The evidence for task-level AI productivity gains is now substantial and broadly consistent. GitHub's research on Copilot showed a 55% reduction in task completion time for code generation. McKinsey's analysis of customer service agents found a 14% increase in issues resolved per hour with AI assistance. Boston Consulting Group's experiment with management consultants demonstrated a 40% improvement in quality scores for tasks within GPT-4's capability frontier. Academic studies across writing, analysis, and data processing report similar effect sizes, typically in the 20-50% range.
These are not trivial numbers. A 30-55% improvement in task-level productivity, if it translated directly to organisational output, would represent the most significant productivity shock since electrification. It would resolve the AI productivity paradox overnight. It would show up in national statistics within quarters, not decades.
It has not. OECD total factor productivity growth across advanced economies averaged 0.3% in 2024. US non-farm labour productivity growth was 1.5% between 2020 and 2025. The NBER reports that 90% of firms see no measurable productivity impact from AI. Something is absorbing, diluting, or fundamentally transforming these task-level gains before they reach the organisational level.
Understanding what happens in that gap is not merely an academic question. It is the key to understanding where AI creates enterprise value — and why the accounting systems designed to capture that value are looking in precisely the wrong place.
The gap between individual task gains and organisational outcomes is not unique to AI. It is a recurring feature of general-purpose technology adoption that economists have studied across multiple transitions. But the specific mechanisms through which AI task gains dissipate are worth examining, because they point directly to the intangible assets being created in the process.
When individual contributors become faster at producing outputs, the organisation must process more outputs. Reports that used to take three days now arrive in one. Code that used to require a week of development is ready in two days. The bottleneck shifts from production to review, integration, and decision-making.
Ethan Mollick's research at Wharton documents this pattern clearly. Consultants who used AI produced higher-quality individual outputs but created coordination challenges: more alternatives to evaluate, more drafts to review, more decisions to make. The task-level gain was real; the organisational gain was diminished by the increased coordination burden.
Faster individual output without corresponding changes to review processes, decision-making structures, and integration workflows creates organisational congestion. The AI makes people faster; the organisation's ability to absorb that speed has not changed.
When a customer service agent resolves 14% more issues per hour with AI assistance, the quantity metric improves. But the more significant transformation is often qualitative: the agent handles more complex issues, provides more thorough explanations, and creates better documentation. These quality improvements are real but largely invisible to traditional productivity measures.
This is a recognised problem in productivity measurement. William Nordhaus demonstrated that quality-adjusted price indices for computing equipment showed productivity improvements two to three times larger than unadjusted measures. Brent Moulton's work at the Bureau of Economic Analysis identified similar quality-adjustment challenges across services sectors.
AI-assisted work frequently shifts the quality frontier rather than simply accelerating existing quality levels. A developer using Copilot may not just write code faster — they may write better-tested, better-documented, more secure code. An analyst using AI may not just produce reports faster — they may explore more scenarios, surface more insights, and identify more risks.
The quality dimension of AI productivity is precisely where intangible assets are being created. Better code creates technology capital. Better documentation creates organisational capital. Better customer interactions create customer relationship value. All invisible to GDP per hour.
When AI automates a task that previously consumed 40% of an employee's time, the question is not how much time was saved — it is how the saved time is used. If the freed time is spent on activities that are productive but unmeasured (innovation, relationship building, strategic thinking, process improvement), measured productivity may not change at all.
This is Daron Acemoglu's "so-so automation" critique applied at the individual level. The automation of routine tasks creates capacity for non-routine work. But non-routine work — particularly the kind that creates intangible assets — is exactly the work that productivity statistics struggle to capture.
When AI frees a knowledge worker from routine tasks, the most valuable use of the freed time is typically investment in intangible assets: learning, relationship building, creative problem-solving, process innovation. These activities create enormous long-term value but register as zero short-term productivity improvement. The more effectively a firm reallocates AI-freed time to high-value intangible investment, the worse its measured productivity performance may appear.
The economic history of general-purpose technologies is unambiguous on one point: technology deployment without complementary organisational investment has never delivered economy-wide productivity gains. Paul David's seminal analysis of electrification demonstrated that factories required 25-30 years to capture the full productivity potential of electric motors — not because the technology was slow to improve, but because realising its potential required redesigned factory layouts, new management practices, retrained workforces, and restructured supply chains.
