Beyond the Answer
As AI becomes both coder and tutor, teachers ask what students still need to learn and what role they should play.
As AI becomes both coder and tutor, teachers ask what students still need to learn and what role they should play.
According to Satya Nadella, artificial intelligence (AI) writes 20 to 30 percent of the code in Microsoft’s repositories, with some already passing the halfway point. Sundar Pichai puts Google’s figure at roughly a quarter to a third. At my own startup with 15 engineers, no one handwrites code anymore. Two years ago, any of these claims would have sounded like science fiction.
Linus Torvalds, who created Linux, the operating system that powers every Android phone and most of the world’s servers, has pointed out that compilers write 100 percent of everyone’s code and nobody boasts about that. He is right that this has happened before. The difference is that a compiler is reliably correct, and a large language model is not.
If professional software is increasingly written by machines, what do schools have left to teach?
I put that question to four people from very different disciplines who do this for a living: in schools, in coding boot camps, in research, and in the creation of Japanese school textbooks. Surprisingly, there was a good deal of consensus.
That Which Survives
Simon Guest, a computer science educator and former chief technology officer of Code.org, the nonprofit behind the Hour of Code and the largest provider of free computer science curriculum in US schools, teaches AI at both the high school and undergraduate levels. He explains what he thinks endures using house construction as a metaphor.
“The student could ask an AI to ‘build me a house’ in one shot and simply accept whatever it produces,” he said. “Or they could design the house first, deciding on the structure and what each room needs to accomplish, and then use AI to help build each part. Using the second approach, AI may still do much of the construction, but the student is the ‘builder’ of the system.”
Stefania Druga holds a PhD in AI tools for creative coding, created the Cognimates platform, and is currently a staff research scientist at Sakana AI in Tokyo. She arrived at the same place after a decade of studying children learning alongside these tools. When agents write more of the code, she said, “knowing how to decompose a problem, specify what you want, and evaluate what comes back matters more, not less.”
Kani Munidasa cofounded Code Chrysalis in Tokyo in 2017, initially helping people transition into software engineering, and today works primarily with companies to develop their in-house software development and organizational capabilities. He put it in the language of hiring.
“As building software becomes faster and cheaper, the scarce capability is increasingly not simply, ‘Can you code?’ but ‘Can you decide what to build, and can you build it effectively as a team?’ That’s an organizational capability,” he said.
This is not new. I have hired senior engineers for 15 years by asking why they built the thing they are describing—what prompted the migration, who decided, and whether they understood the reasoning. If somebody simply told them to do it, the interview is largely over. What’s new is that I now ask the same of juniors. The judgment that used to earn a promotion is becoming the price of entry.
AI as Teacher
Let’s do a little philosophical exercise. AI could be the perfect teacher. It is infinitely patient. It never tires of the same question asked a fourth time. It explains at whatever level you need, in whatever language you speak, and it will do it at two in the morning.
Munidasa does not hedge: “I think AI will eventually do a lot of what we traditionally think of as teaching, and probably do some of it better than we can,” he said. “It can already explain concepts, answer questions endlessly, adapt to the individual, review code, generate exercises, and give immediate feedback. We shouldn’t pretend those capabilities are somehow protected from AI.”
Of the years Code Chrysalis spent running consumer boot camps, he said: “People weren’t paying us for access to information. Even when we started, you could learn JavaScript, algorithms, databases, or almost anything else online for free. What they were paying for was an environment designed to transform how they learned and worked.”
The evidence for that predates AI. When Harvard put CS50, its introductory computer science course, online for free, 150,349 people registered in the first year. Just 1,388 of them finished. On campus that same year, 703 of 706 students completed the identical course, taught by the same instructor from the same material.
Guest does not think the difference is pedagogical.
“When a student signs up for an in-person course, they’re accountable to another person,” he said. “If they don’t show up or complete the work, there’s a teacher or professor who notices.” Online, most of that disappears. Anyone who has joined a gym in January and stopped going by spring already knows how that story ends.
“I am much more optimistic about AI as a ‘teaching assistant’ rather than a teacher—an AI that can provide immediate, personalized help, while a human teacher remains responsible for the relationship, motivation, and accountability,” he said.
“AI may become an extraordinary teacher, but education still has to develop the human being.”
Who It Fails
“The people who can benefit most from self-directed AI learning are often the people who have already learned how to learn without it,” Munidasa noted.
“The danger,” he said, “is when you don’t yet know what you don’t know. AI can give you an answer that feels complete, and even produce something that works, without you understanding why it works.”
Druga explained it this way: “An answer machine optimizes time to answer. A teaching tool optimizes what you can do after it leaves the room.”
In her most recent study, North American participants already had AI at home, while several from other regions had never used it. “The tool amplifies whatever head start a child already has,” she said. “That is why access has to be designed, not assumed.”
The Adult in the Room
Japan has already tested whether access on its own is enough. In 2018, Japanese students spent less classroom time on digital devices than students in any other country in the Organisation for Economic Co-operation and Development. The government’s response, announced the following December, was the Global and Innovation Gateway for All, or GIGA, School Program, which put a networked device in the hands of every public school student, with 1,769 local governments, or 97.6 percent, having completed planned delivery by the end of March 2021.
Distribution was never the hard part.
The education ministry’s own reviews now describe uneven use between regions and between schools.
The constraint is the adult at the front of the room. A survey by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) for FY2022 found that 23 percent of junior high school teachers responsible for technology were teaching under a temporary license or an out-of-field arrangement.
That is easy to read as a competence problem, but I do not think it is one. “The gap is teacher confidence, and tools built for learning rather than for answers,” Druga said.
