In 1987, the famous line from economist and Nobel laureate Robert Solow landed like a mic drop: you can see the computer age everywhere except in the productivity statistics. The parallel with today’s AI shift is hard to miss. The AI productivity story started with bright promises: computers would reshape work, boost efficiency, and multiply output. Yet the data have been stubbornly ambiguous. The Solow paradox reappears, now wearing a hoodie labeled 2026 and hinting that adoption alone does not guarantee macro gain. The era of AI adoption is real, but the curtain call of broad productivity leaps remains unwritten.
Corporate chatter about AI is loud. A Financial Times analysis shows that 374 S&P 500 firms mentioned AI in earnings calls from 2024 to 2025, often with upbeat tone about pilots and dashboards. Yet these optimistic notes seldom translate into immediate macro numbers. The gap between adoption and aggregate productivity is a familiar melody. The headline could read: many firms love AI in principle, but the macro chorus remains cautious. In practice, AI tools are used but not always integrated to the point of moving the macro needle, and the AI productivity headline often hides a more modest tempo.
Data from executive surveys add texture to the song. A National Bureau of Economic Research study found that among 6,000 CEOs and other leaders, the vast majority saw little impact from AI on operations. About two-thirds reported using AI, but that usage averaged roughly 1.5 hours per week. Nearly 90% of firms said AI has had no tangible effect on employment or productivity over the past three years. Still, expectations remained high. Executives forecast AI will lift productivity by about 1.4% and raise output by around 0.8% over the next three years, even as they anticipated a small employment wobble. The tension between promise and measured impact is very Solow paradox—just with more GPUs and fewer punchlines.
MIT researchers, in 2023, claimed AI implementation could boost a worker’s performance by as much as 40% versus nonusers. The subsequent data, however, has been more cautious. Economists wondered when or if AI would deliver a return on the billions spent on such investments, which surged past 250 billion in 2024. The arc still bends toward a gradual payoff rather than a sudden flood of efficiency. As one prominent observer quipped, AI is everywhere except in the incoming macroeconomic data, a contemporary echo of Solow’s quip. The macro data lag behind the workplace experiments, and the job is not done until the productivity numbers sing along with the hype.
AI productivity in the age of AI adoption
When we study AI productivity, we must separate pilots from production lines. Early pilots often show improvements in specific tasks—triaging customer queries, compiling reports, or scheduling with fewer emails. But these gains rarely translate into broad productivity acceleration across the whole firm in the weeks after launch. The real question is whether AI productivity tools become standard operating equipment rather than flashy add-ons. The best-case scenario is a steady climb in output per worker that grows as tools become embedded in daily routines, not just as one-off experiments. This is the practical path toward meaningful macro gains, and it requires disciplined change management, not just the latest algorithm drop.
Solow paradox revisited in 2026
The Solow paradox remains a useful frame because it forces caution without killing curiosity. The 2026 context adds layers: intense AI adoption, a competitive tech landscape, and a global economy that has learned to live with digital headwinds and supply-chain quirks. The paradox now appears as a question of timing and scale. Will the productivity surge come as a J-curve—slow at first, then accelerating as systems learn and scales converge? Or will the gains stay modest and episodic, reflecting a world where AI is powerful in bits and pieces but not yet the backbone of macro performance? In practical terms, companies are learning that AI is a tool, not a silver bullet, and macro statistics often reflect the delay between tool adoption and organizational transformation.
Some researchers see early signs of a payoff that hides in plain sight. A Stanford study using large-scale internet activity data found that generative AI can sharply improve efficiency for routine online tasks, from job hunting to travel planning. Yet the time saved often goes toward leisure or comfort rather than skilling up for higher-value work. The result is a productivity signal that is real but subtle, and sometimes misinterpreted as stagnation. This nuance matters: AI productivity may be incremental for now, with the biggest jumps coming as more complex processes are reimagined around AI capabilities rather than simply enhanced with them.
