Artificial intelligence could significantly increase software engineering productivity, with financial markets pricing in the equivalent of a 32.6% permanent productivity gain between November 2022 and December 2025, according to research from the National Bureau of Economic Research (NBER).

The finding comes as AI coding tools such as Claude Code and OpenAI Codex become increasingly common in software development. However, the research does not mean that every software engineer will automatically become 32.6% more productive. Instead, the figure represents what financial markets expected regarding the long-term productivity value of AI for software engineering.

Can AI Increase Software Engineer Productivity?

Yes. An NBER working paper estimates that AI increased the market’s expected present value of software engineering productivity by the equivalent of a 32.6% permanent increase from November 2022 through December 2025.

The researchers reached this estimate by examining how companies’ stock prices responded to AI-related developments and comparing those responses with how heavily each company depended on software engineering employees.

The result provides an economic estimate of AI’s potential impact, rather than a direct measurement of individual developer performance.

What the NBER Study Found

The research paper, “The Macroeconomic Effect of AI: Sizing the Software Engineering Channel,” was written by Alex Blumenfeld, Jonathon Hazell, Chen Lian, and Andreas Schaab.

The researchers examined companies outside the software and semiconductor industries and looked at the relationship between their stock-market performance, exposure to AI developments, and reliance on software engineering labor.

Their approach was based on a simple economic idea: if AI makes software engineers more productive, companies that employ more software engineers should potentially receive greater economic benefits from advances in AI.

The researchers then used differences in stock-market reactions to estimate the productivity improvement that investors appeared to expect.

According to the paper, the result from November 2022 to December 2025 was equivalent to a permanent 32.6% increase in software engineering productivity.

How Researchers Estimated AI Productivity Gains

The study did not simply ask developers how much faster they could write code.

Instead, the researchers used financial-market data to build a forward-looking measure of AI’s economic effect.

The process can be simplified into four steps:

1. Track AI-related market movements
2. Measure how individual companies respond to those movements
3. Compare companies based on their reliance on software engineers
4. Use an economic model to estimate implied productivity gains

Companies with a larger share of their workforce or payroll connected to software engineering could experience a greater financial impact if AI makes development work more productive.

This approach allowed the researchers to estimate how much productivity improvement financial markets were effectively pricing into company valuations.

What Does a 32.6% Productivity Increase Actually Mean?

The 32.6% figure needs context.

It does not mean that a developer using an AI coding assistant will automatically complete 32.6% more work. It also does not mean that companies have already achieved that level of productivity improvement.

Instead, the study describes the figure as the equivalent of a permanent productivity increase implied by market expectations.

Actual results can vary considerably depending on:

• The type of software being developed
• Developer experience
• AI coding tools being used
• Codebase complexity
• Testing requirements
• Security standards
• Review processes
• Company infrastructure
• How effectively developers use AI-generated code

For this reason, businesses should treat the research as an economic estimate rather than a guaranteed productivity benchmark.

AI Coding Tools Are Changing Software Development

AI-assisted programming has moved beyond basic code autocomplete.

Modern coding agents can help developers with tasks such as:

• Generating code
• Explaining unfamiliar code
• Finding potential bugs
• Writing tests
• Refactoring existing code
• Creating documentation
• Exploring large codebases
• Automating repetitive development tasks

Tools such as Claude Code and OpenAI Codex are examples of systems designed to support developers across broader parts of the software-development workflow.

This shift is important because productivity gains may come from more than simply generating lines of code. AI can also reduce the time developers spend on repetitive tasks, research, debugging, documentation, and codebase navigation.

Does AI Make Software Developers More Productive?

The research suggests that financial markets expect meaningful productivity gains from AI in software engineering.

However, productivity is broader than writing code faster.

A developer’s work can involve:

Planning → Coding → Testing → Debugging → Reviewing → Deployment → Maintenance

AI may improve some of these stages while creating additional work in others.

For example, AI-generated code still needs human review, testing, security checks, and integration with existing systems. Poorly reviewed AI output could create additional technical debt instead of reducing development time.

Therefore, the real productivity impact depends on how AI is integrated into the complete development process.

