A good friend of mine, who used to work as a trainer at a major education company, lost his job during the 2024 downturn. Instead of looking for another job, he decided to go solo with a niche that most people overlook: vocational certification exam tutoring. He partnered with me to inject AI tools into the entire workflow, and within four months, we built a project doing 80,000 RMB in monthly revenue with just two full-time staff and a few part-time tutors. Here is the complete playbook.
Why certification exam tutoring is a goldmine right now
The Chinese job market is brutal. Layoffs are everywhere, and the only thing that gives workers a sense of security is a certificate. From construction safety officers and fire engineers to nutritionists, parenting coaches, and psychological counselors, the demand for exam prep is exploding. These are not optional exams - many industries require specific certifications by law, and companies are willing to subsidize their employees exam fees.
The best part is that almost all of these exams have standardized syllabuses and past exam papers available online. The content is repetitive, predictable, and perfectly suited for AI to handle the heavy lifting.
Step 1: Use AI to build your course library in one week instead of three months
Instead of hiring subject matter experts to write course materials from scratch, we built a systematic process. First, we collected the exam syllabus from the official government website. Then we fed it into GPT-4 and Kimi along with 3 to 5 years of past exam papers, and asked the AI to generate a structured learning outline mapped to each exam point.
For each chapter, we used AI to generate lecture scripts, practice questions, and exam prediction papers. The trick is to iterate: generate, review for accuracy, feed corrections back, and let the AI regenerate the improved version. One person can build the complete course library for an exam in 3 to 4 days. We now have course libraries for 12 different certifications, which would have cost us over 200,000 RMB and six months of work using traditional methods.
Step 2: Use AI voice cloning and digital avatars for video production
We tried recording ourselves explaining concepts, but the production quality was inconsistent and took too long. Instead, we used AI voice cloning tools to clone my friends voice, and then generated lecture videos using a combination of slides and the cloned voiceover. The result sounds exactly like him, but we can produce a 30-minute lecture in about 20 minutes.
For short video marketing on Douyin and Kuaishou, we used digital avatar tools. We recorded a 5-minute real video of the tutor, trained a digital avatar model, and then used text-to-video generation. Now we can pump out 10 to 15 short videos per day, each explaining a single exam point or sharing a test-taking tip. These videos drive organic traffic to our WeChat consultation funnel.
Step 3: Build an AI consultation and sales assistant
The biggest bottleneck in this business is consulting. Potential students ask the same questions over and over: what are the requirements, how long does it take, what is the pass rate, how much does it cost. We trained a custom GPT on our FAQ database and integrated it into our WeChat workflow using a simple chatbot interface.
When a new lead adds us on WeChat, the AI assistant handles the initial conversation, answers basic questions, shares sample videos, and qualifies the lead. Only when the lead shows serious purchase intent does the human tutor take over to close the sale. This freed up 70 percent of the tutors time and allowed us to handle 10 times more leads without hiring additional sales staff.
Step 4: AI-powered study group management
Once a student enrolls, they join a WeChat study group. We used AI to automate daily check-in reminders, send daily practice questions, generate weekly progress reports, and answer common questions at 2 AM when no tutor is online. The AI monitors each students progress and flags at-risk students who havent studied in 3 days or who score below 60 percent on practice exams.
The human tutor then reaches out personally to those students with a targeted pep talk and study plan. This approach increased our completion rate from 45 percent to 78 percent, which directly translates to higher pass rates and better word-of-mouth referrals.
The economics
Each course sells for 1,980 to 3,980 RMB depending on the certification. Our cost structure is minimal: 500 RMB per month for AI tool subscriptions, 1,000 RMB for cloud servers and domain hosting, and 25,000 RMB for two full-time staff salaries. The rest is pure profit.
We currently maintain 3 enrollment cohorts per month, with an average of 40 to 50 students per cohort. At a conservative estimate of 2,000 RMB per student, that is 80,000 to 100,000 RMB in monthly revenue with less than 30,000 RMB in fixed costs.
The biggest lesson
Do not try to create original content from scratch. AI is bad at being creative, but it is exceptional at processing and organizing existing knowledge. For exam prep, the knowledge already exists in government syllabuses, textbooks, and past papers. Your job is to use AI to repackage that knowledge into a structured learning experience, produce it at scale, and deliver it with personalized attention.
The second lesson is that AI cannot replace the human touch in closing sales and supporting struggling students. Use AI for the 80 percent of repetitive work, and let your human team focus on the 20 percent that builds trust and relationships.
If you are looking for an AI business that requires zero technical skills, has clear demand, and can be started with less than 10,000 RMB, certification exam tutoring is one of the best options right now. Pick one certification that you have some knowledge of, build your course library with AI in one week, start creating short videos, and get your first student within 30 days. The market is massive, and AI gives you the leverage to compete with established players on a fraction of the budget. |
