🔍 Read the full analysis: A Practical Way To Evaluate AI Coding’s Costs And Payoff on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Engineering teams report AI coding budgets ranging from $150 to about $3,000 per developer per month, while productivity results vary widely across companies and studies. The source argues that teams should judge the tools by accepted changes, rework, failures and backlog progress—not by spending or pull-request counts alone.
Engineering teams are spending anywhere from $150 to about $3,000 per developer per month on AI coding tools, but a review published October 23 by Thorsten Meyer says many companies still cannot link that spending to delivered value. The report draws on a survey of 27 workplaces, vendor figures and productivity studies, and recommends measuring accepted changes, rework and delivery outcomes rather than tool usage alone.
The budget figures come from engineering leaders and senior engineers interviewed by The Pragmatic Engineer, whose October 20 report also included aggregate data from code-review vendors CodeRabbit and Gitar. In the freely available portion cited by Meyer, spending varied sharply even among organizations of different sizes. Some companies set monthly caps of $150 per developer; others reported typical spending around $1,000, while a small European agency said it spent about $3,000 per developer.
Output measures were similarly inconsistent. CodeRabbit reported that developers on its platform created 113% more pull requests per week than a year earlier. Microsoft reported that engineers adopting Claude Code and GitHub Copilot CLI merged about 24% more pull requests over four months than they otherwise would have, based on an analysis of tens of thousands of engineers. Microsoft cautioned that a merged pull request does not, by itself, establish the value delivered.
Other companies in the 27-workplace sample reported unchanged pull-request frequency, larger pull requests, or gains ranging from roughly twofold to five- or tenfold. Those figures are not directly comparable: they come from different organizations and measures, and some are company reports rather than controlled studies. The source also cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability, a result that makes code quality and security part of any cost calculation.
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
The spread in reported spending and output means a flat monthly allowance cannot show whether a tool is paying off. Costs also vary with model choice, usage patterns and billing rules, while a higher count of generated or merged changes may bring more review work, rework or security exposure. Teams that track consumption alone risk treating activity as value.
Meyer’s proposed measures focus on work that reaches users and the effort required to get it there: cost per accepted change, rework, change-failure rate, and whether teams close more user stories than arrive. These measures can expose cases where an assistant speeds up issue triage or clears a backlog without increasing overall pull-request volume. They can also reveal when larger batches raise review costs or produce defects.
The distinction matters as companies prepare budgets for 2027. The source says some discounted subscriptions are nearing their end, and some companies expect bills to rise sharply if they do not change plans or usage. A team-level outcome measure gives budget owners a basis for deciding which tasks and workflows merit continued spending.
As an affiliate, we earn on qualifying purchases.
Evidence Varies by Task and Study
The evidence cited in the report does not support one universal productivity figure. In a 2025 study, METR found experienced open-source developers took 19% longer with AI on selected tasks. A February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but both confidence intervