🔍 Read the full analysis: What Makes AI Coding Pay Off For Engineering Teams? on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A report drawing on interviews at 27 workplaces and data from AI coding and review tools finds wide variation in spending and productivity results. Teams report clear gains in triage, code archaeology, prototyping and backlog cleanup, but pull-request counts alone do not show whether those gains create business value. Security, rework and cost per accepted change remain key measures.
Engineering teams are spending about $150 to $3,000 per developer per month on AI coding tools, but a review of data from 27 workplaces finds no consistent measure of what that spending delivers. The findings, published by Thorsten Meyer on October 23 and based partly on reporting by The Pragmatic Engineer, suggest AI is most useful for specific tasks such as triage and backlog cleanup, while higher code output does not by itself prove a return on investment.
The underlying report was published on October 20 by Gergely Orosz of The Pragmatic Engineer, who spoke with engineering leaders and senior engineers at 27 workplaces and included aggregate figures from code-review vendors CodeRabbit and Gitar. The accounts describe spending caps ranging from $150 to $4,000 per month in some companies, with typical spending around $1,000 at organizations with higher caps. Company size did not reliably predict the level of spending.
Output measures also differ. CodeRabbit reported that developers using its platform created 113% more pull requests per week than a year earlier. Microsoft researchers studying tens of thousands of engineers during an early-2026 rollout found adopters merged about 24% more pull requests over four months than they otherwise would have. Those figures measure activity, not necessarily quality or business value. Microsoft’s researchers cautioned that a merged pull request is not the same as value delivered.
Results from individual companies ranged from flat pull-request frequency to several-fold increases. At one company, pull-request counts stayed flat while average pull-request size tripled. A SaaS firm with about 600 engineers reported little change in merged-pull-request velocity but described faster customer-issue triage, research into why code had been written a certain way, and prototyping. Its staff engineer said triage that could take a day sometimes became minutes of agent work plus up to an hour of human checking.
Infographic 01 · Where the money goes
From a $150 cap to $3,000 a month per developer.
Anecdotal data from 27 workplaces, mostly anonymized. It shows how wide the range is, not an industry average.
Spend per developer per month
Examples from the free part of the article · scale to $4,000
actual or typical spendbudget cap
What changed in 2026
Why the 2027 budget needs a range
- Peak in June and July: spend has roughly halved since for companies on API plans (Opus 5.5 and open models).
- Caps are common at medium and high spenders, and most people say they are easy to get around.
- Subsidies are ending: some companies must optimize or see the bill rise 5–10×.
- GitHub Copilot bills by token since 1 June 2026, at each model’s API rate, with no fallback to a cheaper model.
- Company size does not predict spend.
ROI is “a question that few companies seem able to answer – at least for now.”Gergely Orosz, The Pragmatic Engineer, 20 Oct 2026
Sources: The Pragmatic Engineer, “What’s the ROI of software engineers’ spending on AI?” (20 Oct 2026, free part); GitHub (27 Apr 2026).
Powered by Thorsten Meyer AI
Where AI Saves Engineering Time
The findings matter as companies set 2027 engineering budgets and some vendor discounts expire. The source report says several companies expect their costs to rise substantially if they do not optimize; one quant trading firm’s monthly AI spending rose from $500,000 in January to $8 million in September, though its spending included traders and was tied to desk-level profit and loss rather than a general engineering productivity measure.
Reported gains appear clearest where work is bounded and the technology stack is uniform: issue triage, code archaeology, prototypes and backlog cleanup. Ix Technologies, a small European agency using C# and Blazor, said it was clearing a backlog stretching back about a decade. Cofounder Rogier van der Hee described the two-person team as feeling able to do work that previously required five to seven people. That is the company’s account, not an independently verified productivity comparison.
There are also costs beyond tool subscriptions. If AI increases code volume without matching review capacity, teams may take on more rework or security risk. Veracode’s latest test, as cited in the source material, found that 44% of AI coding tasks produced a known vulnerability. That result does not establish the vulnerability rate in every company’s code, but it underlines why teams need to track quality and security alongside output.
As an affiliate, we earn on qualifying purchases.
Why Budget Pressure Is Rising
Orosz’s reporting comes as companies reconsider spending and vendors change how AI use is billed. The source article says spending at companies on API plans peaked in June and July and had roughly halved since, which Orosz attributed to the release of Opus 5.5 and a shift toward open models. It also reports that usage caps are common and can be bypassed, limiting their usefulness as a firm budget control.
GitHub changed billing for all Copilot plans on June 1, moving to usage-based charges tied to each model’s API rate and removing the fallback to a cheaper model. GitHub said the change reflected higher compute and inference needs as agentic use became more common. The source article also describes companies whose discounted subscriptions are due to end, with some warning that costs could rise five- to tenfold if they do not adjust. Those are reported company expectations, not a confirmed increase for all customers.
Productivity studies cited in the reporting do not give a single settled answer. METR’s early-2025 study found experienced open-source developers took 19% longer with AI. A February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but both confidence intervals included zero. METR said the data was unreliable, partly because many participants withheld tasks they did not want to do without AI.
“A question that few companies seem able to answer – at least for now.”
— Gergely Orosz, summarizing the reporting for The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
What the Productivity Numbers Miss
There is no shared, independently validated measure showing how much AI coding tools improve engineering outcomes across different teams. Vendor figures reflect activity on their platforms, company accounts vary by task and workplace, and study results have uncertainty. METR also reported that participants in its 2025 survey overestimated AI’s effect by 40 percentage points, according to the source article.
It remains unclear how often increased output translates into faster delivery, fewer defects or lower total costs. The cited security test does not establish how frequently vulnerabilities reach production, and the source material does not provide comparable cost-per-change figures across the 27 workplaces. The source text also ends partway through its description of a DORA report, so no further finding from that report can be established here.
As an affiliate, we earn on qualifying purchases.
How Teams Can Test Their Returns
The source article recommends measuring results rather than tool consumption. Teams can compare cost per accepted change, rework, change-failure rates and the number of user stories closed against the number arriving. These measures can help distinguish a rise in generated code from a rise in work that is accepted, reliable and useful.
As budgets are set for 2027 and discounted plans expire, companies will need to decide whether to retain, limit or change their AI tool use. The available reporting does not identify a single next industry-wide benchmark or deadline. For now, the practical test is whether a team can link spending on a defined task to a measurable outcome without allowing review and security checks to fall behind.
AI-powered coding productivity tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are engineering teams spending on AI coding tools?
The workplaces in the reporting described budgets ranging from about $150 to $3,000 per developer per month, with some company caps reaching $4,000. These are reported examples, not a representative industry-wide average.
Does more code or more pull requests prove AI is paying off?
No. Pull-request counts show activity, but do not establish that changes are useful, reliable or valuable to customers. Microsoft researchers cited in the reporting cautioned that a merged pull request is not the same as value delivered.
Which engineering tasks appear to benefit most?
Company accounts in the report point to triage, code archaeology, prototyping and backlog cleanup, particularly on simple, consistent technology stacks. The reported gains vary by workplace and are not a guarantee of similar results elsewhere.
What should teams measure to assess AI coding costs?
The source article suggests tracking cost per accepted change, rework, change-failure rate and stories closed compared with stories coming in. It also cites security concerns, including a Veracode test in which 44% of AI coding tasks produced a known vulnerability.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
