This is part 2 in a 12-part series on How Primary and Secondary Schools Use AI. The goal is to provide educators with a roadmap for planning AI usage in their schools.
Unfinished learning remains one of the biggest challenges in K–12 education. Teachers see it daily. Students struggle with foundational skills, read below grade level, or stall when a task requires multiple steps. Tutoring has always been one of the most effective ways to help, but finding enough trained tutors to support entire districts is a constant obstacle.
AI tutoring is beginning to change what schools can offer. These tools extend teacher capacity by supporting students in the space between whole-class instruction and limited one-on-one help. They give instant explanations, walk students through complex problems, adjust support in real time, and stay available long after the school day ends.
When paired with human guidance, this hybrid human-AI tutoring model is proving remarkably effective.
A – What It Is
AI tutoring refers to the use of intelligent, adaptive tools that provide guided practice, explanations, feedback, hints, and error analysis to students as they work through academic tasks. These tools typically fall into three categories:
1. AI-Powered Step-by-Step Math and Literacy Tutors
These platforms function like a patient, always-available instructor. Students can ask questions, get explanations in multiple formats, or receive hints that gradually lead them to the solution. The AI tutor detects patterns: common errors, misconceptions, or areas of mastery.
2. Interactive Problem-Solving Assistants
Beyond tutoring specific assignments, these AI tools help students work through new problems. They often include built-in whiteboards, multimodal input (speech, writing, pictures), and the ability to break a complex task into manageable parts.
3. Hybrid Human–AI Tutoring Systems
This is more of an approach than a set of tools. The first two categories refer to tools students can use on their own, with varying degrees of teacher involvement. Here we are talking about the teacher or paraprofessional using AI tools together with students interactively. AI handles practice and explanation, while the human tutor reviews student work, provides motivational coaching, and intervenes when students need conceptual reteaching. Research shows this hybrid approach produces the strongest results, especially for students who need intensive support.
B – Why It’s Important
AI tutoring is important for four major reasons: access, equity, time, and instructional quality.
1. Expands access to tutoring for students who need it most
The students who benefit most from tutoring are those behind grade level, multilingual learners, and students with disabilities. They are often the hardest to reach consistently. AI tutors operate 24/7, giving students extra practice whenever they need it, even if they can’t stay after school or secure a tutoring slot.
2. Increases learning time in a way that’s financially sustainable
High-dosage human tutoring is powerful but expensive. AI tutoring can fill gaps between human sessions or replace less efficient practice time. Districts can extend the benefits of tutoring without overspending, especially when budgets are tight.
3. Gives students immediate feedback
A teacher might not reach every student during independent work time, but an AI tutor is always available. Students get explanations right away, not days later when a teacher returns graded work.
4. Frees teachers to spend time where human expertise matters most
When AI tools handle the mechanical aspects of tutoring such as step-by-step hints, repetitive explanations, identifying common errors teachers can shift their energy toward conceptual instruction, relationship-building, and personalized intervention.
C – How It’s Being Used
AI tutoring is being implemented in schools, districts, and entire countries. The most successful implementations are those where educators remain actively involved in reviewing data, coaching students, and using AI as a tool rather than a replacement.
Below are case studies that show what AI tutoring looks like in action.
Case Study #1: Indiana Department of Education (United States) – Expanding High-Dosage Support Through Statewide AI Tutoring Pilots
Focus: Statewide AI Tutoring Pilots for Math & Reading
Heroes: Indiana math and literacy teachers, district instructional leads, school intervention teams, and state digital learning coordinators
What They Did
Indiana launched statewide pilots allowing districts to adopt AI-powered tutoring and instructional tools as part of high-dosage tutoring. Teachers used AI dashboards to see which skills students practiced, where they struggled, and what concepts needed reteaching. Schools integrated these tools into math and ELA blocks so students could work independently while teachers focused on small groups.
How It Worked
Districts used different AI platforms through the state’s AI-Powered Platform Pilot Grant. Schools reported using AI tutors to support early literacy, middle-school math, and just-in-time coaching for struggling students. Teachers described the tools as a practical way to keep students engaged and supported while managing large classrooms.
What Educators Said
According to state testimony and ECS reporting, more than half of teachers in the pilot rated their experience as positive. The Indiana Department of Education concluded that AI tools “reduce paperwork for teachers, extend instructional time, and empower students to take greater ownership of their learning.”
Case Study #2: EdoBEST After-School Tutoring Program (Edo State, Nigeria) – Using Hybrid AI Tutoring to Accelerate English Learning
Focus: Hybrid AI Tutoring + Teacher Facilitation
Heroes: EdoBEST classroom teachers, learning facilitators, instructional coaches, and state education project leaders
What They Did
Edo State ran a six-week after-school program where students worked with a generative-AI tutor for personalized English practice while teachers circulated to monitor progress, coach students, and intervene when needed. The AI tutor provided structured lessons, hints, and scaffolded practice.
