The introduction of generative AI into the workplace has fundamentally altered the traditional trajectory of professional development. Rather than providing a uniform productivity boost across all experience levels, these tools function as a force multiplier for those at the bottom of the skill distribution, effectively compressing the time required to reach proficiency.
A landmark study involving 5,179 customer-support agents at a Fortune 500 firm provided the initial empirical evidence for this trend. Researchers Erik Brynjolfsson of Stanford University, alongside Danielle Li and Lindsey Raymond of MIT, observed that access to an AI assistant increased successful issue resolutions by an average of 15 per cent. While experienced workers saw minimal gains, novice agents experienced a productivity jump of approximately 30 per cent.
This performance shift occurred because the AI system acted as a real-time knowledge repository. By surfacing diagnostic questions, technical documentation, and suggested responses based on historical data, the tool allowed newcomers to replicate the strategies of high-performing veterans. The model effectively codified institutional knowledge that would otherwise require months of on-the-job repetition to acquire.
Data from the study illustrates a stark compression in the learning curve. Agents without AI support typically required eight to ten months to reach a resolution rate of 2.5 issues per hour. With AI assistance, novices achieved that same milestone in approximately two months, demonstrating that the technology steepens the early phase of professional growth.
The mechanism behind these gains relies on the model’s ability to handle moderately uncommon problems. For routine inquiries, novices often possess sufficient knowledge, while extremely rare cases remain outside the model’s training data. The AI provides the most value in the middle ground, where the human agent lacks firsthand experience but the model has sufficient examples to offer reliable guidance.
Evidence suggests that this productivity gain is not merely a transient effect of system reliance. During periods of AI downtime, agents with longer exposure to the tool continued to outperform their own pre-AI baselines. This indicates that some level of knowledge transfer occurs, as workers internalize the strategies and patterns frequently recommended by the assistant.
Furthermore, the deployment of these tools appears to influence broader organizational metrics beyond individual speed. The study noted that customers became more polite and less likely to request a manager after AI deployment, suggesting that the quality of interactions improved alongside resolution rates. These secondary benefits indicate that the technology supports both the agent and the customer experience simultaneously.
Similar patterns have emerged in other professional contexts, including writing tasks and consulting. A study of 758 Boston Consulting Group consultants found that those below the median on baseline assessments improved by 43 per cent when using GPT-4. These findings consistently point to a redistribution of performance, where the floor is raised significantly for less-experienced personnel.
The broader implication is that expertise is being redefined rather than eliminated. Generative AI excels at redistributing codified patterns, which allows organizations to maintain high output even with a less-experienced workforce. This shift places a premium on the ability of senior employees to exercise judgment, identify when the AI’s suggestions are incorrect, and manage genuinely novel problems that fall outside the model’s competence.
The risk of over-reliance remains a critical watchpoint for organizational leaders. As the consulting study demonstrated, AI can lead to errors when users trust the system on tasks that sit outside its technological frontier. Future research will likely focus on whether this skill compression persists in more complex, non-routine environments where the cost of a wrong answer is significantly higher.
The long-term impact on employee retention and training programs remains to be seen. If AI can successfully bridge the gap between novice and expert performance, firms may need to reconsider how they structure mentorship and career progression. The challenge lies in ensuring that the reliance on automated assistance does not hinder the development of deep, independent expertise over the long term.
