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

TL;DR

Before you orderOffer from Amazon

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

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

A Pragmatic Engineer report based on interviews with engineering leaders and senior engineers at 27 workplaces finds that AI coding costs vary widely, with spending caps common and no clear link between company size and spend. Data shared by CodeRabbit and Gitar points to higher pull request (PR) output among some users and teams, but the report says few companies can yet explain the return on their AI spending.

The Pragmatic Engineer has reported wide differences in how much companies spend on AI tools for software work, while finding that few can quantify the return. The report draws on conversations with engineering leaders and senior engineers at 27 workplaces and aggregate data from CodeRabbit and Gitar; platform figures show more PRs in some developer groups, but do not establish that AI spending caused the increase or paid for itself.

CodeRabbit shared anonymized data from developers using its AI code review platform. It reported a 113% increase in weekly PRs and a 127% increase in monthly PRs. In a separate tracking of the same 20,000 developers over a year, monthly PRs rose 128%, according to the report. These are measures of activity, not a direct measure of software quality, revenue or savings, and the source does not provide enough detail here to establish how much of the change came from AI tools.

Gitar, described in the report as Sonar’s AI code verification agent, shared data from hundreds of customers. It said some “AI-native” teams ship more than 60 PRs per developer per week. That figure describes a category of teams in Gitar’s data, not a typical output level across software companies. The report’s author also cautioned that team culture can shape PR rates.

Among the 27 workplaces, reported spending ranged from below $300 to more than $1,000 per developer per month. Company size did not correspond consistently with spending: one two-person .NET business reportedly spent $3,000 per developer monthly, while a publicly traded company with 1,500 engineers capped spending at $150 per developer. Caps were common among medium and higher spenders, though the report says interviewees often found them easy to work around.

PR Output Does Not Settle ROI

The gap between more coding activity and measurable business value matters as companies set budgets for AI tools. A higher PR count may show that developers are submitting changes more often, but it does not by itself show that those changes are useful, safe, maintainable or cheaper to produce. The report does not claim a causal link between tool spending and output.

For engineering leaders, the findings point to a measurement problem as well as a budgeting one. Companies may need to compare AI costs with outcomes such as review time, defects, delivery speed and the work required to maintain generated code. The source report does not present a shared method or a conclusion that any one metric captures ROI.

How Companies Track AI Costs

The report says AI-agent spending has been growing at the technology companies its author contacted, prompting companies to seek a clearer figure for costs per engineer. PR throughput was chosen as one readily available measure that can be compared before and after adoption, although it captures volume rather than the full value of engineering work.

Costs also depend on how tools are purchased. Some companies use subscriptions or subsidized plans that make per-user spending less visible; the report says several subsidies are ending, which could raise bills or prompt companies to change plans. Among companies paying by API, the author reports that monthly costs peaked around June and July and had roughly halved for some by the time of the report, attributing the decline to a mix of model changes and use of open models. The report also identifies Claude and Opus 5.5, as named in its source material, along with Cursor, Codex and GitHub Copilot, among tools mentioned by respondents.

““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 a quant trading firm, quoted by The Pragmatic Engineer

What the Available Metrics Miss

The report does not establish whether higher PR counts were caused by AI adoption, nor whether the additional PRs improved product quality or reduced overall engineering costs. CodeRabbit and Gitar supplied platform data, so their figures describe users or customer groups rather than the industry as a whole. The material provided also does not specify all methods, time windows or comparison groups behind every company-level figure.

It remains unclear how much companies spend across subscriptions, API usage and internal computing, and whether reported caps reflect actual use. The report’s high-cost examples are accounts from individual workplaces, not representative benchmarks. It also does not provide a standardized ROI calculation or enough evidence to rank tools by financial return.

Companies Face Budget Decisions

The next practical test for companies will be whether they can connect AI bills to outcomes beyond PR volume. Firms whose subsidies are ending may need to choose between paying more, limiting use or changing tools; companies using API plans may continue adjusting models and infrastructure to manage costs. The report says some large spenders are building internal computing capacity, but does not provide a timeline or financial results for those efforts.

Further evidence would need comparable cost and quality measures over time, alongside clear information about which teams use AI and how their work changes. Until companies publish or gather that evidence, the report’s spending figures and platform activity data offer a snapshot of adoption, not a settled answer on ROI.

Key Questions

What did the report find about AI spending per engineer?

Spending varied widely across the 27 workplaces surveyed, from below $300 to more than $1,000 per developer per month in the examples described. The report found no consistent relationship between company size and spending.

Did AI tools cause developers to create more PRs?

The report presents higher PR counts in CodeRabbit and Gitar data, but does not establish that AI tools caused the increases. PR volume is an activity measure, not proof of better quality or financial return.

How much did CodeRabbit’s weekly PR figure rise?

CodeRabbit reported a 113% increase in weekly PRs among developers using its platform. The report separately cited a 128% increase when tracking the same group of 20,000 developers over a year; the figures should be read with their stated populations and time frames.

Can companies currently calculate AI’s ROI?

The report says few companies it contacted could answer the question clearly. It does not supply a standard ROI formula or evidence that higher PR counts alone translate into savings or business gains.

Source: rss