Share:

When automated triage fails, patients are the ones who pay the price

A protest rarely begins without a story behind it.

In the case of Kaiser Permanente, that story has a name and a face: it belongs to a therapist who looks at her patients and thinks thank God they are still alive.

Ilana Marcucci-Morris, a licensed clinical social worker at Kaiser’s outpatient psychiatry clinic in Oakland, California, began noticing something different starting in January 2024. People were arriving in critical condition, weeks later than they should have. Cases that, in her assessment, needed emergency care long before they ever made it to her office.

This is not an isolated scenario, and it is no coincidence that roughly 2,400 mental health professionals in Northern California walked off the job in a one-day strike, represented by the National Union of Health Care Workers, or NUHW. The core of the discussion goes beyond a simple operational change. What is at stake is a question the tech and healthcare industries will need to answer sooner or later: how far can artificial intelligence and automated systems go when it comes to triaging people experiencing severe psychological distress? 🤔

What changed at Kaiser Permanente and why it matters

Kaiser Permanente is one of the largest integrated healthcare organizations in the United States, with millions of members and a structure that combines health insurance, hospitals, and clinics under one umbrella. That model is often praised for its efficiency and ability to coordinate care. But it is precisely within that structure that the most recent point of friction emerged.

Before the change, licensed professionals were almost always the first point of contact for patients with behavioral health concerns at Kaiser, according to Marcucci-Morris. Starting in January 2024, the healthcare giant introduced a new triage process for first-time patients. The new system began using administrative staff, who are not licensed professionals, to ask scripted yes or no questions and assess the severity of patients’ conditions and the urgency of their care.

Around the same time, Kaiser also rolled out a different form of triage for some patients: so-called e-visits, which are essentially online questionnaires patients fill out before being scheduled with a licensed healthcare provider.

What Ilana and her colleagues observed was that patients in serious condition were being funneled into wait queues that did not reflect the actual urgency of their clinical situation. The automated system was not picking up on what an experienced professional would catch within minutes of conversation. And the consequences showed up in treatment rooms weeks later, with people arriving in an advanced state of deterioration.

Five licensed Kaiser therapists confirmed to The Guardian that, since the new assessment process was implemented, patients with high-risk cases have been waiting longer to receive care. At the same time, lower-risk patients are being fast-tracked to clinician appointments, clogging a system that was already overwhelmed.

More than 70 documented cases of system failures

Since January 2025, therapists have reported more than 70 examples of situations where Kaiser’s mental health triage system led to negative outcomes for patients. These cases were compiled in an administrative complaint filed by the Northern California NUHW with the California Department of Managed Health Care.

A separate but similar complaint was also filed by the NUHW in Southern California in 2025. In the Southern California complaint, the union describes how administrative employees ask patients questions about suicidal and homicidal thoughts, then enter the information into a software tool. The tool’s algorithm then generates a score and a suggested response to guide the employee in scheduling the patient for further care.

Receive the best innovation content in your email.

All the news, tips, trends, and resources you're looking for, delivered to your inbox.

By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.

The union alleges that Kaiser is using an algorithm to make triage decisions, in violation of state law. The company denies that this triage counts as a clinical assessment and says its administrative staff are not making assessments or clinical determinations. It remains unclear whether a similar algorithm is also being used in Northern California, although the union suspects it is, and Kaiser has not clarified the matter.

What Kaiser says about all of this

In a statement sent via email, Kaiser Permanente said the NUHW leadership has been making misleading claims about access and care. The company stated that AI and administrative staff are not conducting any assessments, making clinical determinations, or performing clinical triage.

The statement also highlighted that administrative employees are trained to escalate cases to the clinical team through an immediate warm handoff to a crisis therapist. Kaiser further stated that it is expanding its workforce, not reducing it, although NUHW representatives say they believe the number of triage therapists has decreased significantly.

The company also stated: We believe AI can be helpful when it supports clinicians, reducing administrative work or improving efficiency, but it does not replace clinical judgment or human assessment.