The question then shifts from whether AI productivity will eventually triumph to how quickly firms translate capability into capacity. Torsten Slok, Apollo’s chief economist, has been explicit: you may not see AI in the employment or inflation data yet, but the future value is locked in how companies implement AI across sectors. The competitive dynamic—difference-makers who move beyond pilots to integrated platforms—could unlock the kind of productivity acceleration that economists have long predicted. The risk is clear: if a majority of firms treat AI as a cost-center or a trendy upgrade, macro gains will continue to disappoint even as individual departments flirt with success.
Other studies complicate the narrative but keep the center of gravity intact: the gains are plausible, but the timing, scope, and distribution remain uncertain. ManpowerGroup’s 2026 Global Talent Barometer, surveying nearly 14,000 workers in 19 countries, found that regular AI use increased by double digits, yet confidence in AI’s utility fell by a similar margin. Confidence matters—if workers don’t trust what AI promises, adoption stalls or becomes tactical rather than strategic. In short, AI productivity is not a straight line; it resembles a winding road with forks that depend on people, processes, and policy as much as on models and machines.
There are signals of potential turnarounds too. Boston Consulting Group highlighted a phenomenon some call AI brain fry: productivity can rise when users deploy three or fewer AI tools, but can falter with four or more. The human factor—the cognitive load, the tool fatigue, and the risk of mistakes—can offset technical gains. The lesson is simple: smarter tool mixes beat sheer volume. The productivity path is not about cramming more tools into the workflow; it is about choosing the right tools and integrating them wisely. In that sense, AI productivity grows not by adding tools but by refining workflows around them.
Industry leaders also weigh long-term structural effects. IBM’s chief HR officer recently signaled a strategic shift: more hires at the entry level, a move that preserves a healthy pipeline of middle managers and leadership. The fear is that automation could hollow out the middle rungs if not managed carefully. Yet the counterpoint is robust: AI can automate repetitive tasks while human talent moves toward higher-value roles, potentially fueling a productivity expansion that benefits the entire organization. The Solow paradox here is not about inevitability but about the speed and quality of organizational transformation.
Historical patterns offer both caution and hope. The IT revolution of the 1980s and 1990s did deliver a productivity upswing in the 1990s and early 2000s, even after a long early-stage slog. Erik Brynjolfsson and Mohamed El-Erian have noted recent signals that suggest a similar dynamic could be underway: a shift from early AI investment to the practical harvest of benefits in GDP and productivity. The macro data may lag behind the micro experiments, but the trajectory toward higher productivity remains plausible, particularly as firms align AI strategy with real operational needs rather than hype alone.
Taken together, the evidence paints a balanced picture. AI productivity is real, but its macro effects depend on how quickly and how well firms embed AI into core processes. The Solow paradox persists, not as a verdict against AI, but as a reminder that technology’s impact on the economy takes time, discipline, and context to emerge. The current moment is less a triumphal march and more a careful, sometimes humorous, negotiation between promise and practicality. The future will lean toward productive use—if firms push beyond pilots and managers align incentives with long-term value creation.
A version of this story was published on Fortune.com on Feb. 17, 2026, and this piece echoes that material with fresh examples and a lighter touch. Thank you to Fortune for the original material and for sparking the ongoing conversation about AI, productivity, and the Solow paradox.
Original article attribution: Fortune – Thank you for the original material.
Practical steps to translate AI productivity into gains
- Set clear, measurable goals that connect AI pilots to real workflow improvements.
- Embed AI into daily routines, making it part of standard operating procedures rather than isolated experiments.
- Invest in training and change management to deepen adoption and reduce user fatigue.
- Track both task-level improvements and macro indicators like output per hour worked.
FAQ
- Is AI productivity guaranteed to boost output? Not automatically. Benefits depend on how deeply AI is integrated into core processes and how well teams adapt.
- Why does Solow paradox still matter in 2026? Because technology alone doesn’t guarantee macro gains; organizational change and scale matter as much as the tools.
- What should firms do now? Focus on embedding AI into workflows, align incentives with long-term value, and measure real outcomes over hype.