Every school district technology director has faced some version of this conversation: a superintendent walks in energized after a conference, mentions that Khan Academy's AI tutor is getting national press coverage, and asks, "Why don't we just use that?"
It's a fair question—and one that deserves a serious, structured answer rather than a reflexive yes or no.
The choice between adopting an existing AI tutoring platform like Khanmigo and building (or licensing) a custom AI tutor is one of the most consequential EdTech decisions a district can make right now. It affects student outcomes, teacher workflows, budget cycles, and data governance for years to come.
This guide breaks down both paths with the specificity district leaders actually need to make a sound decision.
What Is Khanmigo, and What Does It Actually Do?
Khanmigo is Khan Academy's AI-powered tutoring assistant, built on OpenAI's GPT-4 technology and integrated directly into the Khan Academy platform. Launched broadly in 2023, it offers students Socratic-style guidance across math, science, humanities, and test prep subjects. It also includes teacher-facing features like lesson plan assistance and student progress summarization.
What Khanmigo does well:
- Guides students through problems using questioning rather than giving direct answers
- Integrates with Khan Academy's existing exercise library and video content
- Provides a safe, monitored chat environment with content guardrails
- Offers teacher tools for classroom integration
- Requires minimal setup—districts can be live within days
For individual families and teachers, Khanmigo represents a genuinely impressive product. For districts evaluating it at scale, however, the picture becomes more complicated.
The Case for Khanmigo in a School District Context
Speed of Deployment
Khanmigo's most significant advantage for districts is time-to-value. Because it's a mature, cloud-hosted product, there's no infrastructure to build, no AI model to train, and no content library to populate. A district with 10,000 students could theoretically have Khanmigo available to every student within two to four weeks of signing an agreement.
For district leaders facing immediate pressure—a teacher shortage, a post-pandemic learning gap, or a school board mandate—that speed matters enormously.
Lower Upfront Cost
Khanmigo pricing for schools is structured as a per-student subscription. This makes it budget-predictable and avoids the large capital expenditures that custom development demands. Districts with tight annual budgets and limited technology reserves often find this model more politically viable with school boards and finance committees.
Proven Pedagogical Foundation
Khan Academy has decades of learning science research informing its content and approach. Khanmigo inherits that foundation. For districts without in-house curriculum expertise or instructional design capacity, this is meaningful—the guardrails and pedagogical choices are already baked in.
The Limitations of Khanmigo for School Districts
Despite its strengths, Khanmigo carries constraints that become significant at the district level.
Curriculum Alignment Gaps
Khanmigo is built around Khan Academy's own content library, which follows a generalized curriculum framework. It does not natively align to a district's specific scope and sequence, adopted textbooks, or state-specific standards variations. A fifth-grade teacher in Texas using a specific math curriculum will find that Khanmigo's tutoring references and explanations may not match what students encountered in class.
This misalignment isn't a flaw in Khanmigo—it's an inherent limitation of any universal platform trying to serve every district simultaneously.
Limited Customization and Branding
Khanmigo looks and feels like Khan Academy. Districts cannot white-label the experience, embed it into their existing student portals, or meaningfully customize the interaction style to reflect local instructional priorities. For districts building a coherent digital learning environment, this creates friction.
Data Ownership and Privacy Considerations
Student data generated through Khanmigo interactions resides within Khan Academy's infrastructure. While Khan Academy complies with FERPA and COPPA, districts with stringent data governance policies—particularly those in states with additional student privacy laws—may face complications. Critically, districts have limited visibility into how interaction data is stored, retained, or used for model improvement.
Dependency on a Single Vendor's Roadmap
When a district adopts Khanmigo, they're betting on Khan Academy's product decisions for years. If Khan Academy changes pricing, deprecates features, or shifts strategic priorities, the district has limited recourse. This vendor dependency is a real governance risk, particularly for multi-year technology plans.
What Does "Building a Custom AI Tutor" Actually Mean?
The phrase "build a custom AI tutor" encompasses a wide range of approaches, and districts should understand the spectrum before assuming it means hiring a team of AI engineers.
Option 1: Full custom development A district (or more realistically, a consortium of districts) contracts with an AI development firm to build a tutoring system from scratch—custom models, custom content, custom UI. This is the most expensive and time-intensive path, typically requiring 18–36 months and budgets in the millions.