Kaiser said it provides high-quality care in a timely manner to meet its members’ needs and that its members receive non-urgent mental health appointments, on average, faster than what the state requires.

A track record of mental health access problems

It is worth noting that Kaiser has previously faced state and federal scrutiny over providing timely access to mental health services. In 2023, the company agreed to a $200 million settlement with the state of California to resolve investigations into behavioral health care delays.

More recently, in February 2026, the U.S. Department of Labor announced a $31 million settlement with Kaiser over similar allegations. The department stated that Kaiser used patient responses to questionnaires to improperly prevent patients from receiving care. Kaiser also agreed to reforms, as part of the Department of Labor investigation, that would reduce wait times for appointments and expand access to quality care.

For the workers, the use of automated systems in triage feels like yet another layer added on top of problems that were never properly addressed, not a genuine solution to them.

The complexity of mental health triage

Kaiser employees pointed to several ways in which questionnaires and administrative staff can fall short. And this is one of the most important parts of this entire discussion.

Mental health triage is complicated. Relying on workers who are not licensed professionals and who follow limited scripts has major limitations. Therapists often need to draw on their expertise to uncover the real meaning behind a patient’s statements.

Harimandir Khalsa, a Kaiser therapist in Walnut Creek, California, gave a clear example: if someone mentions suicidal thoughts, a healthcare professional needs to understand whether those thoughts are active or passive. Has the person already considered a method? If they say they are not sure what they are going to do, what are they referring to? The answers to these questions are rarely straightforward.

On the other hand, patients may self-diagnose in a way that exaggerates their symptoms, taking critical resources away from those who need them more urgently.

Khalsa also revealed that her team of nine staff members has been reduced by two-thirds over the past two years. She still cannot imagine leaving her role because she loves using her decades of clinical and research experience to help people at such a vulnerable moment in their lives. But as she watches the triage team in her department shrink around her, she cannot help feeling anxious about the future: Am I next? What is my future?

The real impact on day-to-day care

Therapist Carolyn Staehle started in Kaiser’s intake and assessment department in Pleasanton, California, in May 2023. After the new system was implemented, Staehle, whose role at the time was supposed to focus on non-emergency cases coming from triage, found herself seeing far more people with dangerous delusions and severe suicidal thoughts.

They needed me to call an ambulance for them because they could not guarantee their own safety or work through a safety plan, she says.

More recently, Staehle has been working on a crisis team meant for higher-severity cases. They continue to receive patients who do not need that level of care. This clogs up and slows down the work, meaning people in immediate and desperate need may not be able to get care in time.

It is not the same level of care as being assessed by a licensed therapist, Staehle said. It takes longer for each patient to figure out whether they are going to be a danger to themselves or others, or whether it is an emergency or not. In practice, we have to spend time handling fundamental things that triage used to take care of.

The fear of being replaced by artificial intelligence

Kristi Reimer, a licensed psychologist who used to conduct mental health triage assessments at Kaiser’s Walnut Creek facility, said she left her position preemptively because she saw the writing on the wall. She stated that she would not have moved to another department if Kaiser had not changed the nature of its mental health assessment system so drastically.

In an internal 2025 survey of Kaiser mental health workers in Northern California, obtained by The Guardian, more than one-third of employees reported that Kaiser has already implemented AI or other technologies that they fear could negatively affect their work or the care patients receive. Nearly half of the workers said they were somewhat or very uncomfortable with the introduction of AI tools into their clinical practice.

Many were particularly concerned about transparency and data retention policies tied to the use of Abridge, an AI-powered note-taking software. A Kaiser representative said the tool is not mandatory for employees and requires patient consent. 😔

Why a licensed therapist makes a difference in triage

A patient’s first point of contact when seeking mental health help can determine whether they will see a licensed clinician, and also the type of appointments they will receive, according to the NUHW. That is why the union is so concerned about recent changes to Kaiser’s assessment process and is pushing for more information about how the company uses technology in initial evaluations.