Ayana Murakami, who works on the digital content in Japan’s authorized junior high school textbooks and is a software engineer and PhD candidate at Ochanomizu University, describes a second structural problem. “The national curriculum is revised roughly once a decade, and junior high school textbooks follow a four-year cycle, while AI can change dramatically within months.”
The ministry appears to agree. Its budget documents describe a new junior high subject under consideration, provisionally called Information and Technology, and say work on teaching materials will begin without waiting for the next curriculum revision to take effect.
Murakami is not pessimistic about schools. Asked whether compulsory education is better placed than self-directed learning to help students understand the potential, risks, and use of AI, she said: “AI makes it easier to learn on your own, but it can still be hard to study things you are not interested in. Schools provide a common foundation by giving students a structured learning environment.”
What’s Left
Having granted AI that much, Munidasa draws a conclusion that runs against his own interest. “But if that’s all a boot camp does, then yes, I think AI should replace it.”
What survives is the part that was never about information. “You don’t learn teamwork by reading about teamwork. You learn it by having to build something with another human being when you don’t agree.”
Asked what she would fund in Japan if she could fund one thing, Druga did not name a tool for learners. She chose “AI tools that assume a teacher is present and make her better.” Every person I spoke to for this piece said a version of the same thing.
A country where every child has a device, a teacher, and a teacher confident with both would be the first of its kind. Japan already has two of those three.
The third is the hardest. Teachers need confidence with these tools, as well as AI built for learning rather than for answers. Current AI products will not do that out of the box. They are built to answer. So tools that enhance and support the learning process have to be built.
MEXT is reconsidering the curriculum now, and the obvious move is to add AI to the syllabus. But nothing in a syllabus reaches a student except through a teacher. What we do know is that accountability matters more than it did, and that what is worth teaching is the reasoning rather than the result. As Munidasa put it: “AI may become an extraordinary teacher, but education still has to develop the human being.”
AI in the Workplace
As artificial intelligence gets its own desk in the office, workers and leaders weigh its risks and rewards—and whether age shapes their view.
As artificial intelligence gets its own desk in the office, workers and leaders weigh its risks and rewards—and whether age shapes their view.
A client question nobody has an answer for leads to a stalemate in the middle of a meeting. One colleague takes the old-school approach and books another meeting for next week, spending the days between sketching notes. The other opens an artificial intelligence (AI) tool and has a workable plan within the hour.
That contrast, said Roy Tomizawa, chief executive officer of Reinventing Asia, is the story of AI in the workplace: it is changing the distance between thinking and acting.
“So much organizational time is spent waiting and procrastinating,” Tomizawa said. Unread emails and slow-to-schedule meetings are a kind of quiet tax that most companies never measure. “Half of what was discussed could have been thought through, drafted, or prototyped beforehand by the person who leveraged the available AI tools on his or her own.”
How businesses close the gap between thinking and acting is dividing opinion, forcing leaders to decide what AI should handle and what it shouldn’t, while young professionals are discovering that the skills worth building are shifting day to day.
The Age of the Generalist
Tomizawa explained that the shift goes beyond speed. AI is dissolving the old dependency on specialists. An employee who knows how to assemble credible first versions of a website or a video will have an advantage. Tomizawa called this “the age of the generalist.”
But that doesn’t flatten the value of expertise, he said.
“It actually raises the price of real expertise, since shallow, surface-level work is now cheap and abundant.” Someone still has to know when an AI’s answer is actually good, he added. Without that, the same tools that make people efficient just make them “efficient masters of AI slop.”
Sludge, Slop, and Everything in Between
Not everyone sees the acceleration in such rosy terms. Dr. Greg Story, president of Dale Carnegie Training Tokyo, compared the effect of AI on communication to that of social media. Both are technologies that can make people more insular, offering contact with or through a machine as a substitute for real contact with each other.
Story, who has written nine books and recorded thousands of podcast episodes, said listeners can identify his writing by ear. Hand that same task to AI by default, he argued, and you never develop a personal voice.
“If you’re ever going to have a distinctive style, and you keep outsourcing it to AI, then you’ll never achieve that and you’ll be part of the sludge,” he said. “There’s no discrimination between one piece of sludge and the other.”
For most professionals, the risk, he argued, is that AI tends to default to generic text unless given deliberate direction. That caution shows up in how he runs his own training business. Dale Carnegie has built its own AI system, trained only on its own material, to keep answers on brand rather than generic. But Story keeps it confined to prework and follow-up, away from the live instruction that’s actually being sold.
In sales, he explained, AI becomes a genuine research tool for those who ask the right questions, provided someone checks it for hallucinations—the tendency for AI to make up information when it doesn’t have an answer.
“The human dimension is being reduced,” he said, “so we have to hold on to it as much as we can to counterbalance the swing to impersonal communication.”
“AI is dissolving the old dependency on specialists. An employee who knows how to assemble credible first versions of a website or a video will have an advantage.”
Finding a Voice with AI
Ted Katagi, CEO of Kenja K.K., builds tools based on retrieval-augmented generation, or RAG, that draw only on a company’s own material rather than the open internet, including the system behind Dale Carnegie’s own AI tool. Katagi makes a distinction between content and style.
“We have to separate things—the voicing of it, which to me is the style, versus the content,” he said. The content has improved with AI use, he said, but the style “will tend to aggregate” as writers rely on it. Part of that, he argued, is simply that AI eliminates bad writing—which raises its own question. “If we get rid of the bad writing, isn’t that going to make you a little bit more similar in a sense?”
Preserving an individual voice, he said, takes deliberate work. “I wouldn’t just say, ‘Here, AI, write like me.’ I would say, ‘Here are all my papers. How do you think I write?’” Skip that step, he warned, and a personal voice doesn’t get replaced so much as quietly forgotten.
