AI and the New Economic Order (1/3)

June 2025 

Let’s be clear: artificial intelligence is no longer operating at the margins, but has become the core engine of the modern economy. Over the past two decades working with high-tech and high-growth companies, I’ve witnessed waves of technological disruption: the rise of MEMS, the explosion of smartphones, the scaling of the cloud, the emergence of SaaS, and the proliferation of IoT. Each wave redefined an era. But what we are now witnessing with generative AI, powered by foundation models and autonomous systems, may eclipse them all in both scale and consequence.

This shift isn’t about smarter chatbots or sharper product recommendations. It is about a fundamental rewiring of how value is created, scaled, and captured across the economy. AI is no longer a tool confined to IT departments or academic experiments; it is embedding itself deep within the operating core of businesses, redefining how firms organize, compete, and grow. Much like electricity reorganized the industrial economy, foundation models, general-purpose, multi-modal systems like GPTs, are quietly powering a transformation that reaches across sectors and siloes.

The implications are not cyclical; they are structural. These models represent a break from the legacy of task-specific automation. They are flexible, adaptive, and capable of handling language, code, images, and more. But more than just enhancing productivity, they substitute for human effort. They collapse workflows, compress decision timelines, and automate the sort of knowledge work that once sat safely atop the value chain. This time, the target is not the factory floor, it’s the boardroom, the design studio, the analyst’s desk.

So why is this moment different?

It’s not one factor, but a confluence. Scale has made the difference. We’ve hit a technological inflection point, where internet-scale data meets Moore’s Law-fueled compute, enabled by breakthroughs in neural architecture. The result is systems that don’t just promise general intelligence, they begin to deliver it. Unlike the brittle, rule-based expert systems of the past, today’s models are robust, adaptive, and multi-domain. They can generate text and code, interpret images and speech, simulate reasoning, and even self-improve.

What’s equally important is the inversion of the adoption curve. Historically, new technologies flowed downward: from governments, to corporations, and eventually to consumers. With AI, the pattern has reversed. Adoption began at the edges, by consumers, prosumers, and small startups, and is now scaling inward, upending traditional diffusion models. Tools like ChatGPT and Midjourney were not enterprise-first, they became mass-market phenomena almost overnight, driving demand from the bottom up. This consumer-led adoption accelerates refinement, fuels rapid iteration, and undercuts the power of legacy gatekeepers and regulatory inertia.

While the early breakthroughs in AI were driven by a handful of dominant players—OpenAI, DeepMind, Meta, Microsoft, today, the proliferation of open-source models like Mistral, LLaMA, and Phi-3, along with tools for local inference such as Ollama, has enabled individuals and small teams to fine-tune and deploy powerful AI systems at low cost. Access is becoming more democratised. However, the underlying infrastructure, compute, data centres, and model training remains largely centralised.

This shift is not just technological; it is deeply cultural and economic. Culturally, AI is transforming how people create, communicate, and interact, reshaping media, art, music, writing, and even companionship. Economically, it is redefining productivity: one person equipped with AI can now accomplish what once required entire teams. A new form of social capital is emerging, where fluency in AI tools confers status and leverage.

The disloaction is slowly becoming evident. The new economic order will not be determined solely by who owns the most compute. While control over infrastructure still matters at the frontier, value creation is moving to the application layer, how AI is used across sectors like healthcare, finance, education, and manufacturing. Individuals and small businesses are increasingly capturing outsize returns by building on top of foundation models through agents, plugins, and vertical applications.

Underlying all of this is a profound shift in how we think about intelligence. Once a scarce, human-bound resource, intelligence is becoming ambient embedded across systems, and agentic, as autonomous tools begin to plan, reason, and act independently. AI is not merely augmenting cognition; it is decentralising and distributing it across the economic landscape.

Productivity Without People?