Option 2: Custom configuration on a white-label platform Districts work with an EdTech provider to deploy an AI tutoring solution that is built on proven AI infrastructure but customized to district specifications—curriculum alignment, branding, subject coverage, and integration with existing SIS and LMS systems. This middle path typically takes 3–6 months and costs significantly less than full custom development.
Option 3: AI tools with deep configurability Some EdTech platforms, including tools developed by companies like Evelyn Learning, offer AI tutoring and homework help products that can be configured for specific curricula, grade levels, and institutional branding without requiring ground-up development. This approach can deliver the customization districts need while maintaining deployment timelines closer to off-the-shelf solutions.
For most school districts, Option 2 or Option 3 represents the realistic custom path—not a team of internal engineers building neural networks.
Build vs. Buy: A Framework for District Decision-Makers
Rather than treating this as a binary choice, district leaders should evaluate five dimensions.
1. Curriculum Specificity Requirements
Ask: How tightly does the AI tutor need to reflect our specific curriculum, pacing guides, and instructional language?
If your district uses a specific adopted program (e.g., Eureka Math, Into Reading, OpenSciEd) and teachers want AI support that references the same vocabulary, examples, and sequencing students encounter in class, a generic platform will consistently fall short. The more curriculum-specific your requirements, the stronger the case for a custom or configurable solution.
2. Data Governance Standards
Ask: What are our district's data residency, ownership, and privacy requirements?
Districts in states with robust student privacy legislation—California's SOPIPA, New York's Education Law 2-d, or similar frameworks—need to scrutinize where data lives and who controls it. Custom or white-label solutions can be structured to give districts full data ownership and control, something generic platforms cannot always guarantee.
3. Integration Depth
Ask: Does the AI tutor need to connect to our SIS, LMS, rostering systems, or existing digital tools?
Khanmigo operates largely as a standalone experience. If a district wants AI tutoring embedded in Schoology, Canvas, or Google Classroom—with student data flowing between systems—custom integration becomes a requirement, not a preference. The administrative overhead of managing a siloed platform is often underestimated.
4. Timeline and Budget Realities
Ask: When do we need this operational, and what does our budget cycle actually support?
If a district has an urgent need and is working with per-pupil subscription budgets, Khanmigo or a similar off-the-shelf tool is the pragmatic choice. If there is runway to plan, and if the district is willing to view the investment as a multi-year infrastructure decision, a custom or configurable solution will typically deliver better long-term ROI.
A useful benchmark: districts that invest in custom AI tutoring infrastructure typically recover the cost premium within two to three years through reduced dependency on third-party content licensing, more efficient teacher support workflows, and measurable improvements in student outcomes that reduce remediation costs.
5. Institutional Capacity for Ongoing Management
Ask: Do we have the internal capacity to manage, evaluate, and iterate on a custom tool?
Custom solutions require a district champion—someone who can interface with the technology provider, evaluate performance data, communicate teacher feedback, and drive continuous improvement. Without that internal capacity, even the best custom solution will underperform because it won't be actively maintained and refined.
What the Research Says About AI Tutoring Effectiveness
Before choosing any platform, it's worth grounding the conversation in what actually drives AI tutoring effectiveness.
Studies on intelligent tutoring systems consistently show that the following factors predict student learning gains:
- Curriculum alignment: Tutors that match the specific content and language students encounter in instruction produce stronger outcomes than those operating on parallel content
- Socratic vs. direct-answer approaches: Systems that guide students to answers through questioning—rather than providing answers directly—produce deeper understanding and better retention
- Feedback specificity: Vague encouragement produces minimal learning gains; specific, actionable feedback at the sentence or step level produces measurable improvement
- Teacher integration: AI tutoring that connects to teacher visibility and classroom instruction outperforms tools that operate in isolation from the classroom
This research framework is useful for evaluating any platform, including Khanmigo. The question isn't just "does it use AI?" but "does its AI implementation reflect what learning science tells us about effective tutoring?"
Evelyn Learning's 24/7 AI Homework Helper, for example, is built specifically around the Socratic questioning model, guiding students to discover answers through structured prompting rather than surfacing direct solutions. That design choice reflects a deliberate application of learning science—not just a technical architecture decision.
Practical Scenarios: Which Path Makes Sense?
Scenario A: Mid-Sized District, Urgent Need, Limited Tech Capacity
A 15,000-student district in a midwestern state is facing significant teacher shortages in math and science. The superintendent wants a homework help solution operational before the next school year.