Katagi’s advice to young professionals is to treat AI more as a “sparring partner” than a first-drafter.
“Ask always: What did it do that was smart that I need to learn?” Katagi advised. “What did it do that was bad that I need to check on?” For those who genuinely engage with AI, it “can seem like they gain 10 years of experience.”
“It will get you where you want to go and it will help you finish the last mile,” he said. But human judgment and intuition are still needed. He cited as an example a recruiting-industry deployment that builds an editor directly into the workflow rather than letting AI hand off a polished document unchecked.
Differentiating from Sameness
Meghan Barstow, president and representative director of ACCJ Corporate Sustaining Member Edelman Japan K.K., has watched two prior technology shifts reshape her industry: the internet and mobile phones.
“I experienced the advent of the internet and mobile phones,” she said, “both of which transformed how we create, communicate, and connect.” She expects AI to do the same, viewing it as a tool that “aids and amplifies human creativity rather than replaces it.”
Her concern is closer to what Story called sludge. Barstow called it “sameness.”
“AI is incredibly good at producing something plausible. But plausible isn’t necessarily interesting, original, or culturally meaningful,” Barstow said, adding that her worry is more about the behavior it enables.
“If we all use the same tools, trained on much of the same material, and accept the first answer they give us, there is a real risk that we will create more content but with less creativity.”
As an executive, Barstow, who is also an ACCJ vice president, described adopting AI as “quite fun,” but said the stakes reach into business models themselves, with real winners and losers ahead. Organizations that treat AI as additive to their core value rather than a replacement for it will succeed, she predicts.
For young professionals, she offered this advice: “Raise your hand. Get involved. Be curious. Learn, learn, learn. Use the new tools, absolutely, but don’t do so at the expense of learning the fundamentals or having experiences that develop your judgment and perspective.”
She also viewed the generational exchange as a two-way street. Edelman brought interns back this year for the first time since the pandemic, several through the ACCJ’s own internship portal, and Barstow credited them with making the organization sharper.
“The more opportunities we create for generations to learn from one another, the better all of us—and our organizations—will be,” she said.
What Leaders Need to Know
Digital fluency doesn’t mean becoming a technologist, Tomizawa said. It means understanding how value is created and decisions are made. He said three mindsets matter most:
- Treating information as a strategic asset
- Measuring outcomes, not activity
- Introducing AI deliberately
By treating information as a strategic asset, Tomizawa means that it is important to know where critical data sits, who owns it, and whether it’s reaching the people making decisions. Once that is in place, the value of the technology must be assessed. “High usage of an AI tool, for example, tells leaders little about whether people are saving time, making better decisions, improving customer outcomes, increasing revenue, or reducing risk,” he said. When preparing to implement AI tools, leaders should start by looking at the work itself to find the bottlenecks and errors before choosing the technology. Finally, they should define exactly what AI should recommend and where human approval stays mandatory.
Story said that leaders rarely had time to check employees’ work before AI, and they have even less reason to start now. His solution is to train people on privacy and instill in them a healthy skepticism.
Cybersecurity is another area to be mindful of.
Katagi said a clash between cybersecurity’s oldest principle and AI’s newest capabilities exists, pointing to “zero trust,” the model he said most cybersecurity experts rely on now.
“Zero trust means you say you’re Chris and it looks like it, but I’m not going to trust it until I see some irrefutable proof that you are who you say you are,” he explained. Agentic AI, a system trusted to act on a user’s behalf that Katagi said is at the center of AI right now, asks for the opposite. “Basically, what it’s allowing you to do is say, ‘Okay, trust me and I’ll do the thing.’ But that’s going directly against all the cybersecurity frameworks that people have.”
“The professionals who thrive will be the ones who know when to trust it, when to challenge it, and when to think differently.”
A Youthful Approach
Among young professionals, thoughts on AI lean more toward the opportunities the technology brings.
“AI is one of the most significant shifts I have seen in how we work, and I think young professionals have a real opportunity to shape how it is adopted,” said Fatim Diallo, a director at FGS Global and vice-chair of the ACCJ Young Professionals Forum (YPF). “Used well, it frees up time for the kind of creative thinking and relationship-building that transforms careers.”
The shift has made her more curious about where she can add value than anxious about what AI might replace. At a company dealing with multiple markets, AI has “meaningfully reduced the friction of working across linguistic and cultural contexts,” she said.
Her main concern is what she dubs a “meat proxy”—people who pass AI output straight through without reviewing or questioning it.
“The professionals who thrive will be the ones who know when to trust it, when to challenge it, and when to think differently.”
Kelly Langley, CEO of Gemini Group K.K. and co-chair of the YPF, noted a similar shift in what makes someone valuable. AI increasingly handles the “first 80 percent” of a task, he said, freeing people to spend more time exercising judgment on the final 20 percent.
“That makes me more interested in capabilities that are difficult to replicate: judgment, trust, persuasion, relationship-building, improvisation in moments of crisis, and understanding what motivates people.”
Langley pushed back on the idea that there is a generational divide among AI adopters.
“The bigger distinction is between people who are relentless learners and early adopters and those who remain comfortable with established ways of working.”
Progress in the Age of AI
The next wave of AI innovation isn’t going to come from the massive, companywide rollouts most large organizations are currently betting on, Katagi said.
“People think the big projects are the thing. I think the next innovation is going to occur in agile projects at the middle level,” he said, predicting smaller teams and narrower budgets focused on solving one specific problem well rather than applying a general-purpose tool everywhere at once.
