Who Will Be the Japan of the AI Era?

Drift Signal #249

The responses to the previous edition of Drift Signal (The Great Reset: From Startups to What's Next) highlighted a crucial question: as artificial intelligence reshapes production methods, which country or region will emerge as its pioneering force?

Some readers pointed to China's efficiency gains, others to India's emerging capabilities. Many emphasised Europe's institutional advantages, particularly in workforce management and social cohesion. This diversity of views reflects a deeper truth: the leader in AI-enabled production remains far from predetermined, much like Japan's rise in manufacturing wasn't obvious in the 1950s.

The parallel with Japan's lean production revolution proves particularly illuminating. Few predicted that a country still mastering mass production would pioneer a revolutionary approach to manufacturing that would transform not just how cars were built, but how entire value chains were organised. Yet through a combination of institutional innovation, long-term thinking, and strategic adaptation, Japan and its carmaking giants did exactly that.

Today, as we enter the Synergy stage of the Age of Computing and Networks, AI is emerging not as a new technology but as a revolutionary approach to organising production—much like lean production was for the Age of the Automobile and Mass Production. The question isn't who will develop the most advanced AI, but rather whose institutions are best suited to reimagine how we harness computing and networks across entire value chains.

This edition explores what made Japan's lean production revolution possible, why certain countries proved better at adopting it than others, and what this tells us about who might lead as AI transforms how we organise knowledge, work, and production itself.

1/ The Synergy Stage: Where We Are

As we enter 2025, we stand at an inflection point in what economist Carlota Perez calls a “Great Surge of Development”. The current Age of Computing and Networks (my label, not Carlota’s) has long passed through its Installation phase, characterised by the emergence and frenzied adoption of new technologies, and its subsequent collapse (bursting of the dot-com bubble). We are now somewhere in the middle, or perhaps nearing the end, of the first stage of the Deployment phase: Synergy.

The Synergy stage is when a technology becomes deeply embedded in society's fabric. When it comes to computing and networks, the evidence is all around us:

• Cloud computing and digital infrastructure are now as reliable as electricity

• Remote work and digital-first business models have become the norm rather than the exception

• Digital literacy is approaching the universality that reading and writing achieved in the 20th century

• The benefits of computing and networks are becoming democratised across society

• Business leaders and policymakers now have a deep understanding of network effects and platform dynamics

• Tech leaders have emerged as significant figures in global politics and governance

It is precisely at this plateau of core technology adoption that the most profound innovations emerge—not in the technology itself, but in how businesses organise around it.

As explained in the previous edition, this mirrors the 1970s during the Age of the Automobile and Mass Production, when Japan revolutionised manufacturing with lean production. As documented in the foundational book The Machine That Changed the World, lean production went beyond traditional mass production by emphasising flexibility, waste elimination, and continuous improvement, ultimately transforming not just manufacturing but management philosophy worldwide.

Today's parallel is striking, but in a different way than many assume. While computing and networks—our age's core technologies—are becoming mainstream, their newest iteration, artificial intelligence (AI), is emerging not as a new technology but as a (potentially) revolutionary approach to organising production around now-mature computing and networks. Just as lean production was not a new core technology but rather a profound reimagining of how to organise around standardised parts and assembly lines, AI represents a fundamental rethink of how to harness computing and networks.

---

2/ The Toyota Production System: Transcending False Trade-offs

The most profound innovations often arise from constraint and crisis. The Toyota Production System (TPS) emerged under such pressures, revealing a fundamental tension in business: the conflict between strategy and innovation. While strategy demands trade-offs and embracing constraints, innovation often transcends these very limitations.

In his seminal 1996 Harvard Business Review article What Is Strategy?, Michael Porter observed that “what were once believed to be real trade-offs—between defects and costs, for example—turned out to be illusions created by poor operational effectiveness.”

This happened back in the 1970s. When American carmakers were trapped in what seemed to be an unbreakable trade-off between quality and cost, Toyota proved this dilemma was artificial. The breakthrough wasn't mere improvement. It was a complete reimagining of production itself.

