unions to give workers input at every stage of AI decision-making.
The deal followed a February 2025 prebargaining conference attended by about 300 Alliance leaders and 80 Kaiser leaders. There, employees said they feared for their jobs, and many wanted a say in AI decisions.
Alongside the task force’s oversight, more than 3,500 unit-based teams will identify opportunities for AI to improve workflows or patient experience. Kaiser is also upgrading its internal project tracker to make AI pilots visible across the organization and minimize duplication.
Union representatives will train peer advisers to help co-workers adopt AI tools, a model drawn from Kaiser’s EHR rollout two decades earlier. Representatives have also requested inventories of AI tools and documentation of AI use in hiring and performance monitoring.
At Nashville, Tenn.-based HCA, nurses helped shape the information an AI tool delivers during a shift change.
Front-line nurses at HCA Florida UCF Lake Nona Hospital in Orlando helped develop and pilot Nurse Handoff with Google Cloud. They helped determine which information the tool presents and how it fits into nursing workflows.
The tool organizes EHR information into a shift report that incoming nurses can review on a hospital-provided mobile device. HCA said the tool was in beta at eight hospitals in July. The system expanded it to 15 additional Florida hospitals in September. Nurses remain responsible for handoffs and their clinical decisions, the system said.
HCA said nurses who took part in early testing rated the tool as factual and helpful. The health system’s Sept. 25 news release did not include results measuring the tool’s effect on handoff time or patient outcomes.
Rochester, Minn.-based Mayo Clinic gave nurses a role in identifying problems for AI to address and choosing concepts to advance.
Its nurse-led AI agent competition received more than 450 submissions. Nurses voted on five finalists from a field of 33 semifinalists. Their concepts included an agent intended to streamline Family and Medical Leave Act and short-term disability paperwork and another designed to help nurses find patient-specific guidance.
As of Becker’s Aug. 27 report, the finalists were working with Microsoft on proof-of-concept development. Two of the five finalists, Anna Callahan, BSN, RN, and Duc Chu, RN, said more development and testing were needed before decisions on wider rollout. Neither identified a target pilot or go-live date.
Mayo Clinic Chief Nursing Officer Ryannon Frederick, RN, said the organization wants tools to be reliable and useful before putting them into practice.
“We just don’t add it in to add it in for technology’s sake,” she said.
Ms. Frederick said tools focused on clerical work are likely to reach practice sooner than tools that touch patient care directly, because those require more validation.
