2025 Study: What AI Agents Mean for Educators
December 10, 2025 * 8:32 PM GMT+8
Millions of people now use AI agents for learning and personal productivity. In this study, AI agents are defined as online assistants that can plan and execute multi-step tasks with minimal supervision.
In 2025, Amazon, Google, Microsoft, and OpenAI all launched or expanded digital assistants. A Harvard researcher, working with Perplexity AI, analyzed hundreds of millions of queries from Comet, the company's AI browser and assistant launched in July 2025. The findings were posted online the week of December 10, 2025 and have not yet been peer-reviewed.
Key findings at a glance
- Adoption skews to early tech users and to wealthier, more highly educated countries.
- More than 70% of users work in knowledge-heavy fields (academia, finance, marketing, entrepreneurship). Fewer users come from fields that interact heavily with the physical environment (e.g., energy, agriculture).
- 36% of all agent tasks were productivity and workflow: drafting documents, filtering emails, summarizing investment or information sources, creating calendar events.
- 21% of tasks were learning and research: summarizing course materials, synthesizing sources, clarifying concepts.
- Other common uses: shopping help, travel planning, and job-related searches.
- By context: 55% of queries were personal, 30% work, 16% education.
- Usage evolved: many users started with simple personal tasks (travel, media), then moved into heavier tasks tied to productivity, learning, and careers.
Why this matters for educators
The pattern is clear: people start small, then shift to work that saves serious time. That maps directly to teaching. Begin with a few personal wins to build comfort, then apply agents to lesson prep, grading support, research synthesis, and admin work.
The skew toward knowledge work also signals where agents deliver the most value right now: reading, writing, organizing, and analysis. That's daily life in education.
Practical ways to apply the findings in your workflow
- Productivity and workflow (36% in the study)
- Draft syllabi, rubrics, parent emails, permission slips, and meeting agendas.
- Filter inboxes: auto-label parent vs. admin requests, generate first-pass replies, and extract action items.
- Summarize policy updates, research articles, and long PDFs into bullet-point briefs.
- Create calendar events with attached checklists for recurring tasks (grading cycles, IEP milestones, department deadlines).
- Learning and research (21% in the study)
- Turn course materials into leveled readings, guided notes, and practice questions.
- Generate study guides, concept maps, and quiz banks from your unit outlines.
- Synthesize multiple sources into balanced overviews with citations you can verify.
- Other helpful tasks surfaced by the study
- Plan travel for field trips or conferences (lodging, itineraries, budgets).
- Run job-related searches: role descriptions, interview question banks, portfolio outlines.
Get started in 30 minutes
- Pick one agent you'll actually use daily (from your district's approved list if applicable).
- Choose three repeatable tasks: "summarize PDFs," "draft parent updates," "make lesson outlines."
- Write a simple template for each task. Example: "Summarize this PDF into 5 bullets, 3 discussion prompts, and a 10-question quiz. Flag bias or gaps."
- Connect only what you need (calendar, email). Avoid student PII and sensitive data.
- Track time saved for two weeks. Keep what works, drop what doesn't, and iterate.
Academic integrity, privacy, and policy
Use agents to support learning, not to replace it. Keep identifiable student data out of prompts unless your system provides compliant, approved pathways. Align with your school or district's policy.
Helpful references: UNESCO guidance on generative AI in education and the U.S. Department of Education AI report.
Equity and access
Adoption is higher in wealthier countries, according to the study. If you lead programs, consider low-friction on-ramps: short teacher workshops, shared prompt libraries, and classroom routines that teach AI literacy alongside content.
Keep learning (for educators)
- AI courses by job - find options aligned to education roles.
- Prompt engineering resources - build reusable templates for lessons, grading, and research.
Bottom line
The study shows people begin with simple personal queries, then shift into heavier lifts that save real time at work. For educators, that path makes sense: start with one or two low-risk tasks, prove the value, and expand into lesson design, feedback workflows, and research support-with clear guardrails.
The tools are here. The workflow is yours to shape.
Your membership also unlocks:
