Despite widespread interest in generative AI, business leaders have lacked rigorous, quantified evidence of its real-world impact on productivity and quality—most debates remained speculative rather than data-driven. To address this gap, three controlled experiments examined AI-assisted workflows in customer support, business writing, and software development, measuring throughput, output quality, and skill-development effects. Results showed an average **66% productivity increase** across domains, with AI users completing more tasks while maintaining or improving quality. Notably, lower-performing workers benefited most, narrowing skill gaps, and new hires reached proficiency up to **4× faster**, suggesting AI can simultaneously boost individual output and workforce consistency.
Case Study Source: Nielsen Norman Group
Problem Statement
Business leaders lacked robust, quantified evidence of how generative AI affects real-world productivity and work quality. Most discourse was opinion-led, not data-driven, creating uncertainty about when and where AI delivers value.
Goal
Provide empirical, comparable measurements of generative AI’s impact on employee throughput and output quality across varied job types, and identify which tasks and worker segments benefit most, including effects on learning speed.
Challenges
Debate was largely speculative, with little hands-on, task-based evidence to guide decisions.
Comparing outcomes across domains with different cognitive demands made interpretation non-trivial.
Early-generation AI tools had usability issues and a learning curve, affecting initial results.
Only a subset of tasks in many roles is AI-suitable, so whole-day productivity gains vary.
Limited longitudinal coverage (only **1** of **3** studies tracked users over months) and minimal qualitative UX methods reduced behavioural insight.
Actions
Ran three controlled studies in customer support, business writing, and software development, comparing AI-assisted and non-AI workflows.
Measured task throughput (e.g., items per hour/week) and, where relevant, independent quality ratings.
Used a longitudinal design for customer support to track performance and learning over several months.
Analysed distributional effects to see how outcomes differed for lower- versus higher-skill workers.
Compared cross-domain outcomes to relate productivity lifts to task complexity and cognitive load.
Key Results
Impact
An average lift of 66% equates to roughly 47 years of typical US labour-productivity growth or 88 years in the EU, signalling a step-change rather than incremental improvement.
Gains were largest for cognitively demanding tasks, with AI handling heavy cognitive load so humans could focus on judgement, editing, and creativity.
Performance disparities narrowed and time-to-proficiency shortened, improving workforce consistency and accelerating capability building.
Quantifying Generative AI’s Real-World Impact on Productivity
Until recently, business leaders faced a dilemma. Generative AI promised transformative benefits, yet hard evidence remained scarce. Conversations relied heavily on opinion rather than data, leaving organisations uncertain about where and when these tools actually deliver value.
To address this gap, researchers set out to measure AI’s true effect on workplace performance. The aim was straightforward: gather concrete, comparable data on how these technologies influence both the speed and standard of work across different roles, and pinpoint which employees and tasks see the greatest advantage.
The Challenge of Measuring Impact
Early debates suffered from a lack of rigorous, task-level evidence. Comparing outcomes across professions—each with distinct demands—proved complex. First-generation tools often came with usability hurdles and learning curves that skewed initial findings. Moreover, AI typically suits only certain activities within a role, so day-to-day gains can be modest. Just one of three studies followed workers over several months, limiting deeper insight into how behaviour evolves with sustained use.
A Controlled, Multi-Domain Approach
The research team conducted three controlled experiments spanning customer support, business writing, and software development. Each study compared workflows with and without AI assistance, tracking how many tasks participants completed per hour or week. Quality was independently assessed where applicable. The support centre study ran longitudinally, monitoring agents over months to capture learning trajectories. Analysts also examined how benefits differed between less experienced and highly skilled workers, and related these patterns to task complexity.
Findings: Speed and Quality Both Rise
Output surged by 66% on average across the three trials—a statistically robust result. Breaking this down by sector reveals striking variation. Support staff resolved 13.8% more queries each hour. Business writers delivered 59% more documents in the same period. Developers shipped 126% more projects weekly, more than doubling their throughput.
Speed wasn’t the only metric to improve. Quality climbed too. Support agents increased their successful resolution rate by 1.3 percentage points. Written work quality jumped from 3.8 to 4.5 on a seven-point scale—a significant shift.
Perhaps most intriguing was the effect on newcomers. New support agents typically take eight months to match seasoned colleagues’ pace. With AI, they hit that benchmark in just two months—four times faster.
Lower performers benefited disproportionately. Among support staff in the bottom fifth for performance, throughput rose by 35%. Top-tier agents, already efficient, saw minimal gains. AI appears to level the playing field, narrowing the gap between weaker and stronger contributors.
What It Means
A 66% productivity lift is no incremental tweak. It represents roughly 47 years of typical labour productivity growth in the United States, or 88 years in the European Union—a genuine step change.
Gains were most pronounced in cognitively taxing work. By shouldering the mental heavy lifting, AI freed people to concentrate on judgement, refinement, and creative problem-solving. Performance became more consistent across teams, and skills developed faster. The technology didn’t just accelerate work; it reshaped how quickly and evenly capability spreads through an organisation.
Case Study Source: Nielsen Norman Group
These industry AI case studies featured on our site are based on publicly available sources and are presented for informational and educational purposes only; we do not claim ownership of these case studies or affiliation with the companies mentioned, and attribution is provided where applicable.
