AI’s Productivity Paradox – it’s Still Not Doing the Hard-to-Learn Tasks

AI’s productivity boost is overhyped—excelling at easy tasks but struggling with complex ones. Yet, an ‘always-on economy’ could reshape work, if businesses overcome costly AI adoption hurdles.

The impact of Artificial Intelligence on productivity has sparked a raging debate between Techno-Optimists and Pragmatists. Daron Acemoglu, renowned MIT professor and well-known for questioning AI hype, in a seminal paper titled The Simple Macroeconomics of AIpublished last year, has raised some troubling questions about productivity gains from AI, which he claims would appear to be modest. In sharp contrast, a recent article titled The Always-on Economy: AI’s Real Impact in the Next 5-7 Years published in Sequoia Capital’s portal, the authors are positive that AI will lead to the elimination of traditional business hours, will reduce economic friction and increase asset utilization, thereby implying that it willresult in massive all-round economic gains.

Acemoglu’s paper further argues that early evidence of AI’s impact on productivity or economic impact is from easy-to-learn tasks, whereas some of the future effects will come from hard-to-learn tasks, where there are many context-dependent factors affecting decision-making and no objective outcome measures from which to learn successful performance.

AI might increase productivity (efficiency) by only 0.55% to 0.71% over a decade—equivalent to 0.07% per year. This is far lower than some optimistic predictions (e.g., Goldman Sachs’ 7% GDP boost). GDP growth from AI could range from 0.9% to 1.8% over 10 years, depending on investment levels.

The Question therefore is why the Small Effect?

Per Acemoglu’s research AI excels at “easy” tasks (e.g., writing simple text, coding subroutines) where success is clear-cut. But “hard” tasks (e.g., medical diagnosis, complex decision-making) are harder for AI to master, limiting its broader impact.Current evidence comes from easy tasks, but future gains will depend on harder ones, where progress is slower. He also argues that AI adoption is slow. Few businesses currently invest in AI, and organizational changes take time.

The Always-on Workforce

Techno-optimists on the other hand are confident that large-scale adoption of AI will trigger a shift toward an always-on economy which will have profound effects:
1. Economic efficiency: The elimination of traditional business hours will reduce economic friction and increase asset utilization.
2. Global competition: Time zones will become less relevant as AI enables continuous operation across borders.
3. Work patterns: Traditional 9-to-5 schedules will evolve as systems operate continuously, requiring new approaches to human oversight and intervention.
4. Resource distribution: Peak demand periods will flatten as pricing mechanisms encourage off-hours usage.

The real transformation AI brings in the next 5-7 years, according to the paper published in Sequoia Capital portal, isn’t about replacing human intelligence—it’s about removing temporal constraints from our economy. This shift toward continuous operation will create new opportunities while challenging traditional business models and work patterns. Organizations that adapt to this always-on paradigm will find themselves at a significant advantage in an increasingly fluid and accelerated economic landscape.

Monetization of AI Investments – a Challenge

Exactly a year ago, Goldman Sachs in a report “GenAI: Too Much Spend, Too Little Benefit”wrote that the promise of generative AI technology to transform companies, industries, and societies is leading tech giants and beyond to spend an estimated ~$1tn on capex in coming years, including significant investments in data centers, chips, other AI infrastructure, and the power grid. But this spending has little to show for it so far. Whether this large spend will ever pay off in terms of AI benefits and returns, and the implications for economies, companies, and markets if it does—or if it doesn’t—is top of the mind.

Companies are facing major obstacles in monetization of their AI investments. The challenges range from high infrastructure costs to ethical concerns and an oversaturated market, the path to AI profitability is full of obstacles. Training costs alone can reach millions of dollars, while ongoing cloud hosting and maintenance costs drive up the financial commitment even further. Running AI isn’t just a one-time expense; it requires continuous investment to keep things functioning at peak performance. This turns AI into a high-stakes game where only the well-funded survive.

Acemoglu in his paper observes that it will be the new tasks created with AI that can more significantly boost productivity. However, some of the new AI-generated tasks are manipulative and may have negative social value, such as deepfakes, misleading digital advertisements, addictive social media or AI-powered malicious computer attacks.

Real Productivity to Come from Hard Tasks

The MIT professor, nevertheless, maintains that real productivity gains from AI will accrue when it performs Hard Tasks. The current claims of economic benefits of AI comes from easy-to-learn tasks which may not reliably predict future gains. According to him, hard tasks typically do not have a simple mapping between action and desired outcome. In hard problems, what leads to the desired outcome in a given problem is typically not known and strongly depends on contextual factors, or the number of relevant contexts may be vast, or new problem-solving may be required.

Additionally, there is typically not enough information for the AI system to learn or it is unclear exactly what needs to be learned. Diagnosing the cause of a persistent cough and proposing a course of treatment is a hard problem. There are many complex interactions between past events that may be the cause of the lingering cough and many rare conditions that should be considered. Moreover, there is no large, well-curated data set of successful diagnoses and cures.

In hard tasks, AI models can still learn from human decision-makersbut, because there is no clear metric of success, identifying and learning from workers with the highest level of expertise will not be straightforward either. As a result, there will be a tendency for the performance of AI models to be similar to the average performance of human decision-makers, limiting the potential for large productivity improvements and cost savings. Organizations must overcome their natural tendency to go for the low-hanging and easy-to-learn, automation tasks, and instead think of innovative uses of the technology to deliver new business value that justifies the massive investments.