In 1987, Robert Solow made an observation that has haunted economists for nearly four decades: "You can see the computer age everywhere but in the productivity statistics." Replace "computer age" with "artificial intelligence" and you have the defining economic puzzle of 2026. Global AI investment now exceeds $500 billion annually. OpenAI's valuation tripled from $157 billion to $500 billion in a single year. Yet total factor productivity growth across the OECD remains stubbornly, almost defiantly, flat.
This is not a minor discrepancy. It is a structural disconnect between the most significant technological deployment since electrification and the economic metrics that are supposed to capture its impact. Understanding why this gap exists — and what it means for investors, boards, and policymakers — requires looking beyond the technology itself.
The original Solow paradox emerged during the early personal computer revolution. Between 1973 and 1995, US firms invested hundreds of billions of dollars in computing equipment, yet labour productivity growth averaged just 1.4% per annum — well below the 2.8% rate of the post-war golden age. Economists searched for the missing productivity everywhere. They did not find it for over a decade.
The resolution came slowly. Erik Brynjolfsson and Andrew McAfee demonstrated that IT investment required complementary organisational transformation — new management practices, restructured workflows, retrained workforces — before productivity gains materialised. The technology was necessary but radically insufficient on its own.
Today, the pattern is repeating with remarkable fidelity. BLS data shows US non-farm labour productivity growth averaged 1.5% between 2020 and 2025, a period of unprecedented AI deployment. The Conference Board estimates that total factor productivity across advanced economies grew at just 0.3% in 2024. OECD figures tell a similar story: TFP growth across member countries has not accelerated despite the AI investment surge.
The AI productivity paradox is not a failure of AI technology. It is a failure of organisational adaptation. Technology deployment without complementary transformation has never delivered productivity gains — not with steam, not with electricity, not with computers, and not with AI.
One of the most important — and least understood — dynamics of major technology adoption is the productivity J-curve. When firms deploy a new general-purpose technology, measured productivity initially declines before eventually rising.
This is not a theoretical curiosity. It is an empirical regularity documented across multiple technology transitions. The mechanism is straightforward: firms must simultaneously maintain existing operations while investing time, capital, and management attention in learning, reorganising, and adapting to the new technology. During this transition period, output per hour falls because inputs increase without proportional output gains.
| Phase | Characteristic | AI Example | Productivity Impact |
|---|---|---|---|
| Phase 1: Deployment | Technology installed, workflows unchanged | Adding ChatGPT subscriptions, no process redesign | Negative (cost increase, minimal output gain) |
| Phase 2: Reorganisation | Workflows restructured around technology | Redesigning customer service with AI agents | Flat to slightly negative (disruption costs) |
| Phase 3: Complementary investment | New skills, processes, management practices | Training teams, building data pipelines, new KPIs | Flat (investment without immediate return) |
| Phase 4: Productivity takeoff | Organisational model aligned with technology | AI-native processes, compound efficiency gains | Strongly positive (exponential returns) |
Most firms in 2026 are somewhere between Phase 1 and Phase 2. They have purchased the technology but have not completed the far more difficult organisational transformation required to extract value from it. The productivity gains are not missing — they are deferred.
The historical parallel most instructive for the current moment is not the personal computer. It is electrification.
Thomas Edison opened the first commercial power station in 1882. Yet the productivity impact of electrification did not appear in US manufacturing data until the 1920s — a lag of roughly 40 years. The reason was not that electric motors were inferior to steam engines. It was that factory owners initially used electric motors as direct replacements for steam engines, placing a single large motor where the steam engine had been and running the same belt-and-shaft system.
The productivity transformation only arrived when a new generation of factory managers reconceived the entire production process around the unique capabilities of electric power: individual motors on each machine, flexible floor layouts, single-storey factories, natural lighting. The technology required a complete redesign of the production system.
When Henry Ford adopted unit drive — placing individual electric motors on each machine rather than running them from a central power source — he was able to redesign the factory floor around the logic of the assembly line rather than the constraints of belt-driven power. Productivity did not merely improve; it was transformed. The resistance was not technical but organisational. Managers trained in steam-era factory design could not imagine the alternative until they saw it working.
The AI parallel is precise. Most firms today are using AI as a substitute — doing the same things slightly faster — rather than as a complement that enables entirely new ways of working. The productivity gains from AI will arrive when organisations are redesigned around AI's distinctive capabilities, not when AI is bolted onto existing processes.
There is a second, deeper reason why AI's productivity impact remains invisible: the measurement framework itself is inadequate. National income accounts — the GDP statistics, labour productivity figures, and TFP calculations that economists rely on — were designed for an economy dominated by physical capital. They systematically undercount intangible investment and the assets it creates.
When a firm invests in AI, it creates intangible assets: trained models, proprietary datasets, new organisational knowledge, enhanced brand capability, improved customer relationships. Under current accounting standards (IAS 38, ASC 350), most of this investment is expensed immediately — treated as a cost rather than an investment. The resulting assets do not appear on balance sheets. They do not appear in capital stock calculations. And they do not appear in TFP residuals.