How It Worked
The World Bank study reports that the AI tutor handled step-by-step guidance and immediate feedback, while teachers focused on motivation and targeted support. This hybrid model allowed every student to work at an appropriate level without waiting for teacher availability.
What the Results Showed
Students gained 0.31 standard deviations in just six weeks. That equals nearly two years of typical learning. Researchers concluded that EdoBEST provides “strong evidence that generative AI can function effectively as a virtual tutor” when paired with human support.
Case Study #3: Thomas et al. (2024) – U.S. Middle School Hybrid Tutoring
Focus: Three-Site Hybrid Human–AI Tutoring Model
Heroes: Principal investigators Danielle R. Thomas and research team; participating math teachers, intervention specialists, paraprofessional tutors, and site coordinators
What They Did
Researchers implemented hybrid human-AI tutoring across three low-income middle schools. Students used AI-supported math tutoring software, while human tutors monitored sessions, redirected students, and provided conceptual support.
How It Worked
The AI platform offered problems, hints, and feedback, and tutors used real-time information to adjust their approach. The study compared this hybrid model with schools using software-only tools.
What the Results Showed
Across all three sites, hybrid tutoring improved student proficiency and system usage more than software alone. Lower-achieving students benefited the most, suggesting hybrid tutoring can reduce equity gaps.
Case Study #4: Vanzo et al. (Italy) – Improving English Outcomes With GPT-4 Interactive Homework Tutoring
Focus: After-Hours AI Homework Support
Heroes: Lead researchers (Vanzo, Pal Chowdhury, Sachan); participating English teachers; 9th–10th grade Italian secondary students
What They Did
A high school in Italy replaced traditional English homework with interactive GPT-4 sessions. Students completed assignments through guided conversations with the AI tutor, receiving corrections, follow-up questions, and feedback.
How It Worked
Teachers assigned the same content but delivered through the AI interface. GPT-4 asked students to explain reasoning, revise mistakes, and practice grammar interactively. This mimicked the scaffolding a tutor would offer.
What the Results Showed
In this randomized controlled trial, students using GPT-4 made significantly greater grammar gains and reported higher engagement. Most said they wanted to continue using AI for homework because it kept them from “getting stuck” during assignments.
D – Pro Tips
1. Combine AI tutoring with human guidance
AI tutoring is most effective when teachers or tutors remain involved. Coaching students, reviewing their work, and offering conceptual explanations that AI cannot provide.
2. Start with a focused pilot before scaling districtwide
Launching with one grade level, one subject, or one targeted skill set helps educators learn the tools, refine workflows, and build confidence before expanding.
3. Train educators before introducing AI to students
According to ECS and Indiana DOE, teacher training is the biggest predictor of success. Educators need hands-on practice with dashboards, prompt design, data interpretation, and knowing when not to rely on AI.
4. Use AI to extend learning time, not replace instruction
EdoBEST and Vanzo et al. demonstrate that AI is most impactful when it increases the amount of productive practice students complete between teacher-led sessions. AI should add learning minutes, not replace teacher interaction.
5. Prioritize access for students who need the most support
Hybrid tutoring research shows the biggest gains for students who were previously struggling. Offering 24/7 homework support or structured after-school AI practice can close gaps for students lacking transportation, home support, or consistent access to human tutors.
6. Monitor usage and learning patterns regularly
Dashboards from Indiana’s pilots and the hybrid research studies show that consistent monitoring allows teachers to adjust small groups, reteach skills, and prevent students from practicing on autopilot.
7. Keep messaging clear: AI is an assistant, not a grader
All major studies emphasize that AI is for practice, scaffolding, and feedback, while teachers make instructional decisions. Communicating this clearly helps build trust with families and staff.
References
Education Commission of the States. “AI Pilot Programs in K–12 Education.”
https://www.ecs.org/ai-artificial-intelligence-pilots-k12-schools/
Indiana Department of Education. “Indiana’s Approach to Artificial Intelligence in Education.” Written Testimony of Dr. Katie Jenner to the U.S. Senate Committee on Health, Education, Labor, and Pensions (2025).
https://www.help.senate.gov/download/jenner-testimonypdf
World Bank. “From Chalkboards to Chatbots: Evaluating the Impact of a Large Language Model Virtual Tutor in Nigeria.”
https://documents.worldbank.org/en/publication/documents-reports/documentdetail/099548105192529324
World Bank. “From Chalkboards to Chatbots: Transforming Learning in Nigeria, One Prompt at a Time.”
https://blogs.worldbank.org/en/education/From-chalkboards-to-chatbots-Transforming-learning-in-Nigeria
Thomas, Danielle R., et al. Improving Student Learning with Hybrid Human–AI Tutoring: A Three-Study Quasi-Experimental Investigation.
https://arxiv.org/abs/2312.11274
Vanzo, Alessandro; Sankalan Pal Chowdhury; Mrinmaya Sachan. “GPT-4 as a Homework Tutor Can Improve Student Engagement and Learning Outcomes.”
https://arxiv.org/abs/2409.15981