The power of TPS lay in its elegant simplicity. Built on two foundational principles—Just-in-Time delivery and Jidoka (automation with human intelligence)—it created a system where parts arrived only when needed, and production stopped instantly if quality issues emerged. This challenged mass production's core assumptions that prioritised economies of scale and continuous operation—never stopping the assembly line—over flexibility and quality.

What made TPS revolutionary was its holistic nature. It transformed everything from worker engagement to supplier relationships. While American manufacturers maintained adversarial relationships with suppliers, Toyota built long-term Keiretsu partnerships that fostered collaboration and shared improvement (more on those below). Workers at all levels contributed to Kaizen (continuous improvement), embedding a culture of constant innovation.

The results defied conventional wisdom: Toyota achieved higher quality, lower costs, and faster production times—benefits that mass production experts had long considered mutually exclusive. Most significantly, TPS demonstrated that the most transformative innovations often come not from new technologies, but from new ways of thinking about and organising around existing ones.

---

3/ Born from Crisis: The Origins of Lean Production

Japan's revolution in manufacturing emerged from two moments of crisis that forced fundamental innovation.

The first came in 1950, when Toyota faced near-bankruptcy and intense labour disputes. Unable to maintain its workforce, the company laid off a quarter of its employees, that is, about 1,600 people. The resulting conflict with Toyota's strong labour union could have destroyed the company. Instead, it became the catalyst for transformation.

The resolution set the stage for lean production. After Toyota's founder, Kiichiro Toyoda, resigned to take responsibility, a revolutionary compromise emerged: Toyota would guarantee lifetime employment for remaining workers, while workers would adopt flexible practices—taking on multiple roles and actively contributing to efficiency improvements.

This crisis forced Toyota to reimagine production. Unable to afford the large inventories required for American-style mass production, Toyota had to become more efficient with less. Engineer Taiichi Ohno led the transformation, drawing inspiration from unexpected sources—including American supermarkets, which stocked only what customers needed to maximise return on equity. The result was the Just-in-Time system, where workers were empowered to spot inefficiencies, suggest improvements, and even stop production when necessary.

The second major catalyst came in the early 1970s, when Japan's Ministry of International Trade and Industry (MITI) imposed strict energy efficiency regulations on automakers. This regulatory pressure forced Japanese manufacturers to prioritise fuel efficiency, well ahead of their American counterparts. When the oil crises hit only a few years later, Japanese companies had a critical advantage, producing fuel-efficient cars just as global consumers demanded them.

The resulting surge in Japanese imports transformed the global auto industry. By the early 1980s, much like these days, rising political tension in the US led to protectionist measures and ‘voluntary’ export restraints by the US's main trading partners, notably Japan. But these restrictions had an unexpected effect: they pushed Japanese manufacturers to build production facilities in America. Honda led the way in Ohio, followed by Toyota, Nissan, and others. These “transplant” factories proved that lean production was not just a Japanese cultural phenomenon—it was a universal approach to manufacturing.

---

4/ Europe’s Hidden Advantage: When Institutions Matter More Than Technology

As Japanese manufacturing methods spread globally through transplant factories, a surprising pattern emerged: European manufacturers adapted to lean production more effectively than their American counterparts. This was not due to superior technology or market size, but rather institutional compatibility.

This highlights a crucial but often overlooked truth about innovation: the success of a new production philosophy depends as much on existing institutional frameworks (and macroeconomic context) as on technical capability. While American firms struggled to implement Japanese methods, European manufacturers discovered unexpected advantages embedded in their institutional DNA.

The key lay in workforce management. Lean production thrived in environments that tolerated unionisation, a practice that was deeply ingrained in Europe but often resisted in the US. European carmakers' perceived weaknesses during the mass production era—such as smaller domestic markets, strong union relationships, and seniority-based career systems—became competitive strengths in the lean production era.

By the 2000s, this institutional advantage had profound consequences. European manufacturers—especially premium brands like BMW, Mercedes, and Audi—successfully integrated lean principles into high-margin, high-quality production. Meanwhile, US automakers continued to struggle with inefficiencies, quality issues, and declining market share, leading to a near collapse (and a controversial bailout) in the context of the global financial crisis in 2008.