Story said that “at the end of the day, most business problems are people walking around on two legs talking to each other. That doesn’t go away, but the skill sets are going away in a sense, as people become more isolated.”
But he sees an opportunity ahead in the people using the tools themselves.
“As we’re moving away from voice and body language, we are getting to a stage where AI is just going to accelerate that process,” he said. “If you have the ability to communicate, you’re going to stand out head and shoulders above everyone else. You’ll be clear, empathetic. I think that’s an opportunity rather than a threat.”
Tomizawa argued that most companies are still years from redesigning how work actually gets done around AI, rather than just using it to make things faster.
“The more capable AI becomes, the more important human judgment becomes in deciding what problem deserves attention, what ‘good’ looks like, and which trade-offs are acceptable,” he said.
Five years from now, he added, the people with the greatest value will be the ones who know how to combine machine capability with sound human judgment.
AI 1988
Look back at how members viewed the potential of AI in 1988 in this reprint from the ACCJ Journal archives.
Look back at how members viewed the potential of AI in 1988 in this reprint from the ACCJ Journal archives
Artificial intelligence (AI) is becoming more and more widely used in Japan. Companies in this country are coming to realize that if they want to stay internationally competitive, they have to incorporate this technology. Davis is a popular contributor to these pages who specializes in high-technology subjects.
Japan’s large electronics companies are constantly looking for ways to boost the efficiency of their administrative work while searching for new markets so as to diversify their business. AI, a type of sophisticated computer programming that promises to revolutionize many job-related tasks, is catching on among electronics companies and software businesses here, and is also a target of interest among trading houses. All of these enterprises want to establish a foothold in this up-and-coming technical sector so as to enhance their long-term prospects. AI systems in the years to come may make or break certain companies in highly competitive areas of business. It would not be an exaggeration to say that a “mini boom” is being seen today in Japan’s AI sector.
AI systems under development here are intended to address bottlenecks within corporate product-development departments, and they are being marketed to outside customers, sometimes together with special types of data processing equipment. One of the best ways to sell more computer hardware is to market such equipment with an emphasis on higher value-added features, such as the ability to effectively handle AI writing tasks. However, Japanese companies in this field are all well aware that they have to contend with the likes of Symbolics Inc., a Cambridge, Massachusetts, global leader in AI workstations.
Two years ago (in 1986) the Artificial Intelligence Association of Japan was established in Tokyo by electronics companies, telecommunications businesses, software houses, and others interested in new developments in computer programming. The association cultivates exchanges between researchers in various AI-related fields and disseminates technical information to its members. Moreover, the association promotes specialized training of so-called knowledge engineers and other experts needed for the advancement of the new discipline. Establishment of the special association signifies the maturation of the initial commercial phase of AI here.
In contrast to Japan’s AI infrastructure, state-of-the-art American AI work is typically dominated by clusters of small businesses mainly located around major universities. In fact, many Japanese AI specialists have studied at leading US universities. As a result of the difference in the two paradigms, the large electronics enterprises of Japan have tremendous potential resources to devote to AI studies, whereas in the US, venture capital must typically be raised to fund much of the innovative work in AI.
“The global market for AI systems is likely to grow to as large as much as $10 billion per year sometime between 1995 and 2000, according to Japanese electronics industry estimates.”
The Japanese approach to the AI business often relies as much on proximity to leading US universities as it does on relationships with the top Japanese universities. In other words, Japanese universities are not major actors within the immediate sphere of AI business here. The paradigms are not without exception, however, because some smaller businesses in Japan, such as CSK Corp., are doing work in the field as well.
As AI is widely considered a promising growth market within the information processing sector, electronics companies are offering products that will allow users to develop their own AI systems, such as so-called expert systems. This customized programming is developed on the basis of experts’ knowledge; hence expert systems comprise handy tools for novices—so that they may easily draw upon the comprehensive knowledge of specialists to assist them in complicated tasks, such as writing specific types of software programs.
The global market for AI systems is likely to grow to as large as much as $10 billion per year sometime between 1995 and 2000, according to Japanese electronics industry estimates. The leading AI language today is LISP (LISt processor), and it is widely expected to retain its front-running position. Four of the largest AI applications expected in the mid-to-late 1990s are those for integrated circuit design assistance, manufacture planning, financial planning, as well as computer systems diagnosis and maintenance.
An example of a medical application of AI systems is the so-called RINGS program—rheumatology information counseling system—developed recently by Nippon Telegraph and Telephone Corp. and a medical college in Tokyo. The system is used by those suffering from rheumatism to help them in diagnosing minor problems over the telephone. When more serious problems arise, doctors are to be consulted. A variety of other medical-related AI systems are now under development, in part because the medical sector is likely to see rapid growth due to the aging of Japan’s population.
In the area of nuclear power plant operations, a group of Japanese enterprises is developing an expert system to enhance the safety of pressurized water reactors (PWRs). The LISP-based expert system is intended for use in new types of PWRs to be operated by Kansai Electric Power Co., Inc. and three other electric utilities.
Greater safety in operating nuclear plants can lead to enhanced profits for the utility companies, as they will not need to shut down reactors for prolonged periods in order to do repairs, precautionary tests or other types of maintenance.
The most prominent of Japan’s AI-related development programs is the so-called fifth-generation computer project, which is administered by the Institute for New Generation Computer Technology (ICOT). The institute was established in 1981 under funding from the Ministry of International Trade and Industry’s (MITI’s) Machinery and Information Industries Bureau.