This is not a minor technical point. The ONS estimates that UK intangible investment reached approximately 60% of tangible investment in recent years. The OECD Compendium of Productivity Indicators acknowledges that mismeasured intangible capital is a significant source of TFP residual variation across countries. If a material portion of AI spending is creating unrecorded intangible assets, then the productivity statistics are not merely lagging — they are structurally incomplete.
The SNA 2025 (System of National Accounts) update represents the first major revision to treat certain categories of data and AI investment as capital formation rather than intermediate consumption. However, implementation across national statistical offices will take years. In the interim, the measurement gap persists — and widens with every pound of AI investment.
| Metric | Source | 2024 Figure | Trend |
|---|---|---|---|
| US non-farm labour productivity growth | BLS | 1.5% (5-year avg) | Flat |
| OECD TFP growth (advanced economies) | OECD | 0.3% | Flat |
| Global corporate AI spending | IDC | $500B+ | Rising rapidly |
| Firms reporting measurable AI productivity gains | NBER | ~10% | Slowly rising |
| S&P 500 value attributable to intangible assets | Ocean Tomo | 92% | Rising steadily |
| Executive confidence in measuring AI ROI | Deloitte | 29% | Low, stable |
The gap between the investment column and the productivity column is where the paradox lives. But it is also where the opportunity resides — for firms that can measure the intangible assets AI creates, and for investors who can value what accounting standards cannot see.
If history is any guide, the resolution of the AI productivity paradox requires three things that most firms have not yet attempted.
First, complementary investment in organisational capital. AI tools require new management practices, decision-making processes, and performance metrics. Firms must invest in organisational capital — the processes, routines, and institutional knowledge that determine how technology is deployed — at a level comparable to their investment in the technology itself. My research on the UK productivity puzzle demonstrates that organisational capital is consistently the most underinvested intangible asset category.
Second, workforce transformation beyond training. Adding AI skills to existing job descriptions is not sufficient. The firms that extracted productivity from electrification were not those that taught workers to use electric motors — they were those that redesigned work itself. AI requires a similar reimagining of roles, teams, and decision rights.
Third, measurement systems that capture intangible value. If firms cannot measure the intangible assets AI creates — technology capital, data assets, organisational knowledge, enhanced customer relationships — they cannot manage the transition or demonstrate returns to investors. The Opagio questionnaire is designed precisely for this purpose: making the invisible measurable so that AI investment can be managed as capital formation rather than a cost line.
The gap between AI capability and AI productivity is not a reason for pessimism about AI investment. It is a reason for rigour. The firms that will deliver outsized returns are those that combine AI deployment with complementary organisational transformation — and that can demonstrate measurable intangible asset creation. For PE firms and institutional investors, the ability to distinguish between AI spending and AI value creation is rapidly becoming a core competency. Tools that measure intangible asset formation are no longer optional — they are essential to informed capital allocation.
Opagio exists at the intersection of the AI productivity paradox and the intangible asset measurement gap. Our growth platform provides the measurement layer that connects AI investment to intangible asset creation, productivity improvement, and valuation impact. When 90% of firms report zero AI productivity impact and only 29% of executives can measure AI ROI, the problem is not the technology — it is the absence of a measurement framework designed for an intangible-asset economy.
The productivity gains from AI will arrive. History guarantees it. The question is which firms — and which investors — will be positioned to capture them. The answer, as it was with electrification and with computing, will be those who invest in measurement and organisational transformation alongside the technology itself.
The Productivity 250 series on this site traces 250 years of productivity revolutions. Every one followed the same pattern: technology arrived decades before the productivity statistics reflected its impact. AI is no different. The paradox is not permanent — but resolving it requires the kind of rigorous, data-driven approach to intangible asset measurement that most firms have yet to adopt.
David Stroll is Co-Founder and CTO of Opagio. His PhD research at Birkbeck, University of London explored whether UK SMEs can solve the Productivity Puzzle. He has published research on intangible asset data collection (ESCoE/ONS, 2021), innovation diffusion measurement (ISPIM, 2018), and intangible capital frameworks (Big Innovation Centre, 2017). Previously, he co-founded PayMode, the first B2B internet payment service, and held senior roles at Digital Equipment Corporation. Learn more about the Opagio team.
Traditional productivity metrics were designed for an economy dominated by physical capital and tangible output. In a world where the most valuable firms derive their competitive advantage from software, data, brand equity, and organisational know-how, those metrics are increasingly unreliable. Here is why the gap matters — and what we should do about it.
Read more →
The San Francisco Fed's recent research on AI and productivity is among the most rigorous work in the space. But it also reveals critical blindspots in how economists measure the impact of AI on economic growth. Here is what they got right, what they missed, and why the measurement gap itself is the most important finding.
Read more →
Individual developers using AI copilots write code 30-55% faster. Individual analysts produce reports in half the time. Yet organisational productivity remains flat. The gap between task-level AI gains and enterprise-level outcomes reveals a measurement failure — and an intangible asset opportunity — that most firms are missing entirely.
Read more →Get the latest insights on intangible asset growth and productivity delivered to your inbox.
Book a free consultation to see how the Opagio Growth Platform can help your business.