Teachers Prioritize AI for Productivity, Not Student Chatbots
A new study from Stanford University offers a compelling glimpse into how U.S. educators are integrating artificial intelligence into their professional lives, revealing a clear preference for AI tools that bolster their own productivity rather than those designed for direct student interaction. Analyzing usage logs from 9,000 teachers across the nation who adopted the SchoolAI platform, researchers found that the more frequently teachers engaged with the technology, the more they gravitated towards features supporting tasks like lesson planning, grading, and content creation.
This research provides a valuable counterpoint to self-reported surveys, such as a recent Gallup poll indicating that six in ten teachers use AI for work. By instead leveraging actual usage data from the SchoolAI platform, the Stanford team aimed for a more accurate understanding of AI’s practical application in classrooms. “We all know that humans are flawed at reporting our own behavior accurately,” noted Chris Agnew, director of Stanford’s Generative AI for Education Hub and a key figure in the project.
The study meticulously tracked teachers who first joined SchoolAI between August 1 and September 15, 2024, collecting 90 days of usage data. While not all participants became consistent users, the findings still paint an encouraging picture of adoption. Sixteen percent of teachers used the platform only once, and 43 percent were short-term users. However, a significant 41 percent evolved into “regular users,” logging in between eight and 49 days over the 90-day period. A small but impactful 1 percent emerged as “power users,” engaging with the platform on 50 or more days. This collective adoption, with over 40 percent becoming regular or power users, slightly surpasses typical software retention rates, which often hover around 30 percent after three months, according to platform analytics firm Pendo.
Despite these promising adoption figures, the data suggests that many teachers utilize AI on an as-needed basis rather than embedding it into daily or weekly routines. At any given point, roughly a third of participants were active on the platform, indicating a flexible, demand-driven integration of AI into their workflows.
Crucially, the study uncovered a distinct evolution in how teachers leverage different AI functionalities. SchoolAI offers a diverse suite of tools, from student-facing chatbots to teacher productivity aids like lesson generators, grading assistants, and quiz builders, alongside general teacher chatbot assistants. Initial, lighter users often explored student-facing chatbots, but as teachers became more consistent with the platform, their focus unequivocally shifted towards teacher support features. Power users, in particular, largely bypassed student-focused tools from the outset, dedicating over 80 percent of their time to teacher productivity tools and chatbots.
Agnew interprets this trend as a clear embrace of a “human in the loop” approach to AI in education. He emphasized that this model empowers teachers to filter and contextualize AI-generated output through their extensive professional experience, informing their practice and classroom strategies. This stands in contrast to deploying AI tools directly to students, particularly younger learners who are still developing their expertise and judgment.
The timing of AI usage also presented an unexpected pattern. While the time-saving potential of AI in tasks like grading and lesson planning might suggest after-hours use, the data revealed that most teachers accessed AI tools predominantly on weekday mornings. Although the study did not delve into the reasons behind this timing, Agnew, a former teacher himself, hypothesized that it could signify AI serving as a collaborative partner for educators as they prepare and orient themselves for the school day, perhaps for brainstorming or material preparation before classes begin.
As the educational landscape continues to evolve with AI, researchers acknowledge that tracking active usage days alone doesn’t capture the full impact of AI-assisted work. The next phase of this ongoing project will delve deeper, analyzing the content of teacher-AI interactions and exploring how students themselves engage with the platform, which may reveal entirely different patterns of adoption and utility.