For AI, the required complementary investments are predominantly in intangible assets:
| Complementary Investment | Asset Category | Typical Timeline | Measurement Challenge |
|---|---|---|---|
| Process redesign around AI capabilities | Organisational capital | 6-18 months | Expensed as operating costs |
| Data infrastructure and curation | Data assets | 12-24 months | Treated as IT maintenance |
| Training and upskilling | Human capital | 6-12 months | Expensed as training costs |
| AI-native workflow development | Technology capital | 12-36 months | Expensed as R&D |
| Cultural adaptation and change management | Organisational capital | 18-36 months | Not measured at all |
Every one of these investments is treated as a current-period expense under prevailing accounting standards. None appears as an asset on the balance sheet. The result is predictable: companies that invest most heavily in the complementary assets required to capture AI productivity gains appear to have the worst financial performance in the short term — higher costs with no visible asset creation.
A professional services firm invests £2 million in redesigning its delivery methodology around AI capabilities. This includes process mapping, workflow automation, quality framework revision, training programmes, and change management. Under IAS 38 and prevailing accounting practice, the entire £2 million is expensed immediately. The resulting organisational capital — a fundamentally more efficient delivery methodology — does not appear on the balance sheet. A competitor reviewing the firm's accounts would see higher operating expenses and unchanged revenue, concluding that the AI investment had failed.
A more careful reading of the emerging evidence suggests that AI is creating substantial value — but in categories that traditional productivity measures cannot capture.
Brynjolfsson, Li, and Raymond's 2023 study of AI-assisted customer service agents found something more nuanced than the headline 14% productivity improvement. The gains were dramatically larger for novice workers (34% improvement) than for experienced workers (minimal improvement). This suggests that AI was not simply making individuals faster — it was transferring organisational knowledge from expert to novice, accelerating the creation of human capital at a rate that traditional training programmes cannot match.
Similarly, the BCG study found that consultants using AI produced work that was 40% higher quality — but only for tasks within the technology's capability frontier. For tasks beyond the frontier, AI-assisted consultants performed 23% worse than unassisted consultants. This "falling asleep at the wheel" effect highlights that AI productivity is not a simple multiplier but a complex interaction between human judgement, AI capability, and task characteristics.
The gap between task productivity and organisational productivity is not a failure of AI. It is evidence that AI is creating value in categories that our measurement systems were never designed to detect.
If AI's primary value creation is in intangible assets rather than task-level efficiency, the measurement framework must change accordingly.
Task completion time is a first-order metric. Second-order effects — quality improvements, knowledge transfer, process codification — create more value but require different measurement approaches.
For every AI investment, identify which intangible asset categories are being enhanced: technology capital, data assets, organisational capital, human capital, customer relationships. Use the six-asset framework as a starting point.
The ratio of complementary investment (process redesign, training, data infrastructure) to AI technology spend is a leading indicator of eventual productivity capture. Historical evidence suggests a 5:1 to 10:1 ratio is required.
Paul David's electrification research showed a 25-30 year adoption cycle. Even optimistic estimates for AI suggest 10-15 years for full organisational adaptation. Quarterly ROI assessments of AI investment are measuring the wrong timeframe.
For investors, boards, and management teams, the task-to-organisation productivity gap contains a strategic insight: the companies creating the most AI-driven enterprise value are not necessarily the ones showing the most immediate productivity gains.
A firm that deploys AI copilots and achieves a 30% reduction in code development time has created a measurable efficiency gain. A firm that uses that freed capacity to build proprietary data assets, redesign its customer engagement model, and codify its institutional knowledge into AI-augmented workflows has created something far more valuable — a portfolio of intangible assets that compounds over time and creates sustainable competitive advantage.
The measurement systems that distinguish between these two outcomes — that can identify, quantify, and value the intangible assets being created by AI investment — are what separates informed capital allocation from blind technology spending.
The disappearance of task-level AI gains at the organisational level is not a failure. It is a transformation — from visible efficiency improvements into invisible intangible asset creation. The firms that recognise this transformation, measure it, and manage it will capture the genuine productivity gains that AI makes possible. The firms that measure only task speed will continue to wonder where the productivity went.
The Opagio Growth Platform provides the structured framework for measuring AI's intangible asset creation — connecting task-level deployment to the organisational capital, technology capital, and data assets that drive enterprise value.
David Stroll is Co-Founder of Opagio and leads the company's research into productivity measurement and intangible asset frameworks. His work draws on OECD, ONS, and Productivity Institute research to connect macro-economic productivity analysis with practical enterprise measurement.
Global AI investment has surpassed $500 billion annually, yet total factor productivity growth remains stubbornly flat. This is Solow's paradox reborn: you can see AI everywhere except in the productivity statistics. History tells us why — and what needs to change.
Read more →
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 →
Only 29% of executives can measure AI ROI confidently. The problem is not that AI fails to deliver value — it is that traditional ROI frameworks were not designed for investments that create intangible assets. Here is the 4-layer framework that connects AI spending to board-level outcomes.
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.