Recommendation: A configurable white-label platform is the strongest option here. Full custom development is too slow. A generic platform like Khanmigo can serve as a stopgap but won't align with the district's specific math curriculum, creating teacher friction. A white-label AI homework help solution that can be configured to district standards and deployed in 60–90 days threads the needle between speed and customization.
Scenario B: Large Urban District, Strong Tech Team, Multi-Year Plan
A 90,000-student urban district with a mature EdTech infrastructure team is building a five-year AI learning strategy. They have adopted curriculum materials, robust data governance policies, and a clear RFP process.
Recommendation: Custom or deeply configurable solution with full data ownership. The scale justifies the investment, the internal capacity exists to manage it, and the curriculum specificity requirements exceed what any generic platform can deliver. This district should issue an RFP that evaluates both custom development and white-label platform providers.
Scenario C: Charter School Network, Limited Budget, Need for Consistency
A 12-school charter network wants consistent AI tutoring support across all campuses, using a shared curriculum framework.
Recommendation: White-label platform with network-wide configuration. The shared curriculum makes customization achievable without full custom development. A platform that allows centralized content configuration with school-level reporting gives the network the consistency they need. The Evelyn Learning AI Tutoring Co-Pilot model, which provides real-time teaching support and can be configured for specific instructional approaches, is a strong fit for networks that want to standardize quality across campuses.
Key Questions to Ask Any AI Tutoring Vendor
Whether evaluating Khanmigo, a custom development partner, or a white-label platform provider, district leaders should ask:
- Who owns the student interaction data, and where does it reside?
- Can the system be configured to align with our specific adopted curriculum materials?
- How does the AI handle incorrect responses—does it correct or guide?
- What does teacher-facing reporting look like, and how does it integrate with our existing tools?
- What is the escalation path when the AI encounters a question it cannot handle appropriately?
- How is the AI model updated, and how are we notified of changes that might affect student interactions?
- What does implementation support look like, and what internal capacity do we need to maintain on our end?
Frequently Asked Questions
Is Khanmigo free for school districts?
Khanmigo is available through Khan Academy's school and district pricing, which typically involves a per-student annual subscription fee. Khan Academy has offered subsidized pricing for Title I schools, but it is not universally free for district-wide deployment. Districts should contact Khan Academy directly for current pricing structures.
How long does it take to build a custom AI tutor for a school district?
Full custom AI tutor development typically takes 18–36 months. However, configuring a white-label AI tutoring platform to district specifications can be completed in 60–180 days depending on complexity, curriculum alignment requirements, and integration needs.
Can AI tutoring tools replace teachers?
No. AI tutoring tools are most effective when they extend teacher capacity—handling after-hours support, providing immediate feedback on practice work, and surfacing student misconceptions for teacher follow-up. Research consistently shows that human teacher relationships are irreplaceable for motivation, social-emotional learning, and complex instructional decisions.
What is the difference between Khanmigo and a white-label AI tutoring platform?
Khanmigo is a branded, fixed-functionality product built around Khan Academy's content library. A white-label AI tutoring platform provides the same core AI tutoring capabilities but allows institutions to apply their own branding, configure curriculum alignment, and integrate with existing systems—giving districts significantly more control over the student experience and data governance.
How do school districts ensure FERPA compliance with AI tutoring tools?
Districts should require any AI tutoring vendor to sign a FERPA-compliant data processing agreement (DPA) that specifies data ownership, retention policies, deletion procedures, and prohibitions on using student data for advertising or model training without explicit consent. The Student Data Privacy Consortium (SDPC) maintains a database of vendor agreements that districts can reference.
The Bottom Line
Khanmigo is a genuinely capable product that solves real problems for teachers and students. For districts that need something operational quickly and can accept curriculum alignment limitations, it's a defensible choice.
But "defensible" is different from "optimal."
For districts serious about building AI tutoring capacity that reflects their curriculum, respects their data governance standards, and integrates with their existing technology environment, the custom or configurable path delivers meaningfully better long-term outcomes—even if it requires more planning and investment upfront.
The build-vs-buy decision in AI education isn't really a technology question. It's a strategy question. What level of dependency on an external vendor's roadmap is acceptable? How important is curriculum specificity to your teachers? What does your data governance framework require?
Answer those questions first, and the technology choice becomes much clearer.
If your district is working through this decision and wants to explore what a curriculum-aligned, white-label AI tutoring solution looks like in practice, Evelyn Learning's team of educators and engineers has helped hundreds of institutions navigate exactly this choice—with tools built from the ground up to serve institutional needs, not individual consumers.
