🔍 Read the full analysis: How Teams Can Measure The Payoff From AI Coding on ThorstenMeyerAI.com
Get everyday essentials 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 survey of engineering leaders at 27 workplaces found AI coding budgets ranging from $150 to about $3,000 per developer per month, with no consistent measure of return. Reported output gains vary widely, so the article argues teams should track accepted changes, rework and delivery outcomes rather than tool usage alone.
Engineering teams are setting 2027 budgets for AI coding tools without a reliable way to show what the spending delivers, according to an October 20 report by Gergely Orosz of The Pragmatic Engineer. Interviews across 27 workplaces found monthly spending from $150 to about $3,000 per developer, while reported productivity outcomes ranged from flat pull-request counts to substantial increases.
The findings arrive as some companies face higher bills when discounted subscriptions expire and usage-based pricing makes AI tool consumption more visible. Orosz reported that API-plan spending at some companies peaked in June and July and had roughly halved since, which he attributed to the release of Opus 5.5 and a shift toward open models. The report also said spending caps are common but can be bypassed.
Measures of output do not tell a consistent story. CodeRabbit, an AI code-review vendor, said pull requests created on its platform rose 113% year over year. Microsoft researchers studying tens of thousands of engineers during an early-2026 rollout of Claude Code and GitHub Copilot CLI found adopters merged about 24% more pull requests over four months than they otherwise would have. Microsoft’s authors cautioned that a merged pull request does not necessarily equal delivered value.
Companies in Orosz’s survey reported outcomes ranging from unchanged pull-request frequency to five- or tenfold increases, though the report describes individual company examples rather than a common controlled measure. One company saw pull requests become three times larger while their frequency stayed flat. The article recommends tracking cost per accepted change, rework, change-failure rates and stories closed compared with stories added, rather than treating tokens or pull-request volume as proof of return.
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
Budgets Need Delivery Measures
For engineering leaders, the difference between tool activity and useful output can shape whether AI spending is renewed, capped or expanded. A higher count of generated code or merged changes may not mean customers receive improvements faster if work creates extra review, rework or security risk.
The article points to Veracode testing in which 44% of AI coding tasks produced a known vulnerability. That figure makes review and security outcomes relevant parts of any assessment, alongside speed and cost. It does not establish that all AI-generated code is vulnerable, but it cautions against measuring payoff through volume alone.
A practical measurement approach should compare similar work over a defined period and include both engineering effort and delivery quality. Cost per accepted change, the amount of rework and whether teams close more work than enters the backlog can help distinguish productivity gains from increased code production. The supplied report does not establish a single benchmark that applies across companies.
As an affiliate, we earn on qualifying purchases.
Results Depend on the Work
The examples in the report suggest AI tools can help most with bounded tasks and consistent technology stacks. At a SaaS company with about 600 engineers, overall merged-pull-request velocity barely changed, but a staff engineer described faster customer-issue triage, code-history research and prototyping. These are reported experiences, not a controlled estimate of company-wide savings.
Ix Technologies, a two-person European agency using C# and Blazor, said it spends about $3,000 per developer each month and uses AI to clear a long-standing backlog. Cofounder Rogier van der Hee said the pair felt they could do the work of five to seven people before AI. A multinational team cited in the report said it now closes three times as many user stories each month, for the first time exceeding the number of stories coming in.
Research results also vary. 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 the confidence intervals included zero. METR said the results were unreliable in part because participants held back tasks they did not want to do without AI. The different studies use different populations, tasks and measures, so their numbers are not direct comparisons.
““a question that few companies seem able to answer – at least for now.””
— The Pragmatic Engineer report, citing its author Gergely Orosz
As an affiliate, we earn on qualifying purchases.
The available figures do not establish how much AI coding improves productivity across the industry. Vendor platform data, company reports, controlled studies and self-reported estimates measure different things. The survey’s company examples also do not show that the same results would follow from adopting the same tools elsewhere.
It remains unclear how much reported output gains translate into faster customer delivery after accounting for review, defects, security work and maintenance. The source article also does not provide a standard methodology for comparing spending or outcomes across the 27 workplaces. Its quoted 44% vulnerability result refers to Veracode’s test of AI coding tasks; it should not be read as the share of all AI-generated code that contains a vulnerability.
developer rework tracking software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Companies Set Their 2027 Metrics
Companies facing the end of discounted subscriptions or higher usage-based charges will need to decide whether to renew, cap or change their AI coding budgets. The report points toward measuring outcomes at the level of accepted changes and completed work, while tracking rework and failures so that faster code production is not mistaken for better delivery.
The next useful evidence will come from teams comparing these measures over time and sharing how they account for task complexity, code quality and security. No common industry standard or broadly accepted ROI figure is established in the source material, so budget decisions remain tied to each company’s own work and results.
AI project management tools for developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The October 20 report found budgets ranging from $150 to about $3,000 per developer per month across the workplaces surveyed. Those figures are examples from 27 workplaces, not an industry-wide average.
What measures can teams use to evaluate AI coding?
The source recommends tracking cost per accepted change, rework, change-failure rates and user stories closed compared with stories entering the backlog. These measures focus on delivery and quality rather than tool consumption alone.
Do studies agree that AI coding makes developers faster?
No. Reported results differ by study and task. Microsoft researchers reported about 24% more merged pull requests among adopters over four months, while METR’s estimates varied and had confidence intervals that included zero in its February follow-up.
Where did companies report clearer benefits?
Examples in the report include customer-issue triage, tracing the history of code, prototyping and backlog work on relatively uniform technology stacks. These are company-reported cases, not proof that every team will see the same gains.
Does the 44% vulnerability figure apply to all AI-written code?
No. It refers to Veracode’s test, in which 44% of AI coding tasks produced a known vulnerability. The source does not say that 44% of all AI-generated code is vulnerable.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
