AI training is profitable. There are individuals making 30 million yen a month. However, those who try to profit from training alone will eventually disappear.
Why is that? In what order should an individual enter this market? I will write down the entire procedure, including market rates and institutional figures.
I wrote about the breakdown of why AI training is so profitable in my previous article. If you haven't read it, please start here.
https://x.com/genmai_tokyo/status/2072638502953459727
Premise: The Current State of This Market
First, three key figures:
- Prices for corporate AI training range from tens of thousands of yen to over 5 million yen for hands-on support.
- By using government subsidies (Human Resource Development Support Subsidy / Business Development Reskilling Support Course), SMEs can recover up to 75% of training expenses.
- This subsidy ends on March 31, 2027.
This is almost everything regarding why AI training is profitable and why you must hurry.
So, how can you create sustainable revenue as a business rather than just short-term profit?
STEP 1: Use AI Yourself and Communicate Specific Efficiency Gains and Results
The first thing to create is not a curriculum, but your own track record.
Whether it's Claude Code or ChatGPT, it doesn't matter. First, integrate it into your own work and get to a state where you can state in numbers "what task was changed, how, and how many hours were reduced." For example, "Creating quotes went from 3 hours to 10 minutes" or "Automating inquiry drafts saved 5 hours a week." That's enough.
What corporate clients buy is not a lecture, but reproducibility. There are countless "people knowledgeable about AI" now, but the number of people who have "numbers showing results in their own business" suddenly becomes small. These numbers are more effective than the content of the training itself.
STEP 2: Acquiring Initial Projects
There is no single correct way to get your first project. Use every available route.
Referrals from Free Training: Provide free training to a nearby CEO who is unfamiliar with AI. The compensation is not money, but a track record and referrals. Management networks for SMEs are often interconnected, so if you are truly helpful to one company, it leads to a chain reaction like "Mr. XX in our association was also struggling."
Content Creation: Publish your results on X or YouTube. For those who are good at it, this builds trust with zero advertising costs.
Existing Network: Former business partners, Chambers of Commerce, local management gatherings. This is particularly effective in regional areas. There are still plenty of regions where there are physically no people nearby who can talk about AI.
The routes may differ, but the principle is the same: don't "sell," just show evidence that makes them want to rely on you for AI. The numbers you created in STEP 1 are for this purpose.
STEP 3: Narrow Down the Menu Later While Exploring What Resonates
You cannot create a finished product like "AI Training Specialized for Manufacturing Quote Tasks" right away. Even if you could, whether it would sell is a different story.
Start with more general content for the first few cases. Then, record the reactions of the participants. In which part did they lean forward? For which task did questions increase? What kind of person came to ask questions individually after it ended, and what did they ask?
A specialized menu is not something you think up; it's something you extract from these records.
And narrowing down has a meaning in terms of pricing. As mentioned at the beginning, the market rate for this market ranges from tens of thousands to millions of yen. In a market where the rate is so ambiguous, only those who narrow it down to "this person for that field" can dictate their own price. General training is compared; specialized training is sought out by name.
STEP 4: Subsidies are a Tool to Make Sales Easier
This is the factor causing the current AI training bubble.
By using the Business Development Reskilling Support Course of the Human Resource Development Support Subsidy, SMEs are subsidized for 75% of training expenses. For a 200,000 yen training per person, 150,000 yen is returned, making the actual burden 50,000 yen. Furthermore, a wage subsidy of 1,000 yen per person per hour is provided during training hours.
You can tell the client's decision-maker, "The actual burden is one-fourth." The speed at which a 500,000 yen proposal passes compared to a 2 million yen proposal is on a different level. It's not the client's wallet paying; it's the government's wallet.
There are three points to grasp in practice:
- The client company applies. The instructor doesn't need to be an application agent, just someone who can "accurately guide the system." That alone gives you an advantage over training companies.
- The subsidy target is "Off-the-Job Training (OFF-JT) of 10 hours or more." Half-day seminars are not eligible. If you use subsidies for sales, build your menu for 10 hours or more from the start.
- AI training is usually applied for under the category of "Training associated with DX." In this category, the confirmation from a certified support agency, which is often confused in explanatory articles, is unnecessary. Submit the plan notification to the Labor Bureau at least one month before the start of training.
This subsidy ends on March 31, 2027. A significant portion of the orders currently in the market is demand brought forward by the system, not a genuine desire for learning.
STEP 5: Conduct Training as "Market Research While Getting Paid"
The real asset that a training instructor brings back from a client company is not the fee. It is the inner workings of that company.
Think about it. Is there any other opportunity for an outsider to be in the same room as employees from multiple departments for more than half a day and hear about their work troubles directly from them? (There probably are, but training is currently the easiest way in.)
Therefore, while teaching during the training, you also record.
- Questions from participants. Questions are the verbalization of bottlenecks in the field.
- Who moves the room when they speak? You can see the actual decision-makers who are not on the organizational chart.
- Which tasks are still being handled manually with Excel?
- Which department made comments like "Our data is messy"?
This is why I wrote at the beginning that those who try to profit from training alone will disappear. People who count training as sales will vanish with the drop in unit prices and the end of subsidies. People who use training as research can connect it to the next business.
STEP 6: Create a State Where the Next Contract is Discussed on the Last Day of Training
The records from STEP 5 become the proposal itself.
"Organizing the questions from everyone during the training, your company's bottlenecks are concentrated around order processing. Shall we work together to fix this with AI?"
Whether you can say this on the last day is far more important than the training satisfaction survey.
The exit generally falls into three types:
- In-house Support: Support the implementation of AI into actual work on a monthly basis together with the employees upskilled in the training.
- Advisory Contract: Be continuously involved as a consultant for the CEO in tool selection and internal rule-making.
- Industry-Specific Service: If you find the same bottleneck in multiple companies, it becomes a product. One company's trouble is a project, but a trouble common to 10 companies is a business.
The common thread is the transition from one-time flow revenue to continuous stock revenue. Training is the entrance; the business starts from here.
STEP 7: Imagine April 2027 Before Starting Today
What will happen in this market in April 2027, after the subsidies end? Training that was effectively 500,000 yen will return to 2 million yen. Orders that were brought forward will disappear, and training orders will dwindle. This is critical for operators who relied solely on training.
But it's different for those who have done STEP 5 and STEP 6. You have a list of business flows, bottlenecks, and relationships with decision-makers for dozens of companies. Even if the training disappears, these assets do not. Rather, as training demand thins out, competitors will exit, leaving only the continuous contracts with remaining clients.
Subsidies do not create demand.
They only bring forward demand that should have existed in the future.
Those who read it as a period of consuming future sales will end with the deadline. For those who can read it as a period of market research that can be done while receiving compensation, March 31, 2027, will not be a deadline, but the date when the preparation is complete.
