The AI Tutoring Question: Whether a $7/Month App Replaces What a $200/Month Kumon Centre Does
A prospective franchisee’s guide to the technology risk nobody in the Kumon system is quantifying — and what it means for your five-year investment horizon.
You are considering buying a business that charges parents $150 to $200 per month, per child, per subject, to do paper worksheets at a folding table. Down the road, a parent can open Khan Academy’s Khanmigo AI tutor for $4 to $9 per month and get a one-on-one coaching experience that adapts in real time. ChatGPT costs $20 per month and covers every subject a child will ever encounter. Photomath is free.
The price gap is 20 to 50 times. The product is not identical. But the question every prospective Kumon buyer needs to sit with is not whether AI tutoring is as good as Kumon. It is whether enough parents will perceive it as close enough to stop writing those monthly checks.
Nobody inside the Kumon system is disclosing enrollment trend data — Kumon publishes no Item 19 financial performance representation — so you cannot answer this question with their numbers. You have to reason through it yourself.
The Price Comparison That Should Worry Every Kumon Buyer
Lay the numbers side by side.
- Kumon: $150 to $200 per month per student, per subject. A child doing both math and reading is $300 to $400 per month. That is $3,600 to $4,800 per year before driving to the centre.
- Khanmigo (Khan Academy): $4 to $9 per month for AI-guided tutoring across subjects, built on a pedagogy-first design philosophy.
- ChatGPT: $20 per month. Not designed for K-12 pedagogy, but functionally unlimited in scope. Millions of students already use it for homework help.
- Photomath: Free. Solves and explains math problems step by step from a phone camera.
- Google Gemini “Guided Learning”: Structured study plans and quizzes, bundled into an ecosystem most families already use.
The per-student economics of a Kumon centre depend on maintaining enrollment volume at that price point. If 20 percent of your families decide a $9 app gets their child 80 percent of the way there, your revenue does not drop 20 percent — it drops 20 percent before the word-of-mouth effect hits the next enrollment cycle.
This is not a theoretical exercise. These products exist today, they are improving on a quarterly release cadence, and parents are trying them.
What AI Tutoring Actually Does in 2026
The common objection is that AI tutoring is just a homework-answer machine. That was true in 2023. It is less true now.
Khanmigo is the closest analogue to Kumon’s worksheet method. It does not hand students answers. It asks guiding questions, offers hints, and coaches a student toward a solution — the Socratic method at scale. It tracks mastery across topics and adjusts difficulty. Khan Academy built it on two decades of learning science data from 150 million registered users.
Google Gemini’s education features now include structured study plans, adaptive quizzes, and progress tracking. Google has the distribution advantage: it is already on every Chromebook in every school district.
ChatGPT is a general-purpose tool, not a tutoring product. But that distinction is lost on the 14-year-old who opens it at 9 PM to work through algebra. OpenAI’s usage data suggests students are one of its largest user segments.
The research says blended wins. Work from the Clayton Christensen Institute on blended learning models consistently finds that human-plus-AI approaches outperform either AI-only or traditional-only instruction. This is significant for Kumon, because it suggests the optimal model may not be worksheets alone or apps alone — it may be something Kumon does not currently offer.
The market is moving fast. The global AI-in-education market was estimated at $5.88 billion in 2024 and is projected to reach $32.27 billion by 2030, according to market research from Grand View Research. Those are estimates and should be treated as directional, not precise. But the direction is not ambiguous.
What Kumon Does That AI Cannot (Yet)
If AI tutoring were a clean substitute, Kumon’s roughly 1,710 US centres would already be empty. They are not. There are real things happening inside a Kumon centre that no app replicates today.
Physical presence and routine. A child physically goes somewhere, at a scheduled time, and sits down to work. That structure matters. For many families, the act of going to Kumon is the accountability mechanism. A subscription app lives on an iPad that also has YouTube and Roblox.
Human observation in real time. A Kumon instructor watches a child work. They read frustration before the child voices it. They see the pencil stop. They hear the sigh. AI cannot see a child’s posture or read the moment a seven-year-old is about to give up. Kumon’s methodology is built around this observation — the instructor decides when to advance a student, when to hold, and when to circle back.
