Harness Study Finds 89% See AI Productivity Gains as 94% Say Metrics Miss Key Costs
Updated
Updated · PR Newswire · May 13
Harness Study Finds 89% See AI Productivity Gains as 94% Say Metrics Miss Key Costs
3 articles · Updated · PR Newswire · May 13
700 engineers and managers surveyed across five countries reported a widening AI productivity gap: 89% of leaders say developer productivity improved, but organizations still cannot reliably measure whether those gains are real.
94% said current frameworks miss critical factors such as code quality, validation time, tech debt and burnout, even as 89% still called those metrics accurate; only 6% think existing frameworks can fix the problem.
31% of developer time is estimated to go to "invisible work" such as reviewing AI-generated code, fixing bugs and switching tools, while 81% said code-review time has risen after adopting AI coding tools.
54% fear AI productivity data could be used for individual performance reviews, and managers were nearly four times as likely as practitioners to report no concerns about such use—15% versus 4%.
Harness said teams should add measures for code quality, validation time, cognitive load and burnout, track AI output separately from human work, and keep improvement data distinct from performance evaluation.
If 94% of leaders admit their AI metrics are flawed, why are they still using them to judge performance?
AI promises speed but triples production incidents. Is the software industry building a ticking time bomb of unmaintainable code?
AI is automating junior tasks while burning out seniors. Is the software engineering career path fundamentally broken?
Beyond the Hype: Measuring Real Productivity, Hidden Costs, and Workforce Impact in the Age of AI
Overview
This report explores the growing gap between high expectations for AI-driven productivity and the limited, sometimes negative, real-world results seen so far. While executives and managers are enthusiastic about AI's potential, the actual impact often falls short, echoing past technology revolutions that took time to deliver broad economic benefits. The so-called AI productivity paradox arises because AI tends to redistribute value rather than create new wealth, giving early adopters an edge while others lag behind. This uneven value creation leads to a divergence between perceived and real productivity, highlighting the need for new strategies and better measurement to realize AI's true potential.