Yale labor unions push for AI protections

Helen Helpwell smiles at you as she speaks. She enunciates every word. She is “performing correctly,” according to a note on her personal webpage, when she responds to questions in a “concise, confident, warm, and policy-aligned” manner.

“Think of me as your shortcut to the right person, the right process, and the right place to start,” she says, explaining her role as a “research administration concierge” at the Yale School of Public Health (YSPH) — a job she’s had for about a year.

Helen doesn’t have the training to take over the former job of Dana Greene, who retired in May after 32 years as a senior administrative assistant at YSPH.

Still, Greene is worried about her former employer’s decision to work with Helen. Unlike Greene, Helen doesn’t decorate her desk with union buttons. Helen is always available, and she never asks for a raise.

That’s because Helen isn’t human.

She — it — is an artificially intelligent chatbot, accompanied by an AI-generated video of a young white woman, not a hair out of place in her shiny brown bob.

“Helen” is just one example of how Yale has been delegating work to AI-powered chatbots.

According to Yale officials, AI is not intended to replace or reduce the number of human staff employed at the university. “Yale has not reduced positions in favor of AI tools but instead is focused on ways to make more time for staff to do human-centered work or to increase external access to Yale’s resources,” an unidentified university spokesperson wrote in an email to the Independent.

“The contributions of Yale’s staff are essential to the university’s mission, and AI is intended to support their expertise, judgment, and service to the community, not replace them.”

Even so, Yale’s approach to AI has raised alarm within Yale unions about the future of human labor at the university. The number of unionized clerical and technical workers at the university has reportedly declined by 179 in the last year and a half.

Concerns about job security in an AI-saturated future have become a flash point in the ongoing contract negotiations between Yale and its clerical and technical workers union, Local 34. In early October, the union released a report lambasting Yale’s rollout of AI technology.

Local 34’s report claims that “AI has led to profound changes in the fields of grant writing, memo writing, email outreach, customer service, information retrieval, and many other tasks that Local 34-UNITE HERE members do on a regular basis,” without input from the unions themselves. According to union spokesperson Ian Dunn, the report was produced entirely by human labor. (Side note: so was this article.)

The report includes personal narratives from a number of current and former union members, including Greene, describing how they were asked to incorporate AI into their work. One administrative assistant, Carolyn Kyrtopolous, described how generative AI cut the time it took her to produce scholar biographies by three quarters. Another assistant, Jennifer Ongley, wrote that the technology would produce a “confusing jumble of word salad” with errors that ended up taking extra time to fix.

A key claim in the union’s report, which Yale refutes, is that the university is reducing human labor by way of “passive automation” or “passive replacement” — using AI to replace or redistribute the labor of employees who quit or retire from their positions.

The report cites a recent hiring freeze and retirement incentive as examples of how Yale has “cut the total amount of union workers across departments and units.” It argues that “by supplementing a smaller workforce with AI, the University can simply avoid hiring workers to replace those who have retired or have quit, thereby reducing the amount of union work needed to fulfill the same amount of work that was done previously.”

Local 34 is calling for the university to work with unions on a “responsible transition” to an AI-assisted workplace, as well as to protect existing jobs, bolster layoff benefits, and raise the salaries of people trained to incorporate AI into their work.

According to the report, the number of Local 34 members has declined from 4,112 in April 2025 to about 3,933 now.

Meanwhile, Yale appears to be hiring for at least 17 administrative assistant positions, according to its public job postings.

Since August 2024, Yale has been rolling out an internal proprietary platform called Clarity, which provides Yale employees access to versions of Anthropic’s Claude, OpenAI’s ChatGPT, and Google’s Gemini that are designed to adhere to certain security standards.

Yale’s website states that the platform, among other tools, “enables Yale faculty, staff, and sponsored students to utilize secure and scalable API offerings allowing AI capabilities to be built into applications and workflows.” The website also outlines how Clarity allows “Yale faculty, staff, and sponsored students to create and manage Custom Agents at no cost.”

A few Yale departments have launched custom agents, some designed for public use and others intended to assist staff. In addition to “Helen,” there’s “Harvey” at Yale’s medical library, which can “search across the catalog, PubMed, and our website.” There’s “Reggie” at the Registrar’s Office along with a nameless advising chatbot, which can each search for information posted to the university’s website. And there’s Yale Finance Chatbot, an internal agent described as “the best source for answers to most self-service Finance-related questions, including information about Yale’s finance-related policies, procedures, processes, forms, reporting, and other finance-related content.”

