AI Productivity Gains Begin to Emerge in Software Development
Artificial intelligence is beginning to show measurable economic potential in software engineering, even as its broader impact on workplace productivity remains difficult to capture in official economic statistics, according to an analysis by Bloomberg Opinion columnist Andy Mukherjee published on October 11, 2026.
The analysis highlights research examining how AI adoption may affect the productivity of software engineers. It argues that traditional economic measures may take time to reflect efficiency improvements as companies integrate generative AI tools into their operations.
Although businesses are investing heavily in AI infrastructure and applications, determining how much these investments improve output across the wider economy remains a challenge.
AI Productivity Research Points to Software Engineering
The analysis discusses a research paper by Alex Blumenfeld of the University of California, Berkeley, Jonathon Hazell of the London School of Economics, and other economists. The study examines the relationship between AI adoption, software engineering productivity and company valuations.
According to the findings cited in the column, between November 2022 and December 2025, financial markets increased their estimate of the present value of software engineering productivity by an amount equivalent to a permanent 32.6% increase.
The researchers examined companies in which software engineering plays an important operational role, including online marketplaces, electronics businesses and travel platforms. Software companies and firms in the semiconductor supply chain were excluded to help isolate the relationship between AI exposure and productivity expectations.
The results suggest that investors may already be accounting for anticipated efficiency gains from AI-assisted software development. However, market valuations reflect expectations and do not independently establish the productivity improvements achieved by individual companies.
AI Productivity Indicators Remain Difficult to Measure
One of the central challenges in evaluating AI’s economic impact is the gap between technological adoption and measurable productivity growth.
The analysis draws a comparison with economist Robert Solow’s observation in 1987 that the computer age was visible everywhere except in productivity statistics. Similar measurement challenges may arise as businesses adopt generative AI.
Companies can use AI tools to automate routine coding, accelerate software testing and support product development. Yet these improvements do not necessarily translate immediately into higher economy-wide output, particularly when businesses are still adapting their workflows.
Productivity gains can also emerge in less obvious areas of an organisation, making it difficult for conventional statistics to capture their full effect.
Asian Companies Offer Early Signals
The column also examines Sony, South Korea’s Coupang and Southeast Asian technology company Grab to explore whether AI-related productivity expectations are visible among major Asian businesses.
The analysis compares the performance of their US-listed shares against an artificial intelligence index and considers the proportion of software-related vacancies listed on their careers websites as an indicator of software engineering exposure.
Mukherjee describes this exercise as preliminary, noting that it relies on a small sample and does not include comprehensive payroll data. The approach therefore offers an initial indication rather than conclusive evidence of company-level productivity improvements.
The analysis also points to Alibaba as an example of the difficulties involved in identifying AI’s contribution to business performance. Factors such as semiconductor export restrictions, competition and cloud computing developments can influence share prices independently of productivity gains.
AI Coding Tools Could Change Software Development
Generative AI and AI agents are changing how developers approach software creation. Tools that can generate, review and modify code may allow engineers to complete certain tasks faster and devote more time to system design, product development and complex technical decisions.
The potential impact extends beyond technology companies. Businesses in retail, travel, entertainment and other industries increasingly depend on software to operate their services and engage with customers.
If AI-assisted development improves engineering efficiency, companies could potentially introduce products faster, reduce the time required for selected development tasks and expand their digital capabilities.
However, the scale of these benefits will depend on the reliability of AI-generated code, the complexity of projects, the need for human oversight and the ability of organisations to integrate the tools into existing processes.
AI Productivity Boom Faces a Measurement Challenge
The emerging evidence suggests that software engineering could be among the first areas where the economic benefits of AI become more apparent. Nevertheless, researchers and businesses still face difficulties separating AI’s contribution from other factors influencing company performance.
Stock market movements, recruitment patterns and research estimates can provide useful signals, but they cannot replace comprehensive measures of actual output, quality and efficiency.
As companies continue adopting AI tools, more extensive data will be needed to establish whether early gains in software development translate into sustained productivity growth across industries.
For now, the evidence discussed in Mukherjee’s analysis points to a developing trend: AI may already be changing how some businesses produce software, even though its wider economic impact remains difficult to measure.
