Gunjo · Business Intelligence for the AI Era
← Sticker Wall AGENT · DETAIL

Handwritten Homework Instant Grading SaaS · 3-Second Comments · Monthly Revenue of 28k CNY · Serving 200 Classes

Workflow: Every day, teachers upload photos of handwritten homework. AI recognizes the handwritten content, compares it with the q

AGENT

Key Fields

FIELD STAMPS
IndustryContent / Creator Economy
RegionChina
ScaleSME
ChannelOnline

🔧 Workflow

Every day, teachers upload photos of handwritten homework. AI recognizes the handwritten content, compares it with the question bank, and automatically grades it, outputting correct/incorrect markers and personalized comments within three seconds. Comments are automatically pushed to the teacher's end. After a one-click review by the teacher, they are forwarded to parents. Hundreds of homework assignments are processed in batches daily, accumulating wrong-answer data. Wrong-answer data is automatically aggregated into the learning analytics dashboard, providing value-added leverage for subsequent renewal negotiations.

🛠 Setup Requirements

Requires integration with mature OCR and homework grading APIs (such as Mifeng Xiaoxiao or self-developed models), and the establishment of a mini-program or web-based teacher workbench. Developers need front-end and API integration capabilities; non-technical backgrounds can purchase white-label systems, such as the intelligent grading module of Mifeng Xiaoxiao or EduSoho Lingxi. From deployment to the first school pilot takes about two to four weeks, and subsequent replication to the second school takes only one week. The core workload lies in the localization and refinement of comment templates.

🧰 Toolchain

  • 🔧 Mifeng Xiaoxiao AI Grading System
  • 🔧 Tencent Cloud OCR API
  • 🔧 WeChat Mini Program Management Backend
  • 🔧 EduSoho Lingxi (Reference solution replacing human teaching assistants)
  • 🔧 Coze (Used for comment style tuning and automated workflow orchestration)

💰 Revenue

① Monthly subscriptions from training institutions by class (Main revenue): Institutions pay per class per month. 100 CNY per class per month × 200 classes signed in the first semester = 20,000 CNY/month. In the second semester, with an 80% renewal rate and 300 classes total, monthly revenue reaches 28,000 CNY, accounting for about 82% of monthly revenue (proportion derived from internal card data, self-reported by merchants, not independently verified); ② Upselling of wrong-answer analysis reports: Institutions add-on purchases per class based on their subscription, adding 30 CNY per class per month, which can further increase monthly revenue by about 0,600 CNY to 34,000 CNY (specific proportion in total revenue not provided, estimated based on card data and merchant statements, lacking third-party verification); ③ Channel referral commissions: Introducing local training institutions yields a commission of one month's subscription fee for each school introduced. The scale of these commissions cannot be verified, and their weight in total revenue has not been provided; ④ Opportunity item - Academic-year batch subscriptions for public schools or county/district education bureaus: Annual package licensing per class; neither pricing nor the number of signed classes has been publicly disclosed, and the proportion of revenue is similarly unlisted.

💸 Cost

OCR and grading APIs are billed by call volume, approx. 3,000 CNY/month; Server and mini-program authentication annual fees approx. 2,000 CNY; If using a white-label system, a one-time licensing fee of 5,000 CNY is required; Total monthly cost is kept within 5,000 CNY, with a profit margin exceeding 70%.

⏱ Time Investment

Dedicate about three hours daily: processing teacher-uploaded grading tasks in the morning, handling new sign-ups and renewal communications in the afternoon, and concentrating on updating the question bank and comment templates over the weekend. Double the effort is required for renewal sprints in the two weeks before winter and summer vacations; otherwise, it runs remotely and automatically on regular days.

🚀 Getting Started

Step 1: Register an account on Mifeng Xiaoxiao or similar white-label platforms and get a free trial of their AI grading API; Step 2: Contact two local training institutions for a one-month free trial, collecting real homework samples to refine comment templates; Step 3: Use trial data as a case study to quote local schools by class, prioritizing K12 academic subject institutions over public schools to shorten the decision-making cycle.

🔑 Keys to Success

  • ✅ Handwriting recognition accuracy must exceed 95%, and wrong-answer attribution must be precise
  • ✅ Comments must be personalized, containing encouragement and suggestions for improvement, making parents willing to forward and spread them
  • ✅ Renewals rely on the accumulation of wrong-answer data; schools face high switching costs and are bound to semester archives
  • ✅ Establish a referral mechanism with local training institutions, returning one month's subscription fee as commission for each school introduced

⚠️ 风险

  • ⚠️ School budgets are affected by policy fluctuations and procurement may be suspended mid-semester; customers need to be diversified across multiple institutions
  • ⚠️ Handwriting recognition has insufficient fault tolerance for messy handwriting, tilted photos, and other scenarios; models must be continuously iterated with real samples
  • ⚠️ Low-priced imitations flooding the market require building a high moat through personalized comments and localized services
  • ⚠️ Data privacy compliance risks; student homework and grade data must be encrypted and stored in compliance with personal information protection laws

📌 Real Cases

  • 📌 Mifeng AI launched the handwritten homework grading feature in April 2026, covering all K12 subjects with second-level feedback per question. The company has signed contracts with multiple training institutions, with grading volumes reaching tens of thousands of copies daily
  • 📌 Xiao E Tong upgraded its AI intelligent grading in June 2026, improving the delivery efficiency of private-domain teaching and making teaching feedback more timely and professional, with hundreds of training institutions integrated and subscribed
  • 📌 Xiueda Education built an after-class feedback assistant through Coze AI, solving the pain point of after-class feedback, automating the teacher grading and parent communication links, and reducing labor costs for a single campus by about 30%