The Great AI Misunderstanding: It Is Not Another App, It Is a General-Purpose Technology
by Denis Huré
Most executives are asking the wrong question about AI.
They ask: Which model should we use? Which copilot should we deploy? Will it save headcount? Is ChatGPT the next electricity?
The more important question is this: are we treating AI as software when it should be treated as infrastructure?
That distinction will separate companies that achieve marginal automation gains from those that redesign their economics, operating model, workforce and competitive position. History suggests that the greatest value from artificial intelligence will not go to the organisation that installs the most AI tools. It will go to the organisation that understands what AI actually is: a potential General-Purpose Technology.
There is, however, one immediate source of confusion to remove. “GPT” can mean two very different things:
Generative Pre-trained Transformer: the technical architecture behind many large language models, including OpenAI’s GPT family.
General-Purpose Technology: an economic category for technologies that transform multiple sectors over decades, such as steam power, electricity, the internal combustion engine, computing and the internet.
The former is a type of AI model. The latter is a claim about AI’s place in economic history.
That distinction matters because no board should mistake a popular chatbot for a transformation strategy.
The New Reveal: AI’s Biggest Effect Will Not Be Where Most Firms Are Looking
The common assumption is that AI’s value lies primarily in faster content creation, better customer service, coding assistance and lower administrative cost.
Those are real benefits. But they are not the central story.
The new reveal is this: AI may matter less as a productivity tool than as a new layer of economic coordination.
Steam power did not merely make factories run faster. It changed where factories could be built, how production was organised, how cities grew and how global trade operated. Electricity did not simply replace steam engines with motors. It enabled entirely new factory layouts, night-time commerce, domestic appliances, refrigeration, telecommunications and modern urban life.
Likewise, computing did not simply speed up accounting. It created software industries, global supply chains, digital finance, e-commerce, data-driven marketing and platform business models. The internet did not merely improve communication; it rewired distribution, media, travel, retail, customer acquisition, employment and geopolitics.
AI could play a similar role. It is increasingly embedded across knowledge work, software development, scientific discovery, design, customer interaction, risk management, logistics, healthcare, fraud prevention and decision support. The OECD has concluded that generative AI has considerable potential to qualify as a new General-Purpose Technology because it is improving quickly, finding applications across sectors and triggering complementary innovation.
The implication for CEOs is uncomfortable: the value of AI will not be captured by deploying isolated tools. It will be captured by redesigning the systems around them.
Contrast Framing: Stop Comparing AI to SaaS
Most firms are currently managing AI as if it were another software procurement cycle.
They compare vendors. They run pilots. They procure licences. They ask IT to manage access. They measure usage. They celebrate a reduction in the time required to write an email, prepare a presentation or summarise a report.
That is the old model.
The historical evidence points to a different model: General-Purpose Technologies create value only when organisations change the complementary systems around the technology.
Electricity is the classic example. Early factories often replaced a central steam engine with a large electric motor but kept the same belt-and-shaft factory design. The result was limited benefit. The real productivity gains came later, once manufacturers redesigned facilities around smaller electric motors, flexible production lines and new methods of work organisation. The productivity payoff from electrification took roughly two to three decades to become visible in aggregate economic data.
Computing followed the same pattern. During the 1980s, economist Robert Solow famously observed that computers could be seen everywhere except in productivity statistics. The apparent paradox was not that computers were useless. It was that hardware investments alone did not create transformation. Organisations required new business processes, software, data structures, management practices and workforce capabilities before the gains became visible.
AI faces the same test.
The contrast is fundamental. A company that gives every employee a chatbot subscription may gain efficiency. A company that redesigns underwriting, product development, loyalty, sales operations, customer experience and governance around AI may change its relative cost base and strategic position.
One is optimisation. The other is transformation.
Why This Matters Now
There is a narrow window in which AI advantage can still be designed rather than merely defended.
The novelty is not that artificial intelligence exists. AI has been researched and deployed for decades. What changed is that generative AI has made powerful capabilities accessible through natural language interfaces, APIs and increasingly autonomous workflows. A technology that once required specialist teams and significant technical infrastructure can now be used by employees, customers and partners with limited training.
That changes the speed of diffusion.
Steam power took decades to reshape economies. Electricity took much of the late nineteenth and early twentieth centuries. Computing and the internet transformed work over several decades. Generative AI reached 100 million users at unprecedented speed after ChatGPT’s launch, although user adoption is not the same thing as deep economic transformation.
The urgency is therefore not that every organisation must rush to automate indiscriminately. The urgency is that AI is becoming a new competitive baseline faster than previous General-Purpose Technologies did.
When every competitor can generate acceptable marketing copy, basic software code, customer responses and research summaries, those capabilities stop differentiating the business. Differentiation shifts to:
Proprietary data and the right to use it.
Superior operating processes.
Human judgment at critical decision points.
Trust, transparency and brand credibility.
Regulatory readiness and risk management.
The ability to combine AI with real customer relationships, distribution and intellectual property.
The speed at which the organisation learns and redesigns.
In other words, AI can reduce the value of generic knowledge while increasing the value of differentiated assets.
