A Pilot Rooted in Curiosity
What happens when AI becomes a teacher’s assistant, a student’s tutor, and a school’s data collector—all in one? We wanted to find out. We wanted to explore what might be possible.
What would happen if teachers and students were given the opportunity to test an AI tool with structure, some support, and space to play? What might they discover?
Curious, Not Convinced: Why We Tried Khanmigo
Like many educators, we were hearing a lot about AI—its potential, its risks, and its growing presence in classrooms. But figuring out how to explore AI in a thoughtful, low-stakes way isn’t necessarily easy.
We were curious about Khanmigo, an AI-powered tutor and teaching assistant developed by Khan Academy, because it offers both student- and teacher-facing tools. It has features for lesson planning and idea generation, and it provides real-time student support.
What stood out was that Khanmigo isn’t focused on delivering quick answers. Instead, its prompts often encourage students to think through problems, explain their reasoning, and reflect as they arrive at the answer independently. Because it is integrated into Khan Academy’s content-rich platform covering subjects like math, history, coding, and computer science, it also offers some built-in structure for classroom use.
A Glimpse at the Khanmigo Pilot
Over two pilot phases, Michigan Virtual’s Research and Development team partnered with several Michigan school districts to see what might happen if teachers and students explored Khanmigo together.
Pilot 1—Spring 2024:
- 19 teachers, 687 students, 14 school districts
- Focused on Algebra I (9th-grade)
- Emphasis on Khanmigo as a math tutor and support tool
Pilot 2—Fall 2024-Spring 2025:
- 24 teachers, 1102 students, 8 school districts
- Open to all subjects and grade levels
- Greater emphasis on teacher-facing tools, instructional use cases, and student behavior
We gathered pre- and post-survey data, facilitated professional learning sessions, and asked participants to reflect on their experience. Their feedback about Khanmigo and AI tools in general helped us understand how comfortable they were with using AI tools, how frequently they used them, and where AI tools had the biggest impact in their teaching and their classrooms.
Real Feedback, Real Classrooms
So what did educators have to say?
Teacher Tools? Surprisingly Helpful.
Khanmigo proved especially valuable for brainstorming questions, writing lesson hooks, reviewing content, and suggesting student activities.
“I used it to get ideas for introducing new concepts and making learning more applicable to the real world. It felt like having a planning partner.”
While the survey participants differed, teachers who responded in Spring 2025 expressed more confidence in using AI for planning than those in Fall 2024. By the end of the second pilot, all responding teachers reported at least some confidence in lesson planning, with the number of teachers reporting they felt “very confident” increasing from pre- to post-survey across several categories, including:
- Supporting diverse learners
- Drafting emails & communication
- Content creation & curation
- Grading, assessment, & feedback
- Lesson planning
Students? Curious…But Complicated.
From the teacher’s perspective, student engagement with Khanmigo varied. Some students, particularly quieter students or those working independently, embraced it and used it productively. Others saw it as a workaround—a way to get answers quickly or to check their work, rather than a tool to deepen their understanding.
“My students struggled to ask Khanmigo the correct question. They didn’t know what they didn’t know—where to start or what to ask.”
Several teachers described this as less of a problem with the actual tool, but more of a reflection of student readiness and maturity. Still, several agreed that with clear modeling and structure, students were more likely to engage with it as intended.
“We have to teach kids how to use it and how not to use it—and be clear about the intended purpose.”
What the Data Told Us
- According to the teachers who responded in the Fall of 2024, compared to those in the Spring of 2025, student use of AI tools increased, moving from monthly to weekly use for most classrooms.
- Teachers’ familiarity with AI tools grew, with a jump in the number who reported using AI several times a week, and one even reported daily use.
- Professional learning priorities included personalizing learning, lesson planning, and supporting diverse learners.
Some Speed Bumps Along the Way
Not everything worked perfectly. Technical challenges like typing equations or interpreting math prompts made things clunky at times. Some students were frustrated with how “chatty” Khanmigo’s AI chatbot tutor was. Others were quick to use it for shortcuts.
And some challenges weren’t technical—they had more to do with timing and student mindset. Teachers noted the importance of starting the year with an AI tool like Khanmigo, rather than trying to integrate it midway through the semester. Teachers also emphasized the importance of setting clear expectations and guidance as to when and why to use the tool, key to ensuring students have the foundational knowledge to understand how to use Khanmigo and other AI tools appropriately.
“Students don’t want to use it how we would like them to. We have to model appropriate usage for them consistently.”
“It worked better for older or more self-motivated students.”
“AI may need to be taught progressively, with different levels of understanding increased gradually from year to year in school.”
So, Where Do We Go From Here?
Throughout the two pilots, one thing became clear: there is real potential, but only if we stay intentional.
Teachers told us they want help with things like:
- Personalizing learning (top interest in both pilots)
- Using AI for lesson planning and brainstorming
- Supporting diverse learners with more targeted content
But they also flagged what’s missing:
- Clearer policies and norms for student use
- Tools that are more visual and interactive, especially in math
- Better examples of what “good AI use” looks like
So, now we’re asking big questions:
- Is AI a shortcut or a scaffold?
- How do we teach students to use it ethically and wisely?
- What kind of support do teachers need to make the most of it?
If you’re a district leader, teacher, or coach wondering how AI could support your work, know you’re not alone. These are big, complex questions. But trying something small, with proper support, can make a world of difference.
We’re not claiming Khanmigo, or AI in general, is the answer. But AI might be one exciting tool in a growing toolbox. And maybe, just maybe, it can help make learning a little more personal, creative, and supported—if we use it with care and intentionality.
Your Turn: What Role Should AI Play?
We’d love to hear how your school approaches AI—cautiously or creatively. We’re happy to share what we’ve learned, what we’d do differently, and what we’re still figuring out.
Because this conversation is just beginning.