The numbers are compelling. Estimates suggest that generative AI could add $2.6 trillion to $4.4 trillion in annual economic value across 63 high-impact use cases, figures that could double if integrated across existing enterprise platforms. Yet this value is highly concentrated. Around three-quarters of it sits within just four domains: customer service, marketing, software development, and research. These functions, particularly in data-rich sectors such as banking, life sciences, and high tech, are proving especially ripe for transformation.

The effect is not simply one of enhancement, but replacement. Traditional organisational logic, more people equals more output, is being upended. Productivity per capita is surging. Generative AI enables a shrinking workforce to deliver greater output, not only through automation of routine tasks but by penetrating deep into the cognitive core of white-collar professions. Legal briefs, software code, market analyses, even pharmaceutical R&D, all are now being accelerated, if not outright performed, by machines.

This shift is already being priced into executive expectations. Over 75% of CEOs anticipate major AI-led business transformation within three years. Yet while the productivity dividend is attractive, the labour implications are unsettling. McKinsey projects that 60–70% of current work activities could be automated. Crucially, this transformation does not spare the elite. It is not factory-floor jobs under siege, but marketing associates, legal analysts, paralegals, and junior software engineers.

The Economic Paradox: Growth Amid Displacement

The paradox is stark: while AI promises accelerated productivity growth, potentially adding up to 3.4 percentage points annually when combined with complementary technologies, it also threatens to hollow out employment structures faster than new roles can be created. Without effective reskilling strategies, the result could be a surge in structural unemployment, compounded by social unrest and political backlash.

The timeline is tightening. Earlier estimates placed peak job displacement in the 2050s. That horizon has now moved forward by a decade. By 2045, as much as half of today’s work could be automated. The coming years will test the adaptability not just of workers, but of governments, institutions, and corporate leadership.

Capital’s New Frontier

For companies and their shareholders, AI presents not just an operational tool but a financial lever. AI enhances cash flows through increased revenue (via personalization, targeting, and customer analytics) and reduced costs (via process automation and predictive logistics). It accelerates those cash flows by streamlining approval cycles, invoicing, and product development. It reduces volatility by improving forecast accuracy and customer retention. Most strategically, it builds residual value through proprietary models and datasets that confer long-term defensibility, what venture capitalists term the “intelligence moat.”

Today, the spread of this value is uneven. Only 26% of companies have moved beyond pilot programs. A mere 4% are true AI leaders, firms that have embedded AI across the enterprise and aligned their strategic and financial planning accordingly. Studies show that these leaders are supposed to capture the lion’s share of the value: 50% higher revenue growth, 60% better total shareholder returns, and 40% superior return on invested capital. What distinguishes them is not their enthusiasm, but their execution: AI is embedded in their core operations, not relegated to experimental “labs” or back-office tinkering.

Power, Policy, and the Future of Work

The rise of AI as a general-purpose technology brings with it broader policy challenges. The historical compact between labour and capital, already fraying, may be tested anew. If AI continues to depress the marginal cost of labour while boosting capital returns, inequality could deepen and political polarisation intensify. Governments will face a delicate balancing act, whether to tax, regulate, subsidise, or simply adapt. The early contours of this debate can already be seen in proposals for robot taxes, AI governance frameworks, and universal basic income experiments.

But in the final reckoning, AI is not just a technological upgrade. It is a structural force, one that reshapes not only how economies grow, but who participates in that growth. Like steam, electricity, and the internet before it, AI offers the possibility of abundance. Whether it delivers inclusion is a matter not of algorithms, but of choices.

AI and the Strains of Economic Transition

The rapid advance of artificial intelligence is driving a profound dislocation in the global economy: while AI systems are dramatically improving productivity, the pace at which economies generate new, meaningful work has not kept up. Breakthroughs in AI, such as image recognition error rates falling from 30% in 2010 to below 2.2% by 2017, have enabled machines to outperform humans in tasks ranging from visual perception to natural language processing. 