For a loyalty business, for example, generic AI can help produce campaign copy, customer-service responses and broad segmentation. But the real strategic advantage lies in proprietary behavioural data, reward economics, merchant relationships, local regulatory knowledge, fraud controls and the ability to deliver relevant value to customers at scale. AI becomes an amplifier of the underlying business system; it is not a substitute for that system.
Bullseye Proof: Every Previous General-Purpose Technology Created a Lag Before a Leap
The evidence from prior technological revolutions is not that every innovation produces immediate growth. It is that transformational technologies frequently create an initial period of disruption, disappointment and uneven benefits before producing broad economic gains.
The most important pattern is the productivity J-curve.
First, organisations invest. Then costs rise: systems need integration, processes need redesign, staff need training, controls need to be established and legacy technology needs to be reworked. During this phase, productivity may not improve visibly.
Then, once complementary capabilities mature, the technology begins to change the organisation’s production function.
This is why executive impatience is dangerous. A board that expects AI to produce enterprise-wide financial impact within one quarter may force the organisation into shallow automation projects that do not build strategic capability. A board that tolerates endless experimentation with no commercial discipline will waste capital and create uncontrolled risk.
The right posture is neither hype nor paralysis. It is a staged portfolio: near-term productivity cases, medium-term process redesign and long-term business-model options.
Did Previous Technologies Create the Same Fears?
Yes, but the historical lesson is not that today’s fears are irrational.
The Luddites, active in England between 1811 and 1816, were not simply anti-technology. Many were skilled textile workers whose livelihoods, bargaining power and working conditions were threatened by mechanised production. Their concern was not merely that machines existed; it was that the gains from machines would be captured by owners while workers bore the cost of displacement.
That concern remains relevant.
Technology has not historically created permanent mass unemployment across entire economies. But it has repeatedly eliminated specific occupations, weakened some forms of worker bargaining power, shifted income toward capital or high-skill labour, and created sharp geographic and sectoral inequalities.
The computer revolution is instructive. Research has found that computerisation did not generate major aggregate job losses, but it did contribute to substantial labour-market disruption: lower-wage and routine occupations tended to lose employment, while higher-wage, computer-intensive occupations gained.
The internet delivered similar outcomes. It undermined travel agents, newspaper classifieds, traditional retail intermediaries, video-rental stores, telephone operators and many other established roles. At the same time, it generated new work in software engineering, cybersecurity, digital marketing, cloud infrastructure, online logistics, user-experience design and e-commerce operations.
The correct historical conclusion is not: “technology always creates more jobs, so there is nothing to worry about.”
It is: technology can create long-term prosperity while causing real short-term harm to specific workers, communities and business models.
That distinction is central to responsible AI strategy.
Jobs: The Question Is Not Whether AI Replaces Work, But Which Work
AI is likely to automate some tasks, augment many others and create entirely new roles. It is less useful to ask whether “jobs” disappear in the abstract than to ask which activities inside each role become automated, accelerated, monitored, redesigned or newly valuable.
For example:
The World Economic Forum’s Future of Jobs Report 2025, based on responses from more than 1,000 employers representing more than 14 million workers, projects that 170 million jobs could be created globally by 2030 while 92 million roles are displaced, for a projected net gain of 78 million jobs.
Those numbers should not be treated as a forecast with false precision. They are employer expectations, not a settled economic outcome. But they reinforce the historical pattern: AI is likely to reallocate work at scale rather than simply eliminate it.
The real risk is transition capacity.
If AI adoption moves faster than organisations can retrain workers, redesign roles, revise education systems and create new demand, the short-term disruption may be much harsher than it was during slower historical transitions. The World Economic Forum explicitly identifies alternative futures, including an “Age of Displacement” in which automation outpaces adaptation and labour-market disruption intensifies.
GDP, Productivity and Human Development
Previous General-Purpose Technologies eventually contributed significantly to economic output, but their effects were indirect, delayed and difficult to isolate.
The internet’s contribution to the United States business sector was estimated by the OECD at between 3.2% and 13.8% of value added in 2011, depending on how broadly the internet economy was defined. The US Bureau of Economic Analysis found that computers’ direct contribution to real GDP growth increased from approximately 0.1–0.2 percentage points annually in the late 1980s and early 1990s to around 0.3–0.4 points in the late 1990s.
Yet GDP is not the whole story.
Electricity, transportation, digital connectivity and education have been strongly associated with improvements in living standards, access to services and human development. Research examining electricity use and the UN Human Development Index has found a positive relationship, but also diminishing returns: once a country reaches a foundational level of energy access and economic development, further improvements in HDI depend increasingly on education, health, institutions and income distribution rather than energy consumption alone.
AI will likely follow a similar pattern.
Its most socially valuable effects may not appear first in GDP figures. They may appear in earlier disease detection, improved access to education, more accessible public services, reduced language barriers, safer industrial processes, faster scientific research and more personalised customer support.
But those benefits are not automatic.
AI can also worsen misinformation, surveillance, discriminatory decision-making, copyright disputes, labour-market inequality and concentration of market power. The quality of outcomes will depend on the rules, data, incentives and governance surrounding the technology.
