TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
A report by The Pragmatic Engineer examined AI spending and pull request (PR) output at 27 workplaces, alongside aggregate data from CodeRabbit and Gitar. CodeRabbit’s data showed roughly doubled PR counts over a year, but the report found that few companies could say whether the added AI spend was producing a measurable return.
AI-related pull request output has risen in data shared with The Pragmatic Engineer, but a report covering 27 workplaces found that few companies could quantify whether their spending on coding tools was paying off. The findings offer comparisons of engineering output and reported costs, not proof that AI tools caused the changes or delivered a financial return.
CodeRabbit, an AI code review platform, shared anonymized data on developers using its service. It reported a 113% increase in weekly PRs and a 127% increase in monthly PRs compared with a year earlier. In a tracked group of 20,000 developers, monthly PRs were up 128%, according to the report. These figures describe activity among the data provider’s users; they do not establish that AI use alone produced the increase.
Gitar, an AI code verification agent from Sonar, provided data from hundreds of customers. It reported that some teams categorized as “AI-native” ship more than 60 PRs per developer per week. That is a vendor-provided measure of output, not an independent assessment of code quality, delivery impact or productivity across the industry.
The report’s interviews and company information showed a wide range of reported budgets, with no apparent relationship between company size and per-engineer spend in this sample. Examples included a two-person .NET business spending $3,000 per developer per month and a publicly traded company with 1,500 engineers setting a $150 monthly limit. These are selected examples from the report, not representative industry averages.
More Output, Unsettled Returns
The findings highlight a measurement gap for companies buying AI coding tools: PR volume is relatively easy to count, while the business value of that activity is harder to establish. More pull requests may signal faster iteration, but the count alone does not show whether code is better, whether work reaches users sooner, or whether review and maintenance costs rise elsewhere.
For engineering leaders and employees, the cost picture also matters. The report says spending caps are common among medium and higher spenders, and that some subsidized subscriptions are nearing their end. Companies facing those changes may see costs rise or need to adjust usage. Yet the report does not provide a consistent cost-per-PR or productivity measure that would let readers compare the return across workplaces.
As an affiliate, we earn on qualifying purchases.
How Spending Differs Across Firms
The Pragmatic Engineer said it gathered information from engineering leaders and senior engineers at 27 workplaces, anonymizing most companies at their request, and combined that material with aggregate data from CodeRabbit and Gitar. The findings are a snapshot of a limited, non-random group, rather than a representative survey of all software companies.
Spending was not uniform. The report says monthly API costs peaked around June and July at many of the companies interviewed and had roughly halved for some by the time of reporting. It attributes the drop, according to company accounts, to a mix of model changes and use of open models. Subscriptions can obscure per-user costs, while the report says several firms expect subsidies to end, potentially increasing bills by five to 10 times if they do not change their arrangements.
Tools and models named by interviewees included Claude, Opus 5.5, Cursor, Codex and GitHub Copilot. The report also described differences in company policy: Meta and NVIDIA were characterized as encouraging heavy AI use, while Apple engineers were described as more cautious. Those descriptions reflect the report’s accounts, not a standardized comparison of each company’s internal policies.
““Lots of traders are using Claude. Engineering is learning the importance of rate limiting for traders who accidentally run expensive queries without realizing it.””
— An engineer at an anonymized trading firm, quoted by The Pragmatic Engineer
AI-powered code verification tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What PR Counts Cannot Show
The evidence does not establish that AI tools caused the rise in PRs. The report does not provide a controlled comparison separating AI use from other changes, such as team size, workflow, company culture or the kinds of work being done. PR frequency also does not measure the importance or quality of the code, or how much rework it requires.
The 27-workplace sample is small and was assembled through interviews rather than a representative survey. The report does not provide a single comparable figure for total AI spending across all companies, nor a common measure of return. It also does not establish whether the reported cost reductions at some firms persisted or how changes in subscriptions and model choices affected individual workers.
As an affiliate, we earn on qualifying purchases.
Companies Face Cost Reviews
The report points to companies reviewing budgets as AI usage grows and some subsidized subscriptions end. Firms may impose or revise spending caps, shift to different models, or run models on internal computing infrastructure; the report says one high-spending trading firm was building internal GPU and CPU capacity.
The next useful evidence would connect spending to outcomes beyond PR counts, including code quality, review time, delivery speed and operating costs. The report does not identify a shared industry method or a date when one will be available. For now, its data offer a view of reported spending and activity, while the return on that spending remains unresolved.
As an affiliate, we earn on qualifying purchases.
Key Questions
Did the report find that AI tools increased developer productivity?
It found higher PR counts in data supplied by CodeRabbit and Gitar, but did not establish that AI tools caused the increase or measure productivity across all parts of software work.
How much are companies spending on AI coding tools?
The report found widely varying examples, from a $150 monthly limit per engineer at one large company to $3,000 per developer per month at a two-person startup. These examples are not industry averages.
What did CodeRabbit report about pull requests?
CodeRabbit reported that weekly PRs rose 113% and monthly PRs rose 127% among developers using its platform, compared with a year earlier. A tracked group of 20,000 developers showed a 128% increase in monthly PRs.
Did companies in the report know their AI spending had paid off?
The report says few companies could answer the ROI question. It did not find a standard measure that linked AI costs to code quality, delivery results or financial returns.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