“Although today’s AI systems can only cope with surface level knowledge, those of the year 2000 are likely to be capable of dealing with more abstract forms of knowledge. ”
Altogether, there are nine private companies participating in the project. Researchers based at MITI’s Electrotechnical Laboratory (ETL) in Tsukuba, Ibaraki Prefecture, are also involved. Moreover, the ETL, which is administered by MITI’s Agency of Industrial Science and Technology, is doing its own independent work in the field. Six to eight researchers from each of the electronics companies work at the ICOT center in Tokyo, and then only for periods generally ranging from two to four years.
When the project began in 1982, it was the subject of considerable attention throughout the world, due to its bold proposals and the perceived threat that it posed to the American and European computer software industries. However, recently it has not attracted much interest because Americans and Europeans have been less than impressed by the meager results of the project. US interest in the ICOT project led to the establishment of Microelectronics and Computer Technology Corp., a research consortium headquartered in Austin, Texas.
AI systems have a long way to go before they reach a phase of maturity. Although today’s AI systems can only cope with surface level knowledge, those of the year 2000 are likely to be capable of dealing with more abstract forms of knowledge. Advances in the memory capacity of computer microchips, parallel processing capabilities of computers, data processing speeds, and knowledge bases will accelerate the progress of the AI business sector.
Let us hope that people will always be able to keep the upper hand of control on such advanced tools as AI systems, and that the sophisticated tools won’t ever “discard” the humans they are supposed to be helping.
Your Own Private AI
As AI evolves, businesses are turning to custom LLMs to unlock corporate resources.
As artificial intelligence evolves, custom systems unlock business resources
Konosuke Matsushita was one of Japan’s greatest entrepreneurs. As the founder of a light socket company that evolved into Panasonic, he inspired legions of salarymen with his business wisdom. Twenty-five years after his death, the “god of management” was effectively resurrected as an artificial intelligence (AI) model. A chatbot trained on his writings and speeches can produce eerily lifelike Matsushita answers, according to one relative, and will eventually be used to make business decisions. It’s a dramatic example of how businesses are using AI to leverage intellectual property built up over decades.
The past few years have seen an explosion of AI applications based on large language models (LLMs) and tools such as OpenAI’s ChatGPT and Google’s Gemini. They have been used for everyday tasks such as writing text for slide decks, lessons, and articles, as well as synthesizing search results as in Google’s AI Overview that now appears with most searches.
LLMs are based on computational systems using neural network transformers that perform mathematical functions. Measured by the number of parameters they contain, LLMs learn by analyzing vast amounts of text from books, websites, and other sources. During training, the model identifies patterns, relationships among words, and sentence structures. This process involves adjusting millions of parameters—values that help the model predict what comes next in a sequence of words.
A major problem with LLMs and generative AI, however, is that they usually draw entirely from online content and thus are prone to inaccuracies. AI hallucinations, as they are called, occur when LLMs observe patterns in the data that are nonexistent, or at least imperceptible to humans.
One solution is private AI. It brings the power of LLMs inside a company, where queries are secure and limited to the company’s own data, reducing the risk of security leaks and incorrect or misleading responses. Private AI has traditionally been limited to government, defense, finance, and healthcare users, but it’s spreading to a broader spectrum of industries due to fears about intellectual property theft.
“[Private AI] brings the power of LLMs inside a company, where queries are secure and limited to the company’s own data, reducing the risk of security leaks and incorrect or misleading responses.”
Kenja KK, a member of the American Chamber of Commerce in Japan (ACCJ), is opening up the market in Japan to private AI. The Tokyo-based company offers AI solutions for enterprises that include purpose-built expert systems, incorporating a relatively new AI technology called retrieval-augmented generation (RAG).
Bearing a name coined as recently as 2020, RAG relies on a predetermined collection of content to improve the accuracy and reliability of generative AI content. Kenja offers a self-service plan for small and medium-sized businesses and a more comprehensive enterprise plan for businesses.
“Private AI is the next frontier,” said Kenja founder and Chief Executive Officer Ted Katagi, who is also chair of the ACCJ’s Marketing and Public Relations Committee. “All companies face the same issues: you have very sensitive data that you don’t want to make accessible to everybody at the same time. Private not just in terms of someone outside the company, but within the company, too. You may not want HR data to be shared with people in finance, for example. That’s an issue you want to solve, and we solve that.”
Kenja users create so-called rooms where they can upload thousands of documents or other content, organizing this into topic-specific folders. The process can be automated, and Kenja can train and fine-tune the system. For instance, it can be taught to forget certain words or trained to understand a balance sheet in order to do financial tasks such as due diligence.
“You are kind of building a wall around a set of information and telling it to only use what’s in this area,” explained Katagi. “Having 85–90 percent accuracy—which is what current generative AI, such as ChatGPT, Gemini, or Claude, will give you—is not good enough. Private AI models that are fine-tuned and query a closed set of materials can close that gap.”
Private AI is being used in surprising applications. Just as Panasonic has cloned its founder in digital form, Dr. Greg Story is using Kenja to share the teachings of another business luminary, Dale Carnegie. The self-improvement guru from Missouri wrote a book in 1936, How to Win Friends and Influence People, that still counts among the world’s all-time bestsellers. As president of Dale Carnegie Tokyo Japan, Story has been teaching Japanese businesspeople about leadership, communications, and other skills in Dale Carnegie seminars for the past 14 years. Dale Carnegie started in Japan in 1963.
Since learning about the impact of content marketing, he has built up an enormous corpus consisting of white papers, e-books, printed books, course manuals, 270 two-hour teaching modules, as well as video and audio recordings that include hundreds of podcast episodes. He has penned a series of books himself in English and Japanese that includes Japan Sales Mastery, Japan Business Mastery, Japan Presentations Mastery, and Japan Leadership Mastery.