AI Productivity Gains Could Affect Company Investment

The research also raises an important question for businesses: How much should companies invest in AI development tools?

AI coding subscriptions and usage-based services add another technology expense. Companies therefore need to compare those costs with measurable improvements in development output.

Useful business metrics could include:

• Development time per feature
• Bug resolution time
• Code review time
• Testing efficiency
• Deployment frequency
• Developer hours spent on repetitive work
• Software maintenance costs
• AI tool costs per developer

Tracking these metrics can help organizations determine whether AI tools are producing measurable value in their own development environments.

AI’s Wider Economic Impact

The NBER research also estimates a broader effect on the economy.

Its baseline model suggests that the increase in software engineering productivity could correspond to a 3.6% increase in the level of GDP. When higher software engineering productivity also improves research and development productivity, the estimated GDP effect rises to 6.5%.

The researchers also report that by mid-2026, amid rapid progress in coding agents, the estimated productivity and GDP effects had more than doubled compared with the end of 2025.

These are model-based estimates rather than direct measurements of realized GDP growth caused solely by AI.

What This Means for Developers

For software engineers, the research highlights the growing economic importance of AI-assisted development.

AI tools may increasingly become part of everyday development workflows, but human engineering skills remain important.

Developers still need to understand:

• Software architecture
• Programming fundamentals
• Security
• Testing
• Debugging
• Databases
• APIs
• System design
• Code quality
• Product requirements

AI can generate code, but developers remain responsible for determining whether that code solves the right problem and works correctly in the target environment.

What This Means for Businesses

Companies considering AI coding tools should avoid treating the 32.6% figure as a guaranteed return on investment.

A more practical approach is to run controlled tests and compare development performance before and after AI adoption.

For example, a company could measure how long teams take to complete similar tasks with and without AI assistance while tracking quality, bugs, review time, and total costs.

This can provide a company-specific productivity estimate instead of relying entirely on a market-wide economic model.

AI Productivity: Key Takeaways

The latest NBER research provides a new way to estimate the economic impact of AI on software engineering.

The main points are:

• AI could substantially increase software engineering productivity.
• Financial markets priced in the equivalent of a 32.6% permanent productivity increase between November 2022 and December 2025.
• The figure represents an economic estimate, not a guaranteed productivity gain for individual developers.
• AI coding agents are expanding beyond simple autocomplete and code generation.
• Businesses need to measure AI costs against actual development outcomes.
• Human review, testing, security, and engineering judgment remain important.
• The research also estimates broader potential effects on GDP.

Conclusion

AI is becoming an increasingly important part of software development, and new economic research suggests that financial markets expect substantial productivity gains from the technology.

The NBER study estimates an equivalent 32.6% permanent increase in software engineering productivity between November 2022 and December 2025, while also finding potentially significant effects on broader economic output.

The key point is that this number should be viewed as a market-based estimate rather than a promise that every software engineer will become 32.6% more productive. As AI coding agents continue to develop, companies will need to measure real-world development speed, quality, costs, and engineering outcomes to understand their own return from AI adoption.

Frequently Asked Questions

How much can AI increase software engineer productivity?

An NBER study estimated that AI increased the market’s expected present value of software engineering productivity by the equivalent of a 32.6% permanent increase from November 2022 to December 2025. This is a market-based economic estimate, not a guaranteed gain for every developer.

Does 32.6% mean developers will code 32.6% faster?

No. The 32.6% figure represents the productivity increase implied by the researchers’ economic model and financial-market data. It should not be interpreted as a guaranteed individual developer performance increase.

Which AI tools are changing software development?

AI coding systems such as Claude Code and OpenAI Codex are examples of tools being used to assist with code generation and broader software-development tasks.

Will AI replace software engineers?

The study does not establish that AI will replace software engineers. Its focus is on estimating the economic effect of AI on software engineering productivity. Developers still perform important tasks involving architecture, testing, security, review, and technical decision-making.

Why is AI productivity important for businesses?

Higher software engineering productivity could allow companies to produce software with fewer developer hours or redirect engineering time toward additional projects. However, the actual financial benefit depends on AI costs, implementation quality, software complexity, and the resulting changes in productivity.