The Optimistic Case
The optimistic view is based on a powerful historical pattern: no prior General-Purpose Technology has resulted in permanent economy-wide unemployment. Over time, steam, electricity, computing and the internet increased output, created industries, expanded consumer choice and raised average living standards.
AI may be especially powerful because it can improve not only production and communication, but also the process of invention itself. It can accelerate software development, drug discovery, materials science, customer insight, document processing and product design. This creates the possibility of faster innovation cycles than those enabled by earlier technologies.
Under this scenario, AI becomes a force multiplier for human capability. Workers become more productive, smaller firms gain access to sophisticated capabilities, and previously scarce expertise becomes more available.
This is the “Supercharged Progress” scenario: broad productivity gains, new business formation, improved services and higher-value work.
The Case for Caution
The cautious view is equally grounded in evidence.
Economist Robert Gordon and other sceptics argue that AI’s macroeconomic effect may be smaller than the effect of electricity, sanitation, automobiles or modern medicine. They note that many AI benefits may remain concentrated among a small number of technology companies and highly productive firms, while the wider economy sees incremental rather than transformative gains.
There is also the distribution problem. Even if AI raises GDP, it does not follow that it raises wages, job security or wellbeing for everyone. An AI system that enables a firm to serve more customers with fewer entry-level employees may improve margins while reducing the pathways through which people develop expertise.
There is a governance problem too. Unlike steam power or electricity, AI makes decisions and generates content using data that can be biased, confidential, copyrighted, incomplete or manipulated. It can be deployed at machine speed, across jurisdictions, in customer-facing environments and within high-stakes decisions.
For European organisations, this makes AI not only a technology and workforce issue, but a regulatory, data protection and intellectual-property issue. The EU AI Act, GDPR, cross-border data-transfer rules, copyright and training-data questions, and sector-specific obligations must all be considered as part of the operating model, not after technical deployment.
The cautious view is therefore not anti-AI. It is anti-naivety.
What CEOs Should Do Now
The lesson from prior General-Purpose Technologies is clear: do not ask whether AI will transform the economy. Ask whether your organisation is building the complementary assets required to benefit when it does.
Five priorities follow.
Treat AI as enterprise architecture, not a collection of tools.
Create a roadmap that connects technology choices to operating model, customer experience, workforce strategy, data governance, intellectual property and risk appetite.Target workflows, not demonstrations.
A successful pilot is not a strategy. Prioritise end-to-end processes where AI can improve speed, quality, revenue, cost, risk or customer outcomes with measurable accountability.Invest in the complements.
Historical productivity gains followed investment in skills, process redesign, data quality, integration and managerial capability. AI will be no different.Design a workforce transition, not simply a headcount programme.
Analyse task exposure role by role. Build reskilling, redeployment and human-oversight pathways before automation decisions create avoidable capability gaps.Build trust as a competitive asset.
Establish clear standards for data use, human accountability, model selection, vendor due diligence, security, bias testing, intellectual-property protection and regulatory compliance.
The central lesson is this: AI’s future is not pre-written by the technology. It will be determined by the decisions leaders make about how it is deployed, governed and shared.
How TLA&C Can Help
TLA&C works with CEOs, boards and senior leadership teams at the point where AI strategy, commercial opportunity, regulation and brand risk overlap.
The firm helps organisations move from experimentation to a practical, defensible AI operating model: identifying high-value use cases, designing governance and risk frameworks, assessing build-versus-buy choices, preparing for EU AI Act and GDPR obligations, protecting intellectual property, and redesigning workforce and customer processes around measurable commercial outcomes.
The aim is not to adopt AI fastest. It is to adopt it with the strategic clarity, regulatory cover and governance architecture required to create sustainable advantage. TLA&C combines strategy and operations advisory, regulatory and data-compliance expertise, and brand and IP protection to address the full executive agenda created by AI.
Bibliography
Stanford University — “Productivity Paradox: Lagging Investments”
US Bureau of Economic Analysis — “GDP and the Digital Economy”
World Economic Forum — “Future of Jobs Report 2025: The jobs of the future”
CEPR VoxEU — “How computer automation affects occupations: Technology, jobs and skills”
UNDP Human Development Reports — “Energy and Human Well-Being”
TL Agency & Consultancy — AI strategy, compliance, IP and brand protection services
About Denis Huré
Denis Huré is the founder & Managing Consultant of TLA&C. His consulting practice is grounded in first-hand entrepreneurial experience, having built, scaled, and operated businesses himself; he brings a founder’s instinct for what actually works alongside the strategic rigor of a seasoned consultant. Denis brings also a rare combination of strategic innovation, platform architecture expertise, and hands-on business building to consulting assignments. He advises organizations on how to modernize their technology base, reduce structural dependency on vendors, and translate emerging capabilities such as AI, compliance tooling, and advanced payment models into scalable commercial outcomes. TLA&C – Denis Huré
Researched and drafted with AI assistance, edited and fact-checked by the author. Illustration: AI-generated.