Yet, paradoxically, U.S. labor productivity growth slowed to just 1.3% annually between 2005 and 2016, down from 2.8% in the prior decade, with gains largely confined to a few tech-centric sectors (more about it later). Simultaneously, automation threatens widespread job displacement: McKinsey estimates that up to 800 million workers could be affected globally by 2030, while studies by Acemoglu and Restrepo suggest AI adoption reduces both employment and wages in directly impacted sectors, with limited reabsorption elsewhere. AI models like GPT-4 and IBM Watson now perform complex cognitive and service tasks, including diagnostics, coding, translation, and customer service, once thought to be uniquely human. In the U.S. alone, over 2.2 million call center jobs face risk, with AI capable of handling 60–70% of interactions. 

However, the benefits of this technological revolution are accruing disproportionately to a small number of "superstar" firms, whose capital efficiency means fewer jobs per dollar of output. The resulting wage polarisation, stagnating median incomes, and regional employment gaps are fueling political instability, as displaced workers gravitate toward populist movements in areas with little economic diversification. Together, these trends reveal a structural asymmetry: AI’s ability to displace labour is accelerating faster than our capacity to reimagine and rebuild pathways for inclusive economic participation. 

Despite dystopian forecasts, AI is unlikely to eliminate work entirely. Instead, it will reconfigure it. Most roles will undergo partial automation, routine tasks will be offloaded to machines, while humans take on responsibilities that require oversight, integration, and adaptability. The labor market will demand more from fewer people. What once required five employees may soon require only two, and previously distinct roles may collapse into a handful of streamlined functions. The question, then, is what becomes of the displaced? This compression will be felt most acutely in highly automated economies, where the pace of technological adoption outstrips the creation of new roles. As demand for routine cognitive skills declines, workers who cannot pivot toward roles requiring social, emotional, or advanced technical fluency risk permanent exclusion from the labor market.

This disequilibrium is not self-correcting. Organizations may embrace automation to boost margins, but absent deliberate reskilling, they are contributing to a widening chasm between capital and labor. Moreover, even workers willing to adapt often find themselves navigating unclear pathways. Educational systems remain lagging, public policy tools insufficiently targeted, and corporate training programs reactive rather than strategic. The result is a mismatch between displaced capabilities and future demand, further exacerbating wage suppression, job insecurity, and political volatility.

The Rise and Strain of Shareholder Primacy

One reason this challenge has proven so difficult to address is the prevailing model of corporate governance. In Anglo-American capitalism, the principle of shareholder primacy still dominates. AI, by increasing scale without proportionally increasing labor, reinforces this logic. Cost savings and efficiency gains accrue directly to owners of equity, while displaced labor bears the brunt of adaptation. From a shareholder perspective, this is rational. From a societal standpoint, it may prove unsustainable.

Worse, AI further entrenches winner-take-most dynamics. Companies with access to the best models, largest datasets, and proprietary infrastructures enjoy compounding advantages, erecting high barriers to entry and creating de facto monopolies. This reinforces concentration not only of market power, but of political influence. Shareholder primacy in an AI-transformed economy risks morphing into algorithmic rentierism, a form of value capture increasingly decoupled from productive employment.

Calls for reform are growing louder. Some advocate for a broader stakeholder model, where workers, consumers, and communities have formal standing in governance decisions. Others push for structural reforms: antitrust enforcement, algorithmic transparency mandates, or data commons. The core question is whether companies will continue to treat AI as a margin-maximizing tool for shareholder gain, or as a general-purpose technology capable of diffusing benefits across society.

Deflation and the End of the Labor-Centric Economy

Over time, the impact of artificial intelligence will stretch well beyond labor displacement and job redesign. At a deeper level, AI is transforming the price structure of the global economy itself. As digital production scales and the marginal cost of many goods and services falls toward zero, entire categories once shielded from tradability—such as healthcare, education, and professional services are becoming subject to persistent downward price pressure.

