evidoria

← Back to browse

Good practice

atama+ — AI-Adaptive Root-Cause Diagnosis for Japanese Cram Schools

Japan · Tokyo · See the Japan profile

atama+'s AI traces each student's errors back to the exact missing prerequisite, sometimes years below grade level, then sequences personalised video and practice. Used in 3,000+ Japanese cram-school classrooms and one public-school pilot; outcome claims remain mostly company-rep

3,000+ classrooms
Cram-school classrooms deployed (by March 2022)
300 million+
Cumulative answers logged (by March 2022)
~46 $M (¥5bn)
Series B funding raised (2021)
103 students
METI pilot students at Oiso High School (Sept 2021–Feb 2022)
~1.5x
Self-reported mock-exam score increase (single juku) (2-week intensive)
65 to 97 points
Self-reported classroom maths score jump
22,000–24,200 JPY
Monthly fee per subject
atama+ — AI-Adaptive Root-Cause Diagnosis for Japanese Cram Schools atama+ — AI-Adaptive Root-Cause Diagnosis for Japanese Cram Schools

Details

Maturity
Established
Promoter
atama plus inc. (now a Kumon subsidiary)
Period
2017-present
Keywords
adaptive learning, diagnostic assessment, cram schools (juku), edtech

Context

atama plus inc., a Tokyo AI company founded in 2017, builds a tablet-based platform used inside Japanese cram schools (juku) and preparatory schools. It diagnoses precisely which prerequisite a struggling student is missing — sometimes several grade-levels below the topic at hand — by analysing answer patterns, response times and concentration signals, then sequences short videos and practice problems to close that specific gap.

Results

By March 2022, atama+ was deployed in more than 3,000 juku classrooms nationwide with over 300 million cumulative answers logged, used by major prep-school brands including Sundai, Z-kai, Gakken and Jonan Yobiko. Japan's METI tested atama+ under its 'Future of the Classroom' programme (Sept 2021–Feb 2022) at private Iwakura High School and public Kanagawa Prefectural Oiso High School (103 first-year students, roughly two months of AI-assisted maths study); Oiso now runs atama+ mock exams six times a year. Circulating outcome figures — a single juku's report of mock-exam scores rising roughly 1.5x over a two-week intensive, or classroom-level jumps such as 65 to 97 on a maths test — are self-reported by individual cram schools with no control group, and the METI pilot's design has not been matched by a published results report.

Conclusions

Independent, peer-reviewed evaluation is essentially absent. Reviews also flag a real access concern: monthly fees of roughly ¥22,000–24,200 per subject are pricier than sub-¥10,000 online alternatives, and some students report reduced autonomy since the AI, not the learner, sets the topic sequence. In 2026 atama plus was fully acquired by Kumon Educational — Kumon's first acquisition since 2000 — signalling durable institutional backing despite the thin independent evidence base.

Implementation

Indicative cost
Medium (€50k–€500k) — Monthly fees of roughly ¥22,000–24,200 per subject are explicitly described as pricier than sub-¥10,000 alternatives; cost_band set to medium.
Time to results
Short (< 1 year) — Tablet platform is deployed within existing juku classroom settings without describing a lengthy institutional rollout; timeline_band set to short.
Staffing & skills
Works within existing juku instructor-student settings; AI diagnostic runs largely autonomously

Conditions for success

  • Partnership with major prep-school brands (Sundai, Z-kai, Gakken, Jonan Yobiko)
  • Institutional backing from a public-sector pilot (METI, Kanagawa Prefecture) and, since 2026, full ownership by Kumon Educational

Common failure modes

  • No independent RCT or peer-reviewed evaluation found; circulating score gains are self-reported by individual schools with no control group
  • Monthly per-subject fee (~¥22,000–24,200) pricier than sub-¥10,000 alternatives
  • Some students report reduced autonomy since the AI sets the topic sequence

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful