What’s The ROI Of Software Engineers’ Spending On AI?

TL;DR

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

The Pragmatic Engineer gathered AI spending and pull-request data from 27 workplaces, alongside aggregate figures from CodeRabbit and Gitar. CodeRabbit data showed weekly PRs up 113% among surveyed platform users, but the report found that few companies could say whether higher output translated into measurable financial returns.

The Pragmatic Engineer has published findings from 27 workplaces on what companies spend on AI tools for engineers and whether that spending is associated with more software output. The report found wide variation in monthly costs and rising pull-request (PR) counts in data shared by CodeRabbit, but said few companies could answer whether AI spending produces a measurable return on investment.

CodeRabbit shared anonymized data from engineers using its AI code-review platform. It reported a 113% increase in weekly PRs and a 127% rise in monthly PRs. For a tracked group of 20,000 developers, the report described a 128% increase over a year. These are platform-level output measures; the material does not establish that AI tools alone caused the increases or that each PR represents equivalent value.

Gitar, an AI code-verification agent from Sonar, shared figures from hundreds of customers. Its data showed that some teams categorized as “AI-native” ship more than 60 PRs per developer per week. The report also noted that PR rates differed substantially among customer groups, and suggested team culture may influence how much engineers ship. That explanation is the author’s interpretation, not a finding that isolates culture’s effect.

Among the 27 workplaces, reported spending ranged from below $300 per developer per month to more than $1,000. The report said company size did not predict spend: one two-person .NET company reportedly spent $3,000 per developer monthly, while a publicly traded company with 1,500 engineers capped spending at $150 per developer. Several firms use spending limits, and some subsidized subscriptions are nearing their end, which could raise costs for companies that do not change plans or usage.

Output Gains Do Not Prove ROI

The findings show why companies are trying to measure AI costs alongside engineering output: more PRs may signal increased activity, but do not by themselves establish faster delivery, better software, lower labor costs, or greater revenue. PR volume can also vary with the size and type of changes, review practices, and team norms.

That distinction matters to engineering leaders deciding whether to expand access, set usage limits, or pay for higher-cost tools. The report’s data suggests spending can rise quickly and differs sharply even between organizations of very different sizes. Yet without a consistent baseline and measures of quality, delivery time, or business results, the financial return remains hard to compare.

For engineers, the question is not only whether AI tools help create more code changes. Companies also need to know whether those changes are useful, maintainable, and worth their direct subscription or API costs. The report offers a snapshot of spending and output, rather than a controlled assessment of productivity or profitability.

How the Spending Survey Was Built

The report began with a reader asking how much other companies spend on AI per engineer and how PR throughput changed before and after adoption. Its author said spending on AI agents was growing among technology companies and that businesses were seeking clearer cost estimates as 2027 approached.

To investigate, the author spoke with engineering leaders and senior engineers at 27 workplaces, anonymizing most companies at their request, and obtained aggregate data from CodeRabbit and Gitar. The report describes PR throughput as a readily available metric that can be compared over time, while recognizing that it is only one measure of engineering work.

The report said spending at many companies peaked in June and July and later fell by roughly half among firms on API plans. It attributed the drop, according to companies it contacted, to a combination of model changes and a move toward open models. Claude and Opus 5.5 were frequently named as major destinations for spending in the report’s limited sample; Cursor, Codex, and GitHub Copilot were also mentioned.

“On average, devs are creating more than twice as many PRs as they did a year ago.”

— The Pragmatic Engineer report, describing CodeRabbit’s data

What the PR Data Cannot Show

The report does not establish that AI tools caused the reported increases in PRs. The CodeRabbit figures cover platform users, while Gitar’s figures come from its own customer base; neither sample is described as a controlled comparison with teams not using AI. The supplied material also does not provide enough detail to assess how representative either group is of software engineers overall.

It remains unclear whether higher PR counts led to better quality, shorter delivery times, lower costs, or increased revenue. The report says few companies could quantify ROI, and does not provide a common definition or calculation for return across the 27 workplaces. Spending figures may also omit or obscure costs where subscriptions are subsidized. The reported changes in costs and tool use reflect the companies and period covered, not a forecast for all organizations.

Companies Face New Cost Decisions

Companies whose subsidized AI subscriptions are ending will need to decide whether to absorb higher bills, change tools, or limit usage. The report says some firms are preparing for costs that could rise five to ten times when subsidies end, although that estimate reflects what participating companies told the author and is not a universal projection.

The next test for businesses will be linking tool costs to results beyond PR counts, such as review burden, defect rates, delivery speed, and the value of completed work. The report does not identify a shared measurement standard or provide a date for follow-up findings. For now, its survey documents substantial variation in spending while leaving the core ROI question unresolved.

Key Questions

How much are companies spending on AI tools per engineer?

The 27-workplace survey found a broad range, from below $300 per developer per month to more than $1,000. One two-person company reportedly spent $3,000 per developer monthly, while another company capped spending at $150. These examples are from the report’s sample, not industry-wide averages.

Did AI cause the increase in pull requests?

The report shows higher PR counts in data from CodeRabbit users and Gitar customers, but it does not prove AI caused the increases. The figures are platform-based observations, not a controlled comparison that rules out other changes.

Does a higher PR count mean a better return on investment?

Not on its own. PR volume measures activity, but the report does not show whether additional PRs improved quality, reduced delivery time, cut costs, or increased revenue.

What did CodeRabbit report?

CodeRabbit shared anonymized platform data showing a 113% rise in weekly PRs and a 127% rise in monthly PRs. For a tracked group of 20,000 developers, it reported a 128% increase over a year.

Why might AI bills change?

The report says some companies saw API costs peak in June and July and later fall, which they attributed to model changes and adoption of open models. It also reports that subsidies for some subscriptions are ending, potentially increasing costs for affected companies.

Source: rss