McKinsey Global Institute (2023) notes that generative AI tools are significantly reducing the marginal cost of producing code, content, and customer interactions by enabling reusability and automation. GitHub Copilot case studies show productivity boosts of 55–120% for software developers, effectively lowering the labor cost per unit of code. This trend is emblematic of a broader shift: intelligence and cognitive labor, once expensive and scarce—are becoming abundant, replicable, and virtually costless.

According to Goldman Sachs (2023), generative AI could automate 25–50% of tasks across 300 million jobs globally, particularly in legal, education, marketing, and financial services. These are high-margin, high-skill sectors traditionally insulated from technological disruption, now facing accelerating cost compression. In healthcare and education, AI-powered tools like Khanmigo, Scribe, and Gradescope are lowering the per-student cost of instruction, while radiology and diagnostic platforms such as Aidoc and Zebra are reducing scan interpretation times by 30–50%. PwC (2022) estimates that AI-driven automation could cut provider-side healthcare costs by 15–20% in developed countries.

Simultaneously, the rise of open-source AI ecosystems—such as Hugging Face, Mistral, and LLaMA—has made powerful models freely available for local deployment, unleashing a wave of zero-cost agents and autonomous tools. AutoGPT-style agents can now perform end-to-end workflows that previously required entire teams of salaried analysts, assistants, or creatives.

This shift does not merely drive efficiency—it breeds structural deflation. Institutions like the OECD and the Bank of England have highlighted how automation and digital services may suppress prices as productivity growth outpaces demand. Ark Invest forecasts that AI will induce “technologically induced deflation” across knowledge-intensive sectors, with potential cost reductions of 15–20% over the coming decade.

The mechanisms are already visible. In manufacturing, AI-powered robotics and 3D printing are closing the gap between custom and mass production, collapsing cost differentials. In healthcare, AI accelerates diagnostics, streamlines imaging, and compresses drug discovery timelines, lowering costs while expanding access. In media and creative industries, generative AI produces text, video, and design outputs that once required teams of skilled professionals, now delivered in minutes at negligible cost. Even energy markets are being reshaped by AI-optimized grid systems and predictive analytics that cut waste and reduce utility costs.

On the surface, this abundance seems to herald a consumer utopia. Goods and services become cheaper, faster, and increasingly tailored to individual needs. In theory, AI could usher in a post-scarcity era across multiple industries, unlocking productivity gains on a scale not seen since the Industrial Revolution. But beneath the surface lies a more troubling macroeconomic reality.

Deflation, even when born of abundance, destabilizes traditional economic assumptions. As prices fall, wages flatten. Consumers delay purchases in anticipation of lower future costs, suppressing aggregate demand. Investment slows, uncertain of future returns. Meanwhile, debts, whether held by households, firms, or governments, become heavier in real terms. Central banks, long reliant on interest rate manipulation to manage demand and inflation, find themselves increasingly impotent. Their instruments were designed for economies governed by scarcity, not surplus.

A striking historical example is Japan’s “Lost Decades” (1991–2020). After the collapse of its 1980s asset bubble, Japan entered a prolonged period of low growth, stagnant wages, and persistent deflation—despite continued technological progress and global competitiveness. Consumer prices flatlined, household savings rates soared, and consumption was delayed in anticipation of further price declines. In response, the Bank of Japan deployed increasingly unconventional tools: zero and negative interest rates, quantitative easing, and fiscal stimulus. Yet these interventions had limited effect. Demand stagnation became self-reinforcing, corporate investment remained subdued despite abundant capital, and public debt ballooned—its real burden magnified by deflation. Japan’s experience underscores how even productivity-driven abundance, if unchecked by demand-side dynamics, can trap an economy in a cycle of inertial deflation that conventional monetary policy cannot easily reverse.

Perhaps most consequentially, inequality becomes more difficult to address through growth. When fewer workers are needed to generate more output, the historic link between employment and consumption begins to break. The postwar social compact, growth drives jobs, jobs drive income, income drives demand, starts to unravel. In its place, a new equilibrium emerges: one where economic participation must be redefined not by labor input, but by access to the dividends of automation ?