TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The Pragmatic Engineer surveyed engineering leaders and senior engineers at 27 workplaces about spending on AI coding tools and changes in pull request (PR) output. Some platform data showed PR counts more than doubling year over year, but the report found that companies generally could not say whether the added output justified the cost.
The Pragmatic Engineer has published a report based on interviews with engineering leaders and senior engineers at 27 workplaces, alongside aggregate data from CodeRabbit and Gitar, examining how much companies spend on AI tools and whether developers produce more pull requests. The report found substantial variation in spending and rising PR counts in some datasets, but said few companies could quantify the return on investment.
CodeRabbit, an AI code review platform, shared anonymized data showing a 113% increase in weekly PRs and a 127% rise in monthly PRs among developers using its platform, compared with a year earlier. In a separate analysis tracking the same group of 20,000 developers for a year, PR output rose 128%. These are platform-specific observations, not a controlled measure of how much AI caused output to change.
Gitar, described in the report as Sonar’s AI code verification agent, shared data from hundreds of customers showing different levels of PR activity. The report says teams categorized as AI-native ship more than 60 PRs per developer per week. The author suggested that company culture may also shape how many PRs a developer ships; that is an interpretation, not a finding establishing the cause of the differences.
Spending reported by the 27 workplaces ranged widely. A two-person .NET company said it spent $3,000 per developer per month, while a publicly traded company with 1,500 engineers capped spending at $150 per developer monthly. Some companies use spending caps, and several told the report that subsidized subscriptions were ending, which could lead to higher visible bills. The report also says some API-plan costs had fallen from their summer peak, with respondents citing a shift to open models and changes in model use.
More Output, Unsettled Returns
The findings highlight a gap between measuring activity and demonstrating business value. PR counts are relatively easy for companies to track, and the report treats them as a useful before-and-after indicator. But more PRs alone do not establish that work is better, that delivery is faster, or that the extra output is worth the spend.
That distinction matters as AI tool costs move from experimental subscriptions into engineering budgets. Companies weighing broader adoption need to understand both direct costs and the effect on software delivery. The report offers examples of spending and output, but does not provide a standardized ROI calculation that readers can apply across organizations.
As an affiliate, we earn on qualifying purchases.
How Companies Track AI Costs
The author said an engineer asked how other companies spend on AI per engineer and whether PR throughput changed before and after adoption. In response, the report gathered accounts from 27 workplaces and aggregate platform data from CodeRabbit and Gitar. Most company names were withheld at their request, limiting independent comparison of individual cases.
The report describes spending as highly uneven: company size did not correspond neatly with per-engineer outlay, while spending caps were common among medium- and higher-spending companies. Some employees use API plans that expose costs directly; subscriptions can make the underlying usage cost less visible. Respondents also named Claude and Opus 5.5, Cursor, Codex and GitHub Copilot among tools or models used, but the report’s sample is limited and does not establish market-wide usage shares.
“The ROI of AI? A question that few companies seem able to answer – at least for now.”
— The Pragmatic Engineer report
As an affiliate, we earn on qualifying purchases.
What the Metrics Cannot Show
The report does not establish how much of the increase in PR counts was caused by AI tools. It does not offer a control group, a consistent definition of a PR across companies, or a common measurement period for every spending account. The platform figures also cover users of specific products and should not be treated as representative of all software engineers.
It remains unclear whether higher PR volume translated into improved code quality, faster releases, lower labor costs or greater revenue. The report says few companies could answer the ROI question, and anonymization prevents readers from checking many company-level details. The spending figures are self-reported accounts rather than a comprehensive industry survey.
software engineering productivity tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Tracking Costs and Delivery
The report points to companies continuing to examine AI bills, usage limits and output measures as subsidies expire and spending policies evolve. Some respondents said costs on API plans had dropped from summer highs, while others anticipated higher bills as subsidized subscriptions end. Those are reports of current company experiences, not a forecast for the wider industry.
For a firmer ROI assessment, organizations would need to compare AI costs with defined outcomes over consistent periods, while accounting for code quality and the work required to review and maintain changes. The source report does not set out a shared standard for that calculation. Whether companies can establish a reliable link between AI spending and business results remains open.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did the report find about AI spending and ROI?
Spending varied substantially among the 27 workplaces surveyed, and the report said few companies could quantify whether AI tools produced a financial return.
Did PR output increase?
CodeRabbit’s platform data showed weekly PRs up 113% year over year, with monthly PRs up 127%. These are observations from platform users, not proof that AI alone caused the increase.
Does a higher PR count prove AI is making developers more productive?
No. PR volume measures activity, but the report does not show that increases were caused by AI or that they improved quality, delivery speed or financial results.
How much do companies spend per developer?
The reported examples ranged from a $150 monthly cap per engineer at a large public company to $3,000 per developer monthly at a two-person .NET company. They are individual examples, not an industry average.
What remains unknown?
The report does not establish a standardized ROI measure or a causal link between AI use, PR volume and business outcomes. Company-level comparisons are also limited by anonymization and differing measurement methods.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
