Why AI Isn’t Delivering the Productivity Boom We Expected

Introduction

Artificial intelligence (AI) has been hailed as the next engine of economic growth, promising to lift productivity across industries. Yet a growing body of research shows that the promised surge is still largely unrealized. Executives, workers, and scholars alike are reporting a disconnect between AI hype and measurable gains.

• --

The Promise vs. the Evidence

| Metric | Figure | Source |
|---|---|---|
| Estimated productivity gain over ten years (AI) | 0.66 % | Nobel‑prize‑winning economist, 2025 |
| Executives who say AI has not boosted productivity | 90 % | Fortune, 2026 |
| Companies reporting no productivity gains | 80 % | Reddit/Technology survey |
| Firms unsure how to use AI effectively | 38 % | Bain survey |
| Users saying AI does not perform tasks sufficiently | 36 % | Bain survey |

These numbers paint a picture of modest or absent gains, despite the massive capital outlays.

• --

Historical Parallel: The Early Computing Era

Early programmers worked under severe hardware constraints—limited memory, scarce processor cycles. Writing efficient code was a craft; the constraint itself forced deep algorithmic innovation. Decades of breakthroughs emerged directly from those limits. By contrast, today’s AI tools are abundant, reducing the pressure to optimise and potentially dampening the drive for novel algorithmic thinking.

• --

Expertise, Discomfort, and Paradigm Shifts

The literature suggests that true paradigm shifts arise from deep, embodied expertise—the discomfort experienced by seasoned practitioners who recognise the limits of existing frameworks. Novices, including many AI users without computer‑science backgrounds, are less likely to generate disruptive ideas. As Choudhury notes, “the vast majority of situations, AI will be used by people without computer‑science experience,” which can lead to superficial adoption rather than genuine productivity enhancement.

• --

Managerial Response: Measuring Speed and Quality

Productivity is not just about doing more, but doing it well. Organizations should pair usage analytics with quality‑of‑output metrics such as bug density or customer‑facing error rates. This dual approach reveals whether AI accelerates work and maintains standards.

Strategic recommendations:
1. Segment users – junior or lateral hires may benefit from always‑on copilots; experts may need targeted, skill‑diagnostic deployments.
2. Track quality – monitor defect rates, post‑deployment bugs, and customer satisfaction alongside speed.
3. Invest in training – experiments (e.g., Choudhury’s upcoming 1,000‑subject study) suggest that focused training can improve non‑CS employees’ AI effectiveness.

• --

Why the Aggregate Impact May Be Negative

Santarelli et al. (2025) pooled 371 estimates and found no robust, bias‑free relationship between AI adoption and labor‑market outcomes once methodological heterogeneity is controlled. The marginal incentive to innovate cognitively may be eroded when AI handles routine framing tasks, slowing the “invisible production of new ideas” while visible output appears faster.

• --

Employee Sentiment and Market Reactions

Analysis of millions of Glassdoor reviews shows AI‑related comments are markedly more negative than overall sentiment. Managers often expect AI investments and associated layoffs to boost firm value, yet stock‑market reactions to AI‑driven layoff announcements average near‑zero returns. This aligns with earlier findings that AI announcements do not consistently lift share prices.

• --

The Narrow Utility of Generative AI

A Bain survey revealed that 38 % of AI‑using firms are unsure how to leverage the tools, and 36 % say the tools “are not performing the tasks sufficiently.” In Denmark, a study of 25,000 workers found AI chatbots had no significant impact on earnings or recorded hours. These findings suggest generative AI excels in constrained contexts (e.g., drafting a memo) but may become counter‑productive if it narrows the scope of ideas or delays critical discussions.

• --

Policy and Organizational Implications

1. Set realistic expectations – Communicate that AI is a targeted accelerator, not a blanket productivity fix.
2. Align incentives – Reward deep expertise and the discomfort that fuels innovation, rather than mere tool usage.
3. Invest in measurement infrastructure – Combine speed metrics with quality and innovation indicators.
4. Promote inclusive training – Ensure non‑technical staff receive the skills needed to extract genuine value from AI.

• --

Conclusion

AI’s promise of a productivity revolution remains unfulfilled in the near term. While firms see faster task completion, the deeper, innovation‑driving work appears to be slowing. Understanding this nuance, measuring both speed and quality, and investing in expertise‑focused strategies will be essential for turning AI’s latent potential into real economic gains.

• --

References
1. Borrowed expertise: Why AI’s productivity boom may not survive the generation that built it | Brookings.
2. Seven Myths about AI and Productivity: What the Evidence Really Says | California Management Review.
3. Why Artificial Intelligence Isn't a Sure Thing to Increase Productivity | Working Knowledge.
4. Why AI Isn't Actually Boosting Productivity | Trumponomics (YouTube).
5. Why isn’t AI delivering its promised productivity gains yet? | LinkedIn Pulse.
6. 90% of executives say AI hasn't boosted productivity. Some are still cutting jobs | Fortune, 2026.
7. Generative AI’s Productivity Myth | TechPolicy.Press.
8. Over 80% of companies report no productivity gains from AI … | Reddit.