An investigation published July 9 by CalMatters documented call monitoring, predictive productivity software and monthly performance scores used on Kaiser Permanente advice nurses in California. Interviews with seven current and former nurses found that staff are questioned about calls lasting more than 15 minutes and receive monthly scores tracking their speed and activity. Software also records active and inactive time and attempts to predict whether nurses are unproductive or answering too slowly.

Pa Vue, a call center nurse and union representative, recalled having her score reduced after repeating advice to a patient whose symptoms suggested a possible heart problem. She had also seen scores fall when nurses departed from software recommendations. Raquel Alvarez Sanchez remained on the telephone with a suicidal patient for more than an hour until police arrived, knowing the call would affect her average handling time for weeks. As a union steward, she has accompanied colleagues to performance meetings after calls that were handled correctly except for exceeding 15 minutes.

Another nurse withheld the additional time and compassion she believed an elderly patient needed after receiving a terminal cancer diagnosis because she feared damaging her score and being reprimanded. Nurses said they are told to limit themselves to two or three pieces of advice, although calls requiring an interpreter routinely take more than 30 minutes.

Nurses said the time allowed after calls has fallen from about 10 minutes to 30 seconds or less when lines are busy; extra time requires a manager’s permission. In 2024, Kaiser also piloted software that graded tone and apparent empathy in the voices of nurses and patients. One nurse said the program repeatedly scored calls incorrectly. Nurses petitioned against it, and Kaiser suspended the pilot in November, but they were told it could return. Kaiser declined to say whether patients knew their voices were being assessed. Charlotte Capulong, a 22-year call center veteran, told CalMatters: “You aren’t calling Comcast. We’re dealing with life here.”

Kaiser denied using average handling time to evaluate nurses or enforce call-time targets. It said its contact-center tools support quality assurance and are subject to human review. But the company did not explain why nurses are questioned about calls exceeding 15 minutes or why call time appears in the scores they described.

Management’s ability to impose these technologies depends on the trade union bureaucracy suppressing resistance and accepting contracts that preserve management’s authority. The nurses interviewed by CalMatters belong to the California Nurses Association (CNA), which represents 25,000 Kaiser nurses, including about 1,000 in call centers, and has historically remained outside Kaiser’s infamous Labor Management Partnership (LMP). But CNA begins contract talks this month, and its existing contract contains few limits on new technology outside of requiring notice, which even the union says Kaiser has not always provided.

For the 150,000 workers covered by the LMP, the Alliance’s AI agreement confines the unions to notification, discussion and recommendations while explicitly promoting the “accelerated” implementation of new technology.

The Labor Management Partnership covers roughly 150,000 other Kaiser workers represented by the Alliance of Health Care Unions and the Coalition of Kaiser Permanente Unions. More than 31,000 Alliance members represented by UNAC/UHCP (United Nurses Associations of California/Union of Health Care Professionals) struck Kaiser from January 26 to February 24 over wages, staffing and working conditions. The bureaucracy ended the strike without a tentative agreement or membership vote. The contract was not released until March and was subsequently ratified. Its national AI provisions had been settled on August 22, 2025, five months before the strike.

The AI agreement creates a task force with five management and five labor representatives. Kaiser must notify the union before introducing AI for productivity or performance monitoring, but disputes are relegated to local contracts and cannot use the national agreement’s dispute procedure. The task force may share information, discuss deployments, guide training and make recommendations. It has no authority to reject or suspend a system or to prevent information produced by it from being used in discipline.

The agreement lists among its expected outcomes the “successful, accelerated, and sustainable implementation” of AI and technology. Unit-based teams of workers, union representatives and managers are to identify innovations for review and possible use across Kaiser. The preceding national agreement assigns these teams roles in setting metrics, evaluating performance and participating in staffing and scheduling while requiring their goals to align with Kaiser’s regional and national objectives. Proposals are entered in a tracker for review and possible use across Kaiser.

An MIT Sloan report on more than 150 hours of negotiations records management’s demand for rapid adoption to keep Kaiser competitive. Union participants argued that more information would “reduce fear of AI” and that their ties to frontline workers could make employees more receptive. The report says management viewed labor officials as important in shaping the workforce response. It notes that jobs could disappear as work was reorganized; labor participants focused on keeping restructured jobs inside the bargaining unit.

Kaiser stated another purpose of the partnership in a federal lawsuit filed five days before the strike. In its complaint, Kaiser cited strike activity, the unions’ suspension of partnership activities and public accusations concerning unsafe staffing in support of its allegation that the unions had breached the LMP.

Kaiser sought permission to abandon national bargaining for separate local negotiations and said the partnership protected “labor stability, operational continuity, and business reputation.” Equal seats on a technology task force do not change who owns the systems, controls the data and decides whether to deploy them.

Contracts for other Kaiser unions treat staffing in the same way. A national summary or the UNAC/UHCP contract identifies among its “gains” a staffing dashboard, greater data transparency, a toolkit, committee training and an escalation process. It does not identify a new staffing minimum or an automatic obligation to add staff when the dashboard records a shortage.

None establishes a minimum number of workers or requires Kaiser to add staff when the dashboard records a shortage. That agreement does not cover the CNA advice nurses, but their experience supplies the measure: Better information about call volume does not in itself give a nurse more time with a patient.

Workers’ control would require enforceable provisions: Call-time and activity metrics could not be used for discipline; nurses would determine adequate documentation and recovery time; workers would have access to system data and validation results; deployments would require approval by those affected; and clinical staff could override software recommendations without penalty. Staffing minimums would be based on patient need and could not be reduced on the promise of technological savings.

Technology can reduce repetitive documentation, but Kaiser has shown why this cannot substitute for staffing. It cut post-call time to as little as 30 seconds and filled the interval with another patient. Without control of workload, time saved in one task becomes the basis for imposing another. Automation cannot be presented as an alternative to hiring enough nurses.

Rank and file committees independent of the union bureaucracy are needed to investigate systems already in use and unite workers across Kaiser’s bargaining units around these demands. Decisions over staffing, budgets and technology ultimately require public ownership and democratic control of the healthcare system.