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

The Pragmatic Engineer gathered AI spending and pull-request data from 27 workplaces, alongside aggregate figures from CodeRabbit and Gitar. CodeRabbit reported more than twice as many weekly pull requests among surveyed users as a year earlier, but the report says most companies could not yet quantify whether that added output justified AI costs.

The Pragmatic Engineer reported that software companies are spending widely varying amounts on AI tools, while many still cannot say whether the spending is producing a measurable return. Its review of 27 workplaces, together with data from CodeRabbit and Gitar, found that reported pull-request output has risen among some AI-tool users, but the figures do not establish that AI caused the increase or that it translates into business value.

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 comparison tracking the same group of 20,000 developers over a year, monthly PRs rose 128%, according to the report. These are platform-level observations; the material does not provide a control group showing what output would have been without AI.

Gitar, an AI code-verification agent, supplied data from hundreds of customers and categorized teams by PR volume. The report says some “AI-native” teams produce more than 60 PRs per developer per week. The figure describes a particular category of teams, not a general industry average, and PR counts alone do not show whether changes are useful, safe, or ultimately retained in production.

Spending varied sharply across the workplaces contacted. The report describes a two-person .NET company spending $3,000 per developer per month, while a publicly traded company with 1,500 engineers capped monthly AI spending at $150 per developer. Spending limits were common among medium- and high-spending companies, though respondents said employees could often work around them. The report also found that several companies’ costs peaked in June and July and later fell by roughly half for some API users, with respondents citing a move to open models and changes in model use.

Output Is Easier to Count Than Value

Pull requests offer companies a readily available way to track coding activity, and comparing counts before and after AI adoption can help leaders see whether work patterns changed. But a higher PR count is not the same as faster delivery, better software, lower costs, or higher revenue. A team can produce more small changes without delivering more useful work, and PR volume does not capture review burden, defects, or maintenance costs.

That gap matters as AI bills grow and temporary subsidies end. The report says some organizations expect costs to rise five to ten times when subsidized subscriptions expire unless they reduce usage or otherwise manage costs. Companies weighing renewals and broader access need measures that connect spending to outcomes, rather than treating activity counts as proof of return.

How the Spending Survey Was Built

The Pragmatic Engineer said it spoke with engineering leaders and senior engineers at 27 workplaces and gathered aggregate data from CodeRabbit and Gitar. Most company identities were withheld at the request of participants. The report presents a mix of direct responses and vendor-provided usage figures, rather than a representative survey of the software industry.

Respondents named Claude and Opus 5.5 as prominent destinations for spending in the data set, alongside Cursor, Codex, and GitHub Copilot. The report also described differences among large technology companies: it said Meta and NVIDIA encourage extensive AI use, while Apple engineers are more cautious. Those characterizations come from the report’s conversations and should not be read as a full accounting of each company’s policies or costs.

One trading firm cited in the report said its total AI spending reached $8 million per month in September, up from $500,000 in January, and averaged $1,600 per developer per month. A source at the firm said traders, rather than engineers, were its largest spenders, and that the company was developing internal computing capacity. This example illustrates how the costs can extend beyond software engineering, but it is one company’s account.

“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 the trading firm, quoted by The Pragmatic Engineer

Whether More PRs Mean Better Returns

The report does not establish that AI tools caused the reported increase in pull requests. It gives no matched comparison with non-AI teams, and the supplied material does not detail whether the PRs were similar in size, quality, or completion rate across the periods compared. The observations may also reflect changes in staffing, work practices, or company expectations.

Nor does the report provide a common ROI calculation across the 27 workplaces. It does not compare AI bills with measured savings in engineering hours, delivery speed, defect rates, or revenue. Some spending figures are estimates shared by participants, and subscription subsidies can make actual costs harder to see. The data set is limited and anonymized, so readers cannot independently examine most company-level claims.

Companies Need Outcome Measures

The report points to a growing need for companies to track AI usage and costs as subscriptions, model choices, and internal infrastructure change. It does not announce a shared industry standard or a future measurement date. For now, the next practical step for companies is to compare spending with outcomes beyond PR totals, including time saved, review workload, reliability, and the cost of correcting defects.

Whether the reported productivity signals hold up will require more comparable evidence over time. Until companies can link AI expenses to measurable engineering or business results, rising output counts and individual spending examples offer clues about adoption—not a conclusive answer on return.

Key Questions

What did the report find about AI spending and ROI?

The Pragmatic Engineer found that spending differed widely across 27 workplaces and said few companies could quantify AI’s return. The report included examples of higher PR output, but did not establish a consistent financial benefit.

How much did pull requests increase in CodeRabbit’s data?

CodeRabbit reported a 113% increase in weekly PRs among engineers using its platform. It also reported a 127% rise in monthly PRs, and a 128% increase in a tracked group of 20,000 developers over a year. These are platform observations, not proof that AI alone caused the changes.

Does a higher PR count prove that AI improved productivity?

No. PR volume measures coding activity, but does not by itself show code quality, delivery speed, business impact, or the time spent reviewing and maintaining changes.

How much were companies spending per engineer?

The reported figures ranged from a $150 monthly cap per engineer at one large public company to $3,000 per developer per month at a two-person startup. These are examples from the report, not an industry-wide average.

What remains unknown about AI’s return for engineers?

The report does not provide comparable ROI figures across its 27 workplaces or a control group for the PR data. It remains unclear how much AI changes engineering costs, software quality, delivery, and business results over time.

Source: rss