SF Educators Win Protections Against AI, but Tech Expansion Continues

The CEO of San Francisco-based tech company Salesforce, Marc Benioff, wrote an article for Time in November 2024 promising that advances in artificial intelligence will lead to an “unlimited age.” His company is better known for customer service software you might use unwittingly at work than it is for AI development, but Benioff put his own spin on what Salesforce was contributing to this new horizon. He described how an education nonprofit had recently deployed its AI-powered “Agentforce” platform to provide a “virtual college counselor for high school students.” This, Benioff claimed, represents a “powerful expansion of labor potential” and enables “new ways to support students where traditional resources have been limited.”

That vision of AI as the salvation for cash-strapped educators is in line with Salesforce’s gift of $13 million to San Francisco and Oakland public school districts in October that is supposed to go to science and math education and, notably, the integration of AI tools. Salesforce has given the San Francisco Unified School District (SFUSD) over $80 million since 2012 and has made AI an explicit priority in their annual donations since 2023.  

For some workers in the district, the “labor potential” promised by these AI tools seems like it could come at the expense of their own. Their union, the United Educators of San Francisco (UESF), recently went on strike for the first time since 1979 over healthcare, wages, and special education staffing, among other issues. One of those issues was protections against the effects AI tools might have on the security of their jobs. Now, in the tentative agreement reached by the UESF bargaining team after a week on strike, workers won guarantees that AI technologies will not replace their jobs. 

In the midst of the strike, Superintendent Maria Su announced that the district had agreed to a policy on AI regulation that would “set the standard for the state.” SFUSD, however, had resisted the union’s central demand regarding the use of AI, that “any productivity gains resulting from AI implementation shall not be used as a basis for staffing reductions or increased workload expectations.” In their proposal on February 5, the district had conceded only to a “joint artificial intelligence working group” and a promise to negotiate over technology use that could impact workers.

Even as the district eventually accepted the proposals on protections against staffing cuts and a bar on the use of AI to monitor or evaluate workers, they refused demands that would give workers control over “system outputs and decision-making” when using these tools. Workers in the district still need to vote to approve the tentative agreement, as does the San Francisco Board of Education. 

These demands and the concerns that drove them have been long in the making. Alex Schmaus, a middle school educational instructional aide in the district who represents paraeducators on the union’s executive board, spoke with me in May of 2025 about how statements from Salesforce about the transformative effects of their AI raised concerns for him and other workers. “Salesforce,” he noted, “has started to push programs to take the work of counselors.” Other types of education workers could also find their jobs at risk, so it was good the union was "trying to get ahead of this," Schmaus said.

Teachers’ unions in the California Teachers Association shared some common contract demands in the coordinated We Can’t Wait campaign that began last year, including language that would guarantee workers would not lose their jobs because of new AI technologies. UESF took up these demands in bargaining and in their public messaging.

At the heart of this struggle over job protections against technology is the broader fight over staffing and the conditions that can keep educators in their jobs. The district tried to slash 151 jobs for counselors and paraeducators last May before they rolled back the cuts under pressure from teachers and families. Still, job positions remain vacant. Teanna Tillery, Vice President of Paraeducators at UESF, summarized the crisis to KQED: “With almost 400 educator vacancies and health care increasing by 20%, the writing’s on the wall. SFUSD needs to invest in all of its educators or risk losing most of us.” The problem of understaffing has been chronic. 

What might it look like for artificial intelligence to fill positions vacated by education workers? The case of Amira, an AI-powered computer program the district brought in to help third graders learn to read, offers one potential scenario. 

Spark SF Public Schools, the nonprofit formally associated with SFUSD that mobilizes philanthropic support for the district, helped launch the “Ready, Set, Read!” initiative in 2024 to improve third grade literacy. Part of that effort is the introduction of Amira, an AI-powered program that students can speak to as a way to demonstrate reading comprehension. While the homepage for the initiative describes investment in “high-dosage tutoring,” it makes no mention of AI programs that would do that tutoring. The case statement, though, describes Amira as “a digital platform deployed by teachers” that is “a scalable solution across SFUSD from a cost and human capital perspective.”

How would it save costs? Max Raynard, a fifth grade teacher at Clarendon Elementary School in the district who has used Amira with some students, describes its distinct approach to reading instruction. The program collects data and gives feedback based on students reading aloud: “Kids are in one room, they read at the same time,” Raynard explains. 

That noisy environment is different from the method Raynard was used to when he taught third grade, in which a teacher took individual time with a student to record their reading of a physical book. “From the labor perspective, that is expensive,” he notes, because the district needed to pay someone to substitute for the class while the teacher worked one-on-one. He says his impression is that most teachers prefer the one-on-one method. The school’s resistance to limiting the effects of AI comes from their motivation for using it in the first place: “With AI particularly, they want to save labor.” 

Spark SF Public Schools and district administrators have attributed improvement in reading scores to Amira, though that’s not the only benefit they point to. The president of Spark SF Public Schools also publicly highlighted its potential to be “scalable both from a labor perspective as well as a cost perspective.”

This praise for a tool that can save money for a district that claims an impossibly tight budget, specifically by reducing the expense of hiring more workers, echoes Benioff’s account of the AI future. Salesforce is the first benefactor listed on Spark SF Public Schools’ website, and last year the nonprofit distributed at least $4.9 million from Salesforce to SFUSD. Salesforce donations are not, however, funding the Amira program, according to Ginny Fang from Spark SF Public Schools. 

The vision of massive productivity growth and cost saving at the hands of AI remains pointed toward the future, whether you’re a tech mogul promoting a product, a school administrator hoping for an end to budget woes, or a teacher concerned about what this could mean for their classroom. San Francisco school workers did not wait for that future, and the protections they won will make a difference right now. Just earlier this month, the district signed a contract with OpenAI to give ChatGPT accounts to 12,000 teachers and staff. The proposal before the school board to approve the contract also states that the SFUSD will join a cohort of other districts organized by OpenAI and receive training and support from the company “in pursuit of the District’s goal to develop a comprehensive AI policy.”