“If you like the cut of our jib and you want a Dale Carnegie point of view and a curated, trustworthy response, we provide that through this AI.”
The material was scattered in different places, and when clients began asking for on-demand training, Story decided to get ahead of the curve by including all his company’s content in AI-curated form, something public chatbots cannot do.
“ChatGPT will give you everything it can scrape together, but it’s everything and therefore nothing,” said Story. “You get generic answers, and you don’t know if they’re trustworthy. But if you like the cut of our jib and you want a Dale Carnegie point of view and a curated, trustworthy response, we provide that through this AI.”
Story thinks the technology can benefit businesses that have substantial bodies of work to draw on, but those that don’t will get thin answers. He adds that using tools such as those from Kenja will not only help his company learn about the benefits of AI, but it will also give it an edge over competitors. He plans to roll out his AI offerings in 2025, delivering customized responses to students’ questions in English or Japanese on topics ranging from sales to diversity, equity, and inclusion.
Could there be a Dale Carnegie version of the Matsushita chatbot one day?
Kenja has begun working with Dale Carnegie’s global team to do just that, and has developed a prototype revival of Dale Carnegie’s voice, avatar, and writing style. The writing style and word generation are done with Kenja RAG AI technology.
“Carnegie became a global superstar in a non-digital world,” noted Story. “There’s no question we can get an AI to read a script generated in his style, in his voice. It’s amazing.”
AI Audits
AI's ability to analyze considerable amounts of information quickly offers great potential for auditors. How is this rapidly evolving tool impacting the audit process?
What impact does artificial intelligence have on auditing?
In the Fourth Industrial Revolution, with increasingly developed technology, the application of artificial intelligence (AI) has become popular both in everyday life and work. Thanks to AI’s outstanding features, multiple tasks can be performed in less time and with less effort. For this reason, AI is playing an important role in some areas which require the processing of large amounts of information, such as auditing. So how does AI affect the work of an auditor?
AI is defined in the Oxford English Dictionary as “the capacity of computers or other machines to exhibit or simulate intelligent behavior; the field of study concerned with this.” In later use, it is also defined as “software used to perform tasks or produce output previously thought to require human intelligence, especially by using machine learning to extrapolate from large collections of data.” In other words, AI is used to perform tasks that require human intelligence through programmed algorithms.
AI processes supplied information and produces results after conducting an analysis. Thus, AI can convert huge amounts of data in a short time and increase work efficiency.
With those exceptional aspects, AI can be applied at many stages of the audit process. How is it used and how does it affect the efficiency of the audit?
This can be broken down into three stages:
Audit Planning
Based on the customer data input, AI will propose appropriate audit procedures to optimize the audit plan.
Risk Assessment
Using the provided information, AI will analyze past trend fluctuations and financial indicators. Since AI can process and synthesize considerable amounts of information, the analysis will be more specific and more effective, giving auditors a deeper view of the business’s situation. Accordingly, auditors will identify potential risks more accurately and provide more appropriate material. As a result, AI also can assist auditors to predict the potential financial situation and determine the reasonability of financial forecasts as well as potential future risks of the business.
Substantive Procedures
At this stage, auditors must perform many repetitive tasks, such as checking details of documents (e.g., invoices and contracts), matching data among documents, and verifying the accuracy of the financial statements. AI can perform these tasks automatically through programmed algorithms, allowing auditors to review more data promptly with a higher level of accuracy in less time than a traditional audit.
By using AI to analyze and process the large volume of transactions, auditors can easily detect anomalies, errors, and risks in financial data. It allows auditors to focus more on high-risk areas that are prioritized, thereby improving the audit quality. Additionally, AI has a function known as machine learning which allows the system to learn from past data and improve its performance, thus enhancing accuracy and effectiveness.
It can be seen that applying AI in auditing brings many benefits. On the one hand, labor savings and productivity increases are the prominent characteristics of AI system. And with the ability to review and analyze information on a wide scale, AI can help identify fraud or potential risks that may be overlooked, thereby improving risk assessment and strengthening audit quality.
“With the ability to review and analyze information on a wide scale, AI can help identify fraud or potential risks that may be overlooked, thereby improving risk assessment and strengthening audit quality.”
However, using AI still has certain limitations:
AI works based on the provided data, so ensuring that the data is accurate, complete, and taken from reliable sources is crucial. Additionally, with a colossal volume of data, errors are likely to occur during analysis and processing. This can result in inaccurate conclusions and affect audit results.
Another limitation relates to cybersecurity, as using AI requires an internet connection. Therefore, if the internet system is compromised, AI algorithms could be altered and routed to discrepancies in AI operations.
AI is, after all, a machine set up by humans which performs tasks based on pre-established patterns. Hence, AI cannot respond to or handle unforeseen situations. Moreover, maintaining an attitude of professional skepticism is extremely important during the audit process to minimize potential risks. Nevertheless, the nature of AI is mechanical, so it is impossible to possess this skepticism when analyzing information and handling situations as auditors do.
Another limitation is that AI might not be able to satisfyingly detect fraud or window dressing in accounting that may occur in a business, because AI lacks the ability to think and evaluate like humans. Fraud detection requires auditors to possess a professional skepticism to assess the evidence collected during the audit and to evaluate it based on the business operations and internal control activities.
With workloads increasing, the benefits that AI provides are indispensable and will have a positive impact on the audit process. However, auditors should use AI appropriately and not abuse it or rely entirely on it, because AI is precisely a tool and cannot solve complex issues that require human decision-making on a system basis. Consequently, the balance between using AI and manual work in auditing is prerequisite. Furthermore, auditors should be equipped with the necessary knowledge and skills to fully understand AI’s operations, employ its produced results, and avoid cybersecurity attacks.