This pattern of institutional advantage offers a crucial lesson for today. As AI emerges as a new production philosophy, success may depend less on technical sophistication and more on institutional readiness. From this perspective, the key question isn't who has the most advanced AI, but rather whose institutions are best suited to this new approach to production.

---

5/ AI as Production Philosophy: From Browser to Enterprise

Just as lean production wasn't a radical departure from delivering cars to consumers but a revolutionary way of organising their assembly, AI represents not a break from computing and networks but their deepening and reorganisation. This parallel runs deeper than many realise.

Consider that AI has been quietly embedded in our daily tools long before ChatGPT captured public imagination from December 2022 onward. Google's autocomplete function, which became a standard feature as early as in 2008, is a perfect example. Far from being just a convenience feature, it was a sophisticated system serving multiple purposes: helping users articulate their search queries more effectively while subtly guiding them toward phrasing that would yield better results and increase engagement with sponsored links—thereby generating revenue for Google. Similarly, DeepL's translation engine demonstrated how AI (specifically, advanced neural networks and a better training dataset) could surpass traditional rule-based approaches through pattern recognition and learning.

What ChatGPT represented wasn't a technological breakthrough but an interface revolution—making AI accessible through natural language conversation, all while operating an infrastructure able to sustain the resulting workload and to keep the user experience smooth. This mirrors how the web browser, particularly Mosaic and later Netscape, didn't invent the World Wide Web but made it usable for everyone through simple clicking, rather than typing obscure commands.

The precedent of the web browser offers a crucial lens for understanding AI's trajectory. In its early days, OpenAI was a closed system, with early models trained only on highly curated sets of data and unable to surf the Internet. This mirrors how CompuServe and AOL initially positioned themselves as gatekeepers of a carefully curated Internet experience. But then just as the web browser democratised access to the Internet, rendering those walled gardens obsolete, AI in its current form promises to democratise access to knowledge itself.

Building on this parallel (the chatbot as the new browser), the real potential of AI becomes clear when we consider its role in processing unstructured data. The web browser revolutionised access to structured data—information carefully formatted with HTML and other markup languages. AI chatbots (and, soon, AI agents) go further by making any knowledge accessible to humans, regardless of its structure. And while Web 2.0 enabled users to contribute content explicitly through browsers, AI users contribute implicitly through every interaction, continuously training and refining the models they use.

---

6/ AI-Powered Assembly Lines: Reimagining Knowledge Work

The implications of this shift become clear when we examine how AI transforms work within organisations. Just as Toyota revolutionised production by empowering workers on the assembly line, AI promises to revolutionise knowledge work by amplifying human capability.

Lean production rested on two key assumptions about workers: their superior process knowledge gained through experience, and their willingness to share that knowledge given the right incentives. The system worked because workers knew both the production process and the company's information systems intimately. A social contract of lifetime employment and guaranteed advancement ensured they would contribute rather than hoard their expertise.

Yet the pre-AI world imposed hard limits on knowledge sharing. When I worked at Siemens in Munich in 1998-1999, my entire job consisted of locating the right German engineer who could answer precise technical questions from French salespeople bidding on telecom equipment contracts. This manual knowledge brokering was necessary because of fundamental constraints: the capacity of individual memory, limited access to information systems, and the availability of key experts (not even mentioning the very high cultural and language barriers that separate the French from the Germans).

AI removes these constraints. Imagine equipping every authenticated worker with a chatbot capable of querying vast oceans of unstructured data—from technical documentation to email threads to meeting notes. Workers can not only access the entire company's knowledge base according to their clearance levels but also contribute their own insights back into it, creating a virtuous cycle of collective learning.

This reorganisation of knowledge work extends beyond individual enterprises to entire value chains. Just as lean production transformed relationships between manufacturers and suppliers, AI-powered knowledge systems could reshape how entire industries collaborate and innovate. The question isn't just how to implement AI technically, but how to reorganise our institutions to harness its full potential as a production philosophy.

