Hey everyone! I just posted a new tutorial video walking through my complete workflow for creating AI tutors using Flint, and I wanted to share some additional insights here in our community. If you want to try your hand at creating a Flint Tutor but you're not sure where to start. Check out my free TPT download to get you moving in the right direction. Why Flint?
I've been experimenting with various AI tools for the classroom, and Flint has become one of my go-to platforms for creating assessments. What makes it special? It's specifically designed for education, which means it's built with student learning in mind—not just generating generic AI responses.
The Free Account is Surprisingly Robust
Before diving into the workflow, let me address the elephant in the room: yes, this works on the free account. I currently have access to 80 student seats and have only used about 40 of them. I've created numerous activities without hitting any limits on the number of tutors or quizzes I can build. That's pretty incredible for a free tool.
My Workflow: Quality Over Speed
While Flint offers two ways to build activities (manual or with their AI assistant "Sparky"), I've found that using Sparky with a well-structured approach gives you the best results. Here's what I do:
- Step 1: Upload Your Resources
I start by syncing my Google Drive and uploading the PDF resources I want the AI to pull from. For my Eastern Europe quiz example, I uploaded multiple PDFs related to the content. This gives the AI a solid foundation to work from.
Pro tip: Don't skip this step. The quality of your source materials directly impacts the quality of questions generated.
- Step 2: The Secret Sauce—Detailed Objectives
Here's where most people go wrong, and it's the most important part of the entire process: your objectives need to be incredibly specific.
I maintain a Google Doc with custom prompts and detailed learning objectives for each unit. In the video example, even though I was creating a quiz on the Eastern Balkans and Eastern Plains, I included ALL of my Chapter 12 objectives in the prompt. This gives Sparky a comprehensive understanding of what students should actually know.
Why does this matter? If you let the AI just go off your sources without clear objectives, you'll get questions—but they might not align with what you're actually teaching or assessing. Students get frustrated when the AI asks about tangential details instead of core concepts. Clear objectives fix this problem.
- Step 3: Use Plug-and-Play Prompts
I don't reinvent the wheel every time. I've created several prompt templates that I save in Google Docs and simply copy-paste into Sparky. These templates include:
- My standard initial message to students
- Proficiency-based grading criteria
- Time limits and format specifications
- Scaffolding instructions
This makes the creation process quick while maintaining consistency across all my activities.
- Step 4: Preview and Refine
Once Sparky generates the activity, I always run a simulation at different proficiency levels. The preview shows you what students will experience, though it tends to move faster than the actual student experience.
During this phase, I'll ask Sparky questions like "Are you going to scaffold any of this?" or request adjustments to make the activity better. The AI can break content into manageable chunks, adjust difficulty, and modify the approach based on your feedback.
- Step 5: Integration with Google Classroom
Once I'm satisfied with the activity, I create it and share it with my students. I have everyone organized into my World Geography group on Flint, and then I can easily copy the link to post in Google Classroom as an assignment.
My Proficiency-Based Grading Setup
I've moved away from traditional A-B-C-D grading for these AI tutor activities. Instead, I use four proficiency categories that align with my standards-based gradebook. This gives students clearer feedback on their understanding and makes it easier to identify who needs additional support.
The beauty of Flint is that it can adapt the quiz difficulty and provide feedback based on these proficiency levels automatically.
Real Talk: What I've Learned
It takes time upfront, but saves time overall. Yes, creating that first well-crafted prompt and setting up your objectives takes effort. But once you have your template, you can crank out high-quality AI tutors in 10-15 minutes.
Students respond well when it's done right. The key difference between a frustrating AI tutor and a helpful one comes down to those objectives. When the AI knows exactly what to focus on, students feel like they're being assessed on what they actually learned in class.
The preview isn't perfect. I mention this in the video, but the preview simulations don't always perfectly match what students experience. I recommend having a student or two test it before rolling it out to your whole class.
Questions for the Community
I'd love to hear from you all:
- Have any of you tried Flint yet? What's been your experience?
- What other AI tutoring platforms are you using?
- How are you handling assessment in the age of AI?
Want to See It in Action?
Check out the full video tutorial on our YouTube channel where I walk through every step of this process in real-time. You'll see exactly how I upload files, craft my prompts, and refine the activity.
And as always, if you have specific questions about implementing this in your classroom, drop them in the comments here or on the video. I'm here to help!
Resources mentioned:
Keep innovating in your classrooms, and don't be afraid to experiment with these tools. The future of education is being written right now, and we're the ones holding the pen.
—Jeff
AI Social Studies Lab