Tools we use daily

AI-based triage systems tend to operate based on historical patterns. They learn from data on past cases and try to classify new ones based on that accumulated experience. The problem is that this model works well when patterns are stable and the data is rich enough to capture the real diversity of cases. In mental health, that is a fairly fragile assumption.

People in crisis do not follow scripts. They minimize symptoms, feel ashamed to reveal what they are experiencing, respond defensively, or are so exhausted they can barely describe what they are going through. An algorithm trained to identify keywords or response patterns can easily miss the very cases that are most severe.

That does not mean technology has no place in this field. Digital tools can support mental health professionals in very valuable ways: organizing patient histories, identifying patterns over time, facilitating follow-ups between appointments, or flagging sudden behavioral changes. What Kaiser’s workers are questioning is not the presence of technology itself, but the role it is playing within a sensitive clinical process, without adequate oversight.

What this debate reveals about the future of AI in healthcare

The Kaiser Permanente case is far from isolated. Around the world, healthcare systems are experimenting with ways to integrate artificial intelligence into their workflows, often arguing it will increase efficiency, reduce costs, and expand access. These are legitimate goals, especially in a context where the demand for mental health services has skyrocketed in recent years and the supply of qualified professionals simply has not kept pace.

The problem begins when efficiency becomes the primary benchmark in contexts where what is at stake is someone’s life.

What the Kaiser workers’ protest makes abundantly clear is that introducing AI into clinical processes needs to be accompanied by robust governance, transparency about how the systems work, and real accountability mechanisms for when something goes wrong. It is not enough to claim the technology was tested or that the data shows it works on average. In mental health, the cases that fall outside the average are exactly the most critical ones.

The next step: a new contract and concrete guarantees

For now, Staehle and other workers are focused on ratifying a new contract with Kaiser and securing the company’s commitment not to replace licensed social workers and other mental health professionals with artificial intelligence.

The discussion happening in California has implications far beyond that state’s borders. It is foreshadowing a debate that will become central in the years ahead: how do we ensure that the incorporation of artificial intelligence in healthcare is done ethically, safely, and in a way that is truly patient-centered?

That answer is not going to come from a tech lab. It will need the voices of those working on the front lines, of those who know every patient’s face, and of those who know from experience that some warning signs are only visible to someone who knows where to look. 👁️

As Marcucci-Morris put it plainly: human work needs to stay with human beings.

Picture of Rafael

Rafael

Operations

I transform internal processes into delivery machines — ensuring that every Viral Method client receives premium service and real results.

Fill out the form and our team will contact you within 24 hours.

Related publications

Performance and Growth: Nvidia, AI Agents, and Data Centers

Nvidia accelerates revenue with data centers, GB300 NVL72, and Rubin; efficiency and AI Agents demand drive record growth and profit.

AI and Copyright: Supreme Court Denies Copyright Protection for Artistic Creation

Supreme Court rejected the AI-generated art case; in the US only humans can hold authorship — a direct impact on

AI Reveals the Identity of Anonymous Social Media Users

Vulnerable anonymity: how modern AI unmasks social media profiles and why this threatens your online privacy.

Receba o melhor conteúdo de inovação em seu e-mail

Todas as notícias, dicas, tendências e recursos que você procura entregues na sua caixa de entrada.

Ao assinar a newsletter, você concorda em receber comunicações da Método Viral. A gente se compromete a sempre proteger e respeitar sua privacidade.

Rafael

Online

Atendimento

Calculadora Preço de Sites

Descubra quanto custa o site ideal para seu negócio

Páginas do Site

Quantas páginas você precisa?

4

Arraste para selecionar de 1 a 20 páginas

📄

⚡ Em apenas 2 minutos, descubra automaticamente quanto custa um site em 2026 sob medida para o seu negócio

👥 Mais de 0+ empresas já calcularam seu orçamento

Fale com um consultor

Preencha o formulário e nossa equipe entrará em contato.