"It's better to use AI to make existing businesses profitable than to try to profit directly from AI itself."
The dialogue between DMM Chairman Keishi Kameyama and note CXO Takayuki Fukatsu was full of the essence of how to earn in the AI era, so I will summarize it in my own way today.
To get straight to the point, the target isn't "AI itself."
First, here is what Mr. Fukatsu said he feared most:
"Standalone AI services have about an 80% chance of being absolutely crushed by the base models of OpenAI or Big Tech as they naturally get smarter."
Even if you build a slightly convenient AI tool, that function will be integrated into the main platform a year later and released for free. Mr. Kameyama also bluntly stated that "Small to medium AI businesses are high difficulty."
"It might be the era of AI, but it's questionable whether it's the era of AI startups."
I think this one sentence captures the essence.
So where do you make money? The answer is "Industry X × AI."
Logistics × AI, Factories × AI, Buildings × AI. Instead of making AI the protagonist, you move to the side of making existing businesses profitable with AI.
Recently, former Yahoo President Takao Ozawa also mentioned in another talk that "If you're starting a business, Industry X × AI is the only choice," and the conclusions of those on the front lines are perfectly aligned here.
What gave me the biggest chill in Mr. Fukatsu's explanation was the talk about how "Payment and labor will become instantly exchangeable."
Until now, if you wanted to increase manpower, you had to post jobs, interview, train, and increase office space as the headcount grew. It took months and was full of costs.
As AI agents get a bit smarter, it will look like this:
"I paid the subscription; today my development power is 10,000."
"Development is finished, so I'll reduce development power to 10 and increase sales power to 1,000,000."
It's a world where you move sliders to allocate internal labor like a simulation game. Labor will be adjusted instantly from zero to infinity through payments rather than hiring.
In this world, what happens on the human side is the "Super Side Hustle."
If you are someone who can bundle and command AI agents, as long as you design the instruction manuals and business operations, the AI will handle the rest. Therefore, a person in a decent position at a large company could simultaneously serve as a department head for two or three other companies as a side hustle.
Capable people will be in higher demand than ever.
Companies will also shift toward headhunting people who have "experience building their own business with AI" as executives, rather than raising new graduates from scratch. The criteria will be proactive people who have run startups somewhere, rather than those who just have a few years of experience after graduating from a prestigious university.
Conversely, consulting was singled out as being in danger.
Mr. Kameyama's reading is as follows:
In uncertain times, large companies hate the risk of increasing employees, so they use outsourcing, dispatching, and consulting. So, these will grow for the next two or three years. But at some point, they will be cut off abruptly.
Mr. Fukatsu agreed, saying, "In semi-intellectual labor where education and training periods are long, AI is likely to become a rival."
Currently, these are star positions with high salaries. But if you go there just because "the economy is good," you might end up as just another freelancer. If you go, you must be prepared to evolve into someone who can handle everything from gritty field work to high-level upstream design.
According to Mr. Kameyama, if abilities are equal, staying in a large company will protect you for a while. This is because when large companies cut costs, they cut external sources first.
So, where should one choose to go now?
Mr. Fukatsu's answer was interesting:
"Trading companies might be more interesting."
The reason is that at a trading company, you can learn everything from planning to commercial flows, logistics, and money flows. And in the coming era, AI will handle all the practical tasks you've learned.
In other words, it's an environment where you can acquire all the basic skills to become a "One-Person CEO." You can leave the practical work to AI and move to the side of designing the whole system.
Mr. Kameyama's supplement was also practical:
Since large trading companies are highly departmentalized, it's better to join a company of about 100 people and see everything from upstream to downstream right next to the president. If you aren't going to start your own business, join a company where you are close enough to talk directly to the president.
There was also talk about how to choose a market.
Bring specialized AI with unique data into domains where IT adoption is slow, like agriculture, forestry, or fisheries. Alternatively, take the latest AI to factories, hotels, inns, or supermarkets that dominate a local region outside the Tokyo metropolitan area and compete on a regional basis.
Small market domains like tatami makers or bonsai are also strong. Because they are narrow, big players won't come to compete.
In fact, DMM itself has fought this way, being a company built on the idea of "attacking only niches where we can take 1st place, and standardizing only HR, general affairs, and legal. If 60 niches come together, it becomes fairly large."
For product creators, there were more specific guidelines.
Mr. Fukatsu says he tells his investees, "Stop making AI that people have to use."
The focus should be on the type of AI that starts up and works on its own without being 'used.'
For example, something that periodically checks Google Calendar and automatically pastes a summary before a meeting saying, "The person you are meeting is this kind of person from this kind of company."
Instead of making a new app, embed it into everyday tools like Slack, LINE, or email. This "trigger-type" approach is more logical.
This is because most people can't come up with prompts for AI. They can't explain the current situation well, and they can't put what they truly want into words. Leaving that entirely to the user is the current challenge for AI.
And as for performance itself, in another year or two, no matter which one you use, they will all be so smart that there will be no difference.
The competition won't be about performance, but about design that integrates into workflows without the user even being conscious of it.
Finally, the talk turned to the side that won't be eaten by AI.
Mr. Kameyama says, "Almost everything you can see on a smartphone is in trouble."
When AI-made content floods in, something made for 10 billion yen will be buried under a million pieces of content made for 50,000 or 60,000 yen. It becomes a world where you can't tell what's real, and creators feel foolish.
Conversely, the value of real-world experiences like camping, BBQs, and concerts will rise.
And the "Process Economy," where the process itself becomes the product.
If it's a drawing, don't compete with the finished image; instead, have everyone watch a live drawing session while drinking. Listen to opinions via SNS polls and let everyone participate in the process until completion.
If you compete only on results, you will collide with AI. The process and shared experience are difficult for AI to take away.
In the world of talent, AI talents who have no risk of scandal and can work 24 hours a day will become more rational for companies. That's why what remains for humans is the side of "gathering and sharing impressions."
Mr. Kameyama's closing words were the best:
"The troublesome thing is that this is a game you can't quit."
Even if you have mixed feelings about a world where AI spreads, as long as others are doing it, there is no option not to. He continues:
"You're going to fail anyway. Things will work out after about 10 failures, so I want you to accumulate failures. More than seeing it through, just start."
If I were to translate this into actions for tomorrow, I think it would be these two things:
First, think of one multiplication of "Your Field × AI" rather than "AI itself" in your industry. Multiplications that only people who know the field can create are the hardest to compete with.
Second, if you are doing work that is completed entirely inside a screen, think about how you can turn the process or experience into a product.
Those who start can begin clearing those 10 failures.
What kind of "× AI" exists in your field?
