TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
The Pragmatic Engineer’s report draws on interviews with engineering leaders and senior engineers at 27 workplaces, plus aggregated data from CodeRabbit and Gitar. CodeRabbit’s data shows more pull requests over a year, but the report says few companies can yet establish whether AI spending produces a clear financial return.
A report by The Pragmatic Engineer finds that companies’ spending on AI coding tools varies widely, while evidence that the outlay produces a measurable return remains limited. The report combines interviews across 27 workplaces with aggregate pull-request data from CodeRabbit and Gitar, offering a snapshot of costs and output rather than a controlled measure of productivity or profit.
CodeRabbit shared anonymized data from engineers using its AI code-review platform. It reported a 113% increase in weekly pull requests and a 127% rise in monthly pull requests. In a separate view tracking the same group of 20,000 developers over a year, the increase was 128. These are platform observations; the report does not establish that AI tools alone caused the changes or that more pull requests mean more valuable work.
Gitar, an AI code-verification agent, shared data from hundreds of customers. Its figures show some teams producing more than 60 pull requests per developer per week. The report also describes different output tiers across teams, and its author suggests team norms and culture may help explain the variation. The underlying figures are not presented as a standardized, independently controlled comparison of engineering quality.
Spending differed sharply among the companies contacted. The report cites a two-person .NET business spending $3,000 per developer per month, while a publicly traded company with 1,500 engineers capped monthly AI spending at $150 per developer. One quantitative trading firm reported total AI spending of $8 million in September, up from $500,000 in January; its reported average was $1,600 per developer per month, although traders were its largest users. These examples are company-specific, not estimates of industry-wide spending.
Higher Output, Unsettled Returns
The findings show why AI budgets are drawing scrutiny: companies are paying for tools, but the most readily available output measure does not answer whether the spending saves money or improves the software being shipped. A higher pull-request count can reflect more changes moving through review, but it does not by itself measure code quality, user benefit, maintenance costs, or time saved.
That gap matters to engineering leaders deciding whether to expand licenses, impose limits, or move workloads to lower-cost models. The report says spending caps are common among medium- and higher-spending companies, and some caps can be bypassed. It also reports that several organizations are approaching the end of subsidized subscriptions, potentially exposing them to substantially higher bills. Whether those bills will be offset by productivity gains remains unproven in the material reviewed.
AI coding tools for software engineers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Companies Track AI Costs
The report was prompted by a software engineer asking what other companies spend per engineer and how pull-request throughput changed before and after AI tools were introduced. Its author sought responses from engineering leaders and senior engineers at 27 workplaces, and added aggregated data supplied by CodeRabbit and Gitar. Most company identities were withheld at their request.
The report says company size did not correspond neatly with spending in this sample. It also describes companies shifting from expensive API usage toward open models and other options; among respondents using API plans, monthly costs had roughly halved from summer peaks, according to the report. Claude models, Cursor, Codex, and GitHub Copilot were among the products respondents mentioned. These observations reflect the reported sample and are not a full market survey.
““Lots of traders are using Claude. Engineering is learning the importance of rate limiting for traders who accidentally run expensive queries without realizing it.””
— A current engineer at the quantitative trading firm, as quoted by The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
What the Output Figures Cannot Show
The report does not provide a controlled comparison showing that AI adoption caused the reported rise in pull requests. It does not establish whether the additional changes were more useful, whether they required more review or rework, or whether they improved delivery time or business results. The platform figures cover users or customers of the respective services, not all software engineers.
Spending estimates also vary by company and accounting method. Subscription plans can obscure per-engineer costs, and the report notes that some subsidies are ending. It is not clear how representative the 27 workplaces are, how companies calculated their spending, or whether reported output and cost figures cover the same teams and periods. The source material does not establish an overall ROI figure.
As an affiliate, we earn on qualifying purchases.
Budgets Face New Cost Tests
Companies facing the end of subsidized plans will need to decide whether to accept higher bills, limit usage, or change models and tools. The report says at least some organizations are looking at less expensive models and internal computing capacity, but it does not provide a timeline or evidence of savings from those efforts.
The next useful measure will be whether firms connect spending to outcomes beyond activity counts—for example, engineering time, defect rates, review burden, and delivery quality—using comparable periods and clearly defined teams. The report offers no date for a follow-up survey, so it remains unclear when stronger evidence on AI coding tools’ financial returns will be available.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does the report say about AI coding tools’ ROI?
It says few companies can answer the question clearly so far. The report does not calculate a general return on investment or show that spending produces profits or savings.
Did pull requests increase among the developers tracked?
CodeRabbit reported a 128% rise over a year for a tracked group of 20,000 developers. That is a platform observation, not proof that AI tools caused the increase.
How much are companies spending per engineer?
The examples vary substantially: the report cites a two-person company spending $3,000 per developer per month and a larger public company limiting spending to $150. These examples do not represent an industry average.
Does more pull-request activity mean better productivity?
Not on its own. Pull-request counts measure activity, but the report does not establish whether extra changes improved quality, reduced work time, or delivered more value.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
