Thousands of CEOs Admit AI Had No Impact on Employment or Productivity — The 40-Year-Old 'Productivity Paradox' Returns
Survey of 6,000 executives finds 90% report AI had no impact on employment or productivity over the past three years, contradicting Spotify CEO's claims. Solow's 1987 productivity paradox returns as $250B AI investment fails to deliver promised gains.
Thousands of CEOs Admit AI Had No Impact on Employment or Productivity
Solow’s Paradox Returns After 40 Years
In 1987, Nobel laureate economist Robert Solow issued a stark warning about the Information Age: “You can see the computer age everywhere but in the productivity statistics.” Despite the advent of transistors, microprocessors, and integrated circuits, productivity growth had slowed from 2.9% (1948-1973) to just 1.1% after 1973.
In February 2026, history is repeating itself.
6,000 Executives Testify to the Reality
A study published this month by the National Bureau of Economic Research surveyed 6,000 CEOs, CFOs, and other executives from the U.S., U.K., Germany, and Australia. The findings are sobering: nearly 90% of firms reported AI has had no impact on employment or productivity over the last three years.
Key findings:
- AI usage time: Only about 1.5 hours per week
- Non-users: 25% of respondents don’t use AI at all
- Users: About two-thirds reported using AI, but in limited capacity
Intriguingly, expectations diverge sharply from reality. The same executives forecast AI will increase productivity by 1.4% over the next three years. Yet their track record over the past three years? Zero.
The Spotify CEO Contradiction
This survey directly contradicts Spotify CEO Gustav Söderström’s February 14th earnings call statement:
Spotify CEO (Feb 14): “When I speak to my most senior engineers — the best developers we have — they actually say that they haven’t written a single line of code since December… They actually only generate code and supervise it.”
6,000 CEOs (Feb 17): “AI has had no impact on employment or productivity over the last three years.”
Which is true? Likely both. Companies like Spotify represent AI-forward outliers, while the majority of businesses remain far behind in AI maturity.
Warning Signs for the AI Investment Bubble
Apollo chief economist Torsten Slok invoked Solow’s observation: “AI is everywhere except in the incoming macroeconomic data. Today, you don’t see AI in the employment data, productivity data, or inflation data.”
More concerning: Outside the Magnificent Seven tech giants, there are no signs of AI impacting profit margins or earnings expectations.
Corporate AI investments swelled to over $250 billion in 2024. Yet evidence of productivity gains remains elusive.
The Trust Collapse
ManpowerGroup’s 2026 Global Talent Barometer surveyed nearly 14,000 workers across 19 countries:
- AI usage: Up 13% in 2025
- AI confidence: Down 18% over the same period
More people are using AI, but fewer believe in it. This suggests AI tools are failing to deliver promised productivity gains.
The J-Curve Hypothesis: Is There Hope?
Not all is bleak. The IT revolution of the 1970s-80s initially reduced productivity but delivered a 1.5% productivity surge from 1995-2005.
Stanford University’s Erik Brynjolfsson argued in the Financial Times that “the trend may already be reversing.” His analysis shows:
- 2025 U.S. productivity: 2.7% growth
- Q4 GDP growth: Tracking at 3.7% (with job gains of just 181,000)
This suggests a “transition from AI investment to reaping the benefits.” Slok calls this the “J-Curve” pattern: initial performance slowdown followed by exponential surge.
Implications for Coding Agents
Coding agents like Claude Code, Cursor, and GitHub Copilot have proliferated rapidly in 2025-26. However, this survey reveals a crucial distinction: tool adoption ≠ productivity gains.
Key differences from the 1980s IT revolution:
1980s IT Revolution:
- Innovators had monopoly pricing power
- Competitive products took time to emerge
2026 AI Revolution:
- “Fierce competition” among LLM builders drives prices down
- Tools are readily accessible
As Slok notes, “From a macro perspective, the value creation is not the product, but how generative AI is used and implemented in different sectors in the economy.”
Conclusion: Implementation Over Tools
AI coding agents are powerful. Claude Sonnet 4.6 launched February 17th, delivering Opus-level intelligence at Sonnet prices. But tool performance and company-wide productivity gains are separate issues.
The testimony of 6,000 CEOs is clear: for the vast majority of companies, AI has not delivered the promised productivity revolution — yet.
However, the IT revolution’s history suggests the productivity “J-Curve” may still materialize. The question is whether companies can move beyond simply “using” AI tools to deeply integrating them into organizational workflows.
Spotify’s CEO vision of “best developers not writing code” remains confined to a handful of AI-forward companies. Whether the remaining 90% of firms reach that future depends on the next few years of implementation.
References:
- Fortune: “Thousands of CEOs just admitted AI had no impact on employment or productivity” (2026-02-17)
- NBER: Study on CEO AI Usage (2026-02)
- Apollo Global Management: “AI is everywhere except in the incoming macroeconomic data”
- ManpowerGroup: 2026 Global Talent Barometer
- Financial Times: Erik Brynjolfsson on AI productivity
Related Articles
How Claude Sonnet 4.6 Agent Teams Achieve 4x Productivity: Practical Insights from Anthropic's Own Research
Two Anthropic studies—a survey of 132 internal engineers and an analysis of 1M+ real-world agent interactions—reveal the precise delegation strategies and autonomy patterns that enable high-performing teams to multiply output with Claude Sonnet 4.6 agent teams.
Claude Code v2.1.92 Released - forceRemoteSettingsRefresh, Bedrock Setup Wizard, and More
Anthropic releases Claude Code v2.1.92 with forceRemoteSettingsRefresh policy setting, AWS Bedrock setup wizard, /cost command improvements, and numerous bug fixes.
Claude Code v2.1.93 Released - Deferred Permission Decisions, Flicker-Free Rendering, and More
Anthropic releases Claude Code v2.1.93 with deferred permission decisions for PreToolUse hooks, flicker-free rendering option, PermissionDenied hook, and named subagent typeahead support.
Popular Articles
Claude Code v2.1.93 Released - Deferred Permission Decisions, Flicker-Free Rendering, and More
Anthropic releases Claude Code v2.1.93 with deferred permission decisions for PreToolUse hooks, flicker-free rendering option, PermissionDenied hook, and named subagent typeahead support.
Claude Code v2.1.92 Released - forceRemoteSettingsRefresh, Bedrock Setup Wizard, and More
Anthropic releases Claude Code v2.1.92 with forceRemoteSettingsRefresh policy setting, AWS Bedrock setup wizard, /cost command improvements, and numerous bug fixes.
Claude Code v2.1.84 Release - PowerShell Tool Preview and Environment Configuration Enhancements
Claude Code v2.1.84 introduces PowerShell tool for Windows, new environment variable overrides for model selection, idle session handling improvements, and various stability fixes.
Latest Articles
Claude Code v2.1.93 Released - Deferred Permission Decisions, Flicker-Free Rendering, and More
Anthropic releases Claude Code v2.1.93 with deferred permission decisions for PreToolUse hooks, flicker-free rendering option, PermissionDenied hook, and named subagent typeahead support.
Claude Code v2.1.92 Released - forceRemoteSettingsRefresh, Bedrock Setup Wizard, and More
Anthropic releases Claude Code v2.1.92 with forceRemoteSettingsRefresh policy setting, AWS Bedrock setup wizard, /cost command improvements, and numerous bug fixes.
Claude Code v2.1.91 Released - MCP Tool Result Persistence and Improved Edit Tool
Claude Code v2.1.91 introduces MCP tool result persistence override, improved shell execution controls, and enhanced Edit tool efficiency.