AI Homework Grading Comment SaaS · CNY 30K/mo · Unlock Annual Subscriptions from 5 Institutions

💰 Revenue

① Small and medium training institutions subscribe to grading services monthly/annually (main revenue): institutions pay subscription fees, monthly payment 1500 to 3000 yuan × initial 5 to 8 schools = monthly revenue 7,500 to 24,000 yuan (calculated based on the monthly fee and number of schools in the case card); the case card states total monthly revenue of about 30,000 yuan, accounting for about 100% of monthly revenue (calculation basis, from the case, not independently verified); ② Annual payment to lock in customers: institutions pay 30,000 to 60,000 yuan per school annually; the case card states that with improved renewal retention, revenue can reach 80,000 to 100,000 yuan per month in the second year (case basis, no third-party verification), and the share of revenue is not stated; ③ White-label/API licensing to teaching auxiliary and education equipment channels (licensing): licensed by school or by call volume; the licensing price and number of channels are not public, and how much revenue this line accounts for is not mentioned; ④ Opportunity item—teacher/parent individual end billed per sheet (usage-based billing): in the same track, Bee AI has cumulatively graded over 1.2 billion question attempts (case figure, no independent review); per-sheet pricing is not published, and its revenue share likewise cannot be verified.

💸 Cost

Use Coze's free version initially; after upgrading to the team version, 299 yuan per month; OCR recognition costs about 0.01 yuan per image, about 1,000 yuan for 10,000 images per month; large model API call costs are 1,000 to 2,000 yuan per month; server and domain together are about 300 yuan; in the early stage of visiting schools, budget about 1,000 yuan per month for materials and visit expenses. Total monthly cost is about 3,000 to 4,000 yuan; the larger the volume, the lower the average unit cost.

⏱ Time Investment

In the initial product build phase, 4 to 5 hours per day for two months; during operations, 1.5 hours per day to monitor Agent performance quality, handle recognition exceptions, and optimize comment templates; during the sales phase, 8 hours per week for telemarketing, field visits to school academic affairs directors, or attending education equipment expos.

🔑 Keys to Success

• ✅ Comment quality must make parents enjoy reading and teachers willing to use it, rather than mechanical right/wrong; emotional expression matters more than accuracy.
• ✅ Open B2B channels through principal networks or academic affairs director relationships; the education industry has a long decision chain and requires patience.
• ✅ Data security requires commitment to private deployment or encrypted storage, complying with the Personal Information Protection Law and education data regulations.
• ✅ Customize comment templates by subject and grade to form reusable assets; continuous optimization lowers later marginal costs.

⚠️ 风险

• ⚠️ School procurement processes last three to six months and payment collection is slow; public schools must go through government procurement, causing longer delays and high cash flow pressure.
• ⚠️ Large players such as Bee AI and Xiaoe-tech may enter at low prices, using free or low-price strategies to squeeze the survival space of individual developers and small teams.
• ⚠️ Handwriting recognition accuracy is unstable in lower grades or with messy handwriting, requiring manual fallback that increases hidden costs and limits expansion in the younger-age market.
• ⚠️ Education data involves minors' privacy and has high compliance requirements; a data leak would lead to customer loss and legal risks.