For more information, please contact Grant Thornton Japan at info@jp.gt.com or visit www.grantthornton.jp/en
Disclaimer: Opinions or advice expressed in the The ACCJ Journal are not necessarily those of the ACCJ.
Synthetic Savants
Since the introduction of consumer-facing artificial intelligence applications such as ChatGPT and Google’s Bard, generative AI has transformed how people work around the world. How might it impact specific industries in the years to come?
As generative AI sweeps the world, how will it transform the way we work and innovate?
We live in an age of intelligent machines. Since the introduction of consumer-facing artificial intelligence (AI) applications such as OpenAI’s ChatGPT and Google’s Bard over the past year, generative AI has transformed how people work around the world.
From $40 billion in 2022, the market size for generative AI will balloon to $1.3 trillion over the next 10 years, according to Bloomberg Intelligence. First popularized through image generators, the technology has been applied in fields ranging from neuroscience to advertising, sometimes in surprising ways.
Generative AI programs like the large language models powering ChatGPT are trained on enormous volumes of data to sense patterns and predict how they will play out in a piece of content. These models can be trained on linguistic, financial, scientific, sensor, or other data—especially data that is uniform and structured—and can then create new content in response to user input. They have had remarkable success, particularly in image and text generation, and have seen rapid uptake in sectors ranging from education to computer programming. “This technology is set to fundamentally transform everything from science, to business, to healthcare … to society itself,” Accenture analysts enthused in a report. “The positive impact on human creativity and productivity will be massive.”
Powerful New Assistants
Generative AI first gained public attention thanks to its ability to change how we communicate through words, images, and video. It’s no wonder, then, that the world’s largest public relations company has embraced it. Edelman worked with OpenAI to launch the original ChatGPT-2 and delivered the first application in an ad campaign. In the spots for Hellmann’s Mayonnaise, the tool is tasked with finding new ways to use leftovers.
Edelman believes the technology will reconfigure the communications industry, but it won’t replace human ingenuity, strategic advice, and ethical decision-making that builds trust, said Meghan Barstow, president and representative director of Edelman Japan.
“We predict that AI will become an essential assistant in our work, helping to brainstorm, research, summarize, trend spot, monitor media, and generate content, among other tasks,” explained the ACCJ governor and chair of the chamber’s Communications Advisory Council. “The emphasis here is on ‘assistant,’ as we believe there will always be a human in the loop, that AI and people working together will provide the most effective and valuable work output.
“As with any technology, there are risks that require appropriate caution, education, processes, and policies to ensure the safe and trustworthy use of generative AI to protect our work, our clients, and end users from issues related to disinformation, bias, copyright infringement, and privacy.”
Similarly, lawyers such as Catherine O’Connell are also using generative AI as smart assistants. O’Connell is principal and founder of Catherine O’Connell Law and co-chair of the American Chamber of Commerce in Japan (ACCJ) Legal Services and IP Committee.
After taking a course on how to get the most out of ChatGPT, she has been using it for writing keynote speeches, article outlines, posts on social media, and skeletons of presentations. She compares the tool to a human intern, and praises its time-saving efficiencies, but warns that it should not be used for legal work, such as contracts or legal advice. Attorneys in the United States, she noted, have found themselves in trouble after producing legal filings referencing non-existent cases that generative AI simply made up.
“Generative AI is like a teenager that has a lot of promise but has not learned how to be a whole professional yet; it needs guidance,” said O’Connell. “However, in terms of an idea generator or idea expander, a time-saving device, and an assistive tool, generative AI is an asset. The rest falls to me to add my human touch to check and verify, to add my own personality and insights only I have, and to make the output my very own. I think generative AI is so good that its cousin, Google search, may be out of a job sometime soon.”
Smart Tools for Talent
Recruiting is another industry in which workers deal with mountains of structured data, in the form of resumes and online posts, that can be utilized by AI. Robert Half Japan, an ACCJ Corporate Sustaining Member company, uses a system called AI Recommended Talent (ART) to match resumes to client needs. The system speeds up matching for job hunters and employers, allowing staff to spend more time with clients.
“The real power of generative AI is how much it can integrate with our existing systems,” explained Steven Li, senior division director for cybersecurity. “We are piloting ChatGPT-4 integration in our Salesforce CRM. Studies have shown benefits from integrating generative AI into workflows. Other industry examples that highlight the benefit of integration include the GitHub CoPilot generative AI feature.”
The effectiveness of AI in recruiting has led some people to speculate that it could render many human recruiters obsolete. Deep learning algorithms are figuring out what a good resume looks like, and generative AI can craft approach messages and InMails, a form of direct message on the popular LinkedIn platform, noted Daniel Bamford, Robert Half’s associate director for technology.
“However, the real value of agency recruitment is not, and never was, a simple job-description-to-resume matching service,” added Bamford. “Agency recruitment done well is a wonderful journey of problem-solving, involving the goals of organizations and teams and the values and desires of individuals. Excellent recruiters will thrive. They will use AI’s capacity to handle simple tasks like scheduling and shortlisting. This will free up time for high-value interactions, delivering even greater value for their partners and industries through the human touch. The future of excellent recruiters will be brighter with AI’s support.”
Tracking Ships and Patients
Even a traditionally hardware-oriented industry like logistics is being transformed by generative AI. Shipping giant Maersk is using a predictive cargo arrival model to help customers reduce costs with more reliable supply chains. It also wants to harness the power of AI to recommend solutions when shipping routes are congested, advising on whether goods should be flown or stored, and better understand the sales process, Navneet Kapoor, Maersk’s chief technology and information officer, told CNBC.