---

7/ Process Knowledge: The Hidden Factor in Enterprise AI

One of the most important insights about enterprise and business comes from Dan Wang's seminal 2018 essay about (among other things) process knowledge and China's manufacturing prowess. Dan’s analysis of the semiconductor industry reveals something crucial about technological progress. As he explains:

Process knowledge is represented by an experienced workforce... The process knowledge can also be referred to as technical and industrial expertise; in the case of semiconductors, that includes knowledge of how to store wafers, how to enter a clean room, how much electric current should be used at different stages of the fab process, and countless other things. This kind of knowledge is won by experience. Anyone with detailed instructions but no experience actually fabricating chips is likely to make a mess.

Dan argues that “technology ultimately progresses because of people and the deepening of the process knowledge they possess.” Just as semiconductor manufacturing requires deep expertise that can't be captured in manuals alone, any company and any industry can only move forward if it finds a way to harness the process knowledge embedded in its experienced workforce. Here, AI represents a new deal.

However, there are two potential traps in using AI to leverage process knowledge:

• The first trap (Scenario #1) is assuming you don't need experienced workers anymore—that junior workers equipped with AI can deliver the same value without the company having to pay for experience. This mirrors how Uber drivers equipped with GPS could replace experienced taxi drivers who knew the city map by heart.

• The second trap (Scenario #2) is assuming you don't need junior workers anymore because AI can handle simple, repetitive tasks, while you retain only experienced workers who contribute critical process knowledge.

Today's discourse largely focuses on Scenario #1, with frequent predictions that AI will eliminate most senior jobs, leaving only entry-level positions supported by AI. This approach requires access to a cheap and abundant workforce, typically favouring countries with strong demographics or open immigration policies. The parallel with mass production is striking: division of labour, rise of the unskilled worker, high turnover, and a top-down, command-and-control approach—everything that Apple and Steve Jobs denounced in their legendary 1984 advertisement.

Yet Scenario #2 appears more likely, as argued in Steve Yegge's The Death of the Junior Developer. The obvious risk, however, is facing a drought of experienced workers if companies stop hiring and training junior staff. This creates a crucial challenge: how to maintain the virtuous cycle of knowledge accumulation that has historically driven technological progress.

---

8/ Winning in the AI Era: The Institutional Challenge

Let's distill what determines successful AI adoption across industries and companies. The pattern that emerges mirrors the lean production revolution of the 1970s, but with its own distinctive challenges.

The temptation to embrace Scenario #1—the mass production redux—will prove irresistible for many companies, trapping them in false trade-offs created by combining junior workers with powerful AI systems. Steve Yegge's experience as a senior developer illustrates this perfectly. As he told in the aforementioned article, when experimenting with AI to improve his code, the system proposed rebuilding everything from scratch with what appeared to be an elegant solution. Any junior developer might have immediately implemented this AI-generated code, celebrated the quick win, and headed to the beach. But Steve, with his deep process knowledge, burst out laughing when he realised the critical flaws in the AI's approach—flaws that could only be obvious to an experienced developer like him.

This presents a stark choice: whether it’s software development or any other trade, do you want a cheap junior worker-AI combination that produces unstable systems requiring constant rebooting? Or do you follow Scenario #2, where experienced workers use AI for routine tasks while applying their process knowledge to ensure robust, elegant solutions?

Just as with lean production, Scenario #2 promises better quality, lower costs, and faster delivery. But it requires specific institutional conditions to initiate and to succeed:

• First, it thrives in contexts of demographic pressure, where young workers are scarce. This describes much of Europe (particularly given increasing resistance to immigration), China, and Japan. The US currently stands apart, though this could change if immigration restrictions create a talent shortage that forces a shift from Scenario #1 to Scenario #2.

• Second, it demands a solution to what we might call the process knowledge paradox: if companies rely solely on senior workers augmented by AI, they break the pipeline that creates those senior workers in the first place. This requires a cultural and institutional framework where companies commit to training junior staff even when AI could handle their current work more efficiently. It means investing in decades of training before workers accumulate enough process knowledge to fully leverage the company's AI systems. Few countries have institutions capable of supporting such long-term investment—Germany, Nordic countries, and perhaps Japan stand out.

• Finally, it requires incentive systems borrowed directly from lean production: lifetime employment, guaranteed advancement, and continuous learning opportunities that keep workers engaged without the traditional climb up the corporate ladder. These systems must prevent valuable process knowledge from walking out the door to competitors. Japan pioneered this approach, and Europe adapted it successfully. The US, however, faces the same institutional barriers that hindered its adoption of lean production in the 1970s.