Greene said that starting in the winter before her retirement in May, her supervisors began to repeatedly encourage her and her colleagues to make use of Clarity.

At team meetings, she said, she was told to ask questions to a chatbot before coming to a manager. Her takeaway from the announcements, she said, was that “it’ll be learning from you, the answers you give. It will get stronger as it learns from you.”

“Just because they were strongly encouraging it, I said all right, let me see what this is all about,” she recalled. Part of her job involved booking travel arrangements for faculty and students; she recalled asking the chatbot about “government rules related to travel.” She said that the chatbot would cite Yale’s official policies and procedures, the same resource that Greene would have otherwise consulted. “I guess the promise was that it gives it back to you faster,” Greene said.

Greene worked as one of two administrative assistants at the Yale School of Public Health’s Chronic Disease Epidemiology department. She worked with about 15 faculty members, in addition to a rotating pool of postdocs, students, and study collaborators.

On a day-to-day basis as a senior administrative assistant, Greene would coordinate events, reconcile university credit card expenses, help faculty plan for course logistics, and generally respond to “any needs of the department” as they come up.

So far, Yale’s public-facing chatbots have comparatively narrow roles. As “Helen” explains in the pre-set welcome video, “My role is to help you and other faculty and staff find quick, accurate answers about pre- and post-award administration — things like budgets, hiring, or policies.”

The chatbot doesn’t have access to Workday, the university’s billing and personnel platform. It can rattle off travel policies, but it can’t actually book a flight. Not yet, at least.

Greene emphasized that technical limitations are not the only way in which AI cannot replace her former job. “Helping to make the school run, you can’t do that behind a screen,” she said. “You have to connect with people.”

She described often turning to her colleagues in other departments to brainstorm solutions to tricky situations not covered in any procedural handbook. “You can’t get that online, the tips and the strategizing and the figuring out another way,” Greene said.

When Greene retired, she recalled that a group of YSPH students brought her a card — “this huge card that I still have in my house” — thanking her for things she didn’t realize they had noticed.

“They loved the way I ordered a variety of lunches. I heard them when they said I want more of this, I want less of this,” Greene said.

She put in effort, she added, “just to be able to put a smile on somebody’s face.” In the office kitchen, she would frequently post wellness and mental health tips. On her desk, she arranged faux sunflowers alongside “my union gear,” buttons and pins from Local 34. She would print out quotes that moved her and post them on a board titled “Words of Wisdom.”

Over 32 years, she became a source of institutional memory, from “seeing things change for good and not so good.”

The Yale spokesperson argued that AI might actually enable employees to “focus more on human-centered work that is personally fulfilling,” by making other tasks “more efficient.”

“Because AI is transforming many parts of the professional workforce and jobs nationwide, we have observed a strong interest in hands-on workshops and other training opportunities,” the spokesperson wrote. Individual schools and teams have the leeway to implement AI according to their particular needs, they wrote. “Part of this strategy involves careful attention to the human aspects of our mission and the very essential social fabric of a residential university.”

AI has clearly emerged as a point of tension, however, between the university’s leaders and its unions. According to Local 34’s report, Yale’s approach to AI has “made current negotiations more challenging” as the union seeks a new contract. The report calls on Yale to “agree to maintain at least its current level of 3,933 Clerical and Technical positions over the next five years” as a means to “give workers more confidence that innovative AI use will augment University operations instead of eliminating jobs.”

Greene said she worries about what a more automated version of her workplace might look like in a decade. “I’m really, really concerned about the coworkers, the colleagues I left behind,” said Greene. She worries that new employees might be trained by robots, that people might generally work in increased isolation from other humans. “What kind of atmosphere would that be?” she asked.

“I’m not totally against technology,” Greene added. “You gotta have it. And it has done wonders for us… But to do a job like our job, that human connection cannot go away. It cannot be replaced and should not.”

“It was a sense of pride for me,” she said — “just the relationships and with the staff and faculty there. You can’t measure that.”

This story was first published Oct. 28, 2026 by New Haven Independent.