The accountability of cost. This is an underrated moat. Writing a $200 per month check concentrates a family’s attention on the child’s progress in a way that a $7 subscription does not. When something costs real money, parents show up to parent-teacher conferences, enforce the homework schedule, and hold the child to the program. Cheap tools get abandoned in week three.
Sixty years of level-progression data. Kumon’s proprietary worksheet sequence has been refined over more than sixty years of use. The progression from level to level is not a prompt — it is a curriculum refined over decades. Structured repetition builds procedural fluency, the kind of deep automaticity where a child does not think about 7 times 8, they just know. AI-guided exploration may build understanding without building that automaticity.
Community and peer effects. Children working alongside other children in a quiet room normalises effort. A child on a tablet at home is alone with the work.
What AI Does That Kumon Cannot
The advantages run the other direction too, and a prospective buyer who dismisses them is making a bet, not an analysis.
Instant adaptive difficulty. AI adjusts in real time, problem by problem. Kumon’s leveling system requires a human to assess and re-assign worksheets, typically on a weekly or biweekly cadence. A child who masters a concept on Tuesday still does the same worksheet on Thursday.
Available every hour of every day. A Kumon centre is open during a narrow afternoon window, typically two to three hours, two days per week. A child with a question at 8 PM on a Sunday has no Kumon resource. AI is there.
Unlimited subject scope. Kumon offers math and reading. If a parent wants help with science, history, writing, or coding, that is a separate product, a separate vendor, a separate bill. AI covers everything for one price — or for free.
No commute, no parking, no scheduling friction. Parents cite logistics as a primary reason for dropping Kumon. Two round trips per week, in traffic, with siblings in the car, for 15 to 20 minutes of supervised worksheet time. AI eliminates that entire cost.
Cost approaching zero. The fundamental economics of software are that marginal cost drops toward zero at scale. It is already there for basic functionality. Kumon’s fundamental economics are that every centre needs a lease, an instructor, and worksheet inventory. Those costs do not drop.
The Investment Question: What This Means for a 5-Year Agreement
You are not buying a Kumon franchise for 15 years. You are buying a five-year agreement, and the question is whether enrollment holds for five years, not whether it holds forever.
Frame it as two scenarios.
Downside scenario: slow erosion. AI tutoring adoption grows steadily among the affluent, education-focused families who are Kumon’s core demographic. These parents do not switch overnight — they experiment. They try Khanmigo alongside Kumon. Their child improves on both. They start asking whether they need both. Enrollment erodes 5 to 15 percent per year, not from a single exodus but from a slow leak of families who quietly do not re-enroll. On an already modest per-centre income floor, that compression is the difference between a viable small business and a money-losing lease obligation.
Upside scenario: Kumon adapts. Kumon integrates AI tools into its programme and becomes a blended model — human instruction augmented by adaptive technology. This is not unprecedented. Mathnasium has added digital components. Sylvan Learning pivoted years ago. If Kumon makes this move, it could strengthen the value proposition: the structure and accountability of a physical centre, plus the adaptive capability of AI. The brand’s scale — roughly 1,710 US centres within a large international network — gives it negotiating leverage with any edtech partner.
What is not priced in. Neither scenario is reflected in the current franchise economics, because nobody is disclosing enrollment trends. Kumon’s FDD does not include an Item 19. You cannot see whether the average centre’s student count is growing, flat, or shrinking. You cannot see whether newer centres in tech-forward markets are enrolling differently than older centres in traditional ones. This is the single most important variable for your investment, and it is a black box.
The competitive landscape is already shifting. Other supplemental education franchises are moving. The market is not standing still, and a buyer who evaluates Kumon’s position based on 2024 competitive dynamics is underwriting a business that may face a materially different market by 2028.
The honest answer. We do not know how fast this disruption moves. Anyone who tells you AI will kill Kumon in three years is speculating. Anyone who tells you Kumon’s model is immune to technology disruption is also speculating. What we can say is this: the price differential is real, the technology is improving on a steep curve, the adoption is measurable, and the franchise agreement locks you in for five years with no enrollment guarantee.
Your diligence on this question should not rely on Kumon’s corporate messaging or on AI evangelists. Talk to current owners. Ask them directly: are your student counts growing, flat, or declining? Are parents mentioning AI alternatives during enrollment conversations? Is corporate communicating a technology strategy?
If owners will not answer those questions, or if the answers are vague, that is itself information. The absence of data, when data should exist, is never a neutral signal.