This assessment suggests that the nations best positioned to lead in the AI era may be those whose institutions already support long-term investment in human capital and process knowledge. The parallel with lean production's adoption pattern becomes increasingly clear: in any industry, success depends not on technical sophistication alone, but on the ability to build and maintain institutions that nurture and retain process knowledge on the factory floor.

---

9/ Beyond the Enterprise: AI and the Value Chain

There's one final parallel between lean production and AI that we mustn't overlook: neither can succeed if confined within a single enterprise. Just as lean production required the creation of an entire ecosystem—the keiretsu system—AI's potential can only be fully realised through deep integration across value chains.

The keiretsu weren't merely a Japanese curiosity. Nike's early history provides a striking example of how this model could benefit Western companies. As told in the landmark episode of the podcast Acquired about Nike, in 1971, when Phil Knight discovered Japanese trading company Nissho Iwai (now Sojitz), he found not just a financial partner but a strategic ally. Beyond providing crucial inventory financing, Nissho Iwai helped Nike establish manufacturing relationships in Japan, fundamentally transforming the company's trajectory. In exchange for this support and industry knowledge, Nissho Iwai received a 4% royalty on Nike's sales—a partnership that would last decades.

Western companies developed their own versions of value chain integration. As told in Charles Fishman’s excellent The Wal-Mart Effect (published in 2006), Walmart's Retail Link system pioneered a new model of data sharing with suppliers. Rather than maintaining rigid boundaries between retailer and suppliers, Walmart gave its partners access to real-time sales and inventory data. This transformed suppliers from mere vendors into strategic partners, responsible for analysing trends, optimising production, and proposing innovations to boost sales.

As AI emerges as the lean production of our time, we're likely to see a similar transformation in how companies collaborate. Large enterprises in every industry will need to align incentives with smaller, specialised players across their value chains—particularly in technology and AI innovation. This suggests a profound shift in how technology startups will evolve: rather than pursuing the traditional venture capital path of building independent companies, many will thrive by forming deep, keiretsu-style partnerships with established enterprises.

This marks a potential end to the era dominated by VC-backed startups aiming for independent growth or acquisition. Instead, we're likely to see the emergence of a modern, AI-driven twist on the keiretsu system: startups providing specialised AI capabilities while receiving not just funding but also data, domain expertise, and market access from their enterprise partners. Just as Japanese parts makers became integral to the success of Toyota's production system, AI startups will become essential participants in the knowledge systems of tomorrow's leading companies.

The countries and regions that succeed in the AI era will be those that facilitate these new forms of partnership—creating institutional frameworks that support long-term collaboration while protecting the interests of both large enterprises and their smaller partners. In this light, the question of who will be the Japan of the AI era becomes clearer: it will be whoever best enables this reimagining of the value chain for the age of artificial intelligence, and makes it sustainable.

---

10/ Further Reading

• What Is Strategy? (Michael Porter, Harvard Business Review, November 1996)

• Technological revolutions and techno-economic paradigms (Carlota Perez, Working Papers in Technology Governance and Economic Dynamics no. 20, January 2009)

• Disruptive Innovation (video—Clayton Christensen, Oxford University, June 2013)

• How Technology Grows (a restatement of definite optimism) (Dan Wang, July 2018)

• Who’s Disrupting Whom at the Global Level? (me, Drift Signal, June 2020)

• Nike: The Complete History and Strategy (audio and transcript—Acquired, June 2023)

• The Death of the Junior Developer (Steve Yegge, Sourcegraph Blog, June 2024)

• The Death of the Stubborn Developer (Steve Yegge, Medium, December 2024)

• DeepSeek hints that China has mastered the art of ‘kaizen’ — the west should be worried (Leo Lewis, The Financial Times, January 2025)

• The End of Programming as We Know It (Tim O’Reilly, O’Reilly, February 2025)

---

Sign up to Drift Signal if you don’t want to miss the next issues 🤗

---

From Paris, France 🇫🇷

Nicolas