Maurice Lyn, head of Managed by Maersk for Northeast Asia, also sees great potential in the technology. “The biggest changes that I foresee will be related to the enhanced visibility into, and agility of the management of, the global supply chains of our clients on an execution level,” he told The ACCJ Journal. “The data aggregated will allow logistics service providers [LSPs] to deliver predictive and proactive solutions to our clients. If clearly interpreted by the LSPs, stability and uniformity of costs and deliverables will be provided globally and locally to our clients.”
Generative AI may even help us live longer, healthier lives via long-term patient monitoring. Sydney-based medical AI startup Prospection recently launched its first generative-AI model in Japan to analyze anonymized patient data for pharmaceutical companies so they can better understand patient needs. A Japanese drug company, for instance, could look at cancer patient outcomes across the country and find that they are slightly worse in a particular region, possibly because less-effective drugs are prescribed there.
Founded in 2012 and operating in Australia, Japan, and the United States, Prospection now has data on half a billion patients. For the first 10 years, it was using traditional AI methods, but generative AI has opened new services for the company. Users can query Prospection’s AI services about typical pathways for patients who took a certain drug, or what therapy they underwent after quitting the medication. A Prospection model can predict whether a patient will experience a certain event, such as needing to be hospitalized, over the next year.
“The ChatGPT transformer model is trained on billions of sentences consisting of words. We see each patient’s journey as the sentence and events in the journey as the words. That’s the vocabulary,” said Eric Chung, co-founder and co-CEO of Prospection. “The data is very powerful. There are lots of insights to be gained from data on 500 million patients. It’s beyond the power of humans to analyze, but AI can do it.”
Investing in Smart Agriculture
AI gets a lot of attention these days, but its application to farming is not often in the spotlight. Sagri Co., Ltd. uses AI, machine learning, and mapping technologies to solve social problems. I had the opportunity to talk with CEO Shunsuke Tsuboi about the challenges that agricultural technology startups in Japan face when it comes to funding, as well as the benefits of their technology.
Japan startup Sagri is transforming family farming with AI
Listen to this story:
Artificial intelligence (AI) gets a lot of attention these days, but its application to farming is not often in the spotlight.
Recently, I had the opportunity to talk with Shunsuke Tsuboi, chief executive officer of Sagri Co., Ltd., which uses AI, machine learning, and mapping technologies to solve social problems.
On my podcast, Disrupting Japan, Tsuboi and I discussed the challenges that agricultural technology (agtech) startups in Japan face when it comes to funding, as well as the benefits of their technology.
Taking Root
Sagri was founded in June 2018 and uses satellite imaging and data to analyze farmland. The technology scans areas of up to 10 hectares in size, making it particularly suited to Japan, where farms are generally small. The data can be accessed using a smartphone app, and the goal is to help farmers better understand the condition of their soil and identify the best time for harvesting.
The idea for the company took shape in a laboratory at Yokohama National University, where Tsuboi is a mechanical engineering graduate student. His lab is using space-based technology to examine soil, and he and his business partners have been able to apply some of this to their platform, which shares the name of the company.
Applications
If you’ve walked around the Japanese countryside, you’ve probably seen small plots of abandoned farmland. Sometimes these even intermingle with residences in neighborhoods not far outside the capital.
Whether farmland is in use or abandoned makes a difference from a tax perspective, so the government manually checks the status of land each year. The AI behind Sagri’s analysis can determine with 90-percent accuracy whether a field is abandoned, drastically reducing the amount of work required of government staff.
Apart from taxation, the government is also interested in identifying farmland that can be revitalized. Satellite data that provides soil analysis can make that process easier.
Tsuboi noted that a big reason for the abandonment is that the farmers are getting older and are unable to maintain the land. One benefit of the Sagri platform is that machines can receive the data analysis and automatically perform tasks such as applying fertilizer.
Beyond Japan
Agtech is an area in which Japan has a great opportunity to be a world leader, and Sagri is putting its technology to work in India, where there are also many small farms. But getting the financing needed to keep operations going can be difficult. Sagri believes it has a solution.
“Many Indian farmers need loans, but they don’t have the chance to get them,” Tsuboi said. Because there are so many farmers, it is difficult for banks to spread enough money around. To have a better chance of funding, farmers want to show banks that they are a good investment, he explained. Banks cannot get that sort of information using present methods, but the satellite data analysis provided by Sagri can allow them to check the farmland’s condition and potential yields.
Tsuboi sees Africa as the company’s next market, noting potential in countries such as Kenya and Rwanda. Areas of Southeast Asia are also within Sagri’s sights.
Funding
There are not many agtech startups in Japan, but it seems that there should be. With lots of small farms, lots of creative people working on agtech at universities, and venture capitalists (VCs) with money to invest, why don’t we see more?
Tsuboi feels one reason is that VCs and the government both see farmland as low-growth opportunities. And attracting money from abroad, such as from Silicon Valley VCs, is not easy because they are focused on large-scale industrial farming. The farms on which Sagri is focused in Japan and India are too small to attract their interest.
But Sagri has had some success inside Japan, and announced in June that they have secured ¥155 million ($1.4 million) in funding from a group led by Real Tech Holdings Co., Ltd., who was joined by Minato Capital Co., Ltd., Senshu Ikeda Capital Co., Ltd., and Hiroshima Venture Capital Co., Ltd. Also participating is Bonds Investment Group Co., Ltd., whose Hyogo Kobe Startup Fund, established in March, is making its first investment.
Sagri is a great example of a Japanese startup that can assist people at home and also have a much bigger impact—and earn a much bigger profit—abroad. Globally, the company can help millions of small family farms thrive, and they can bring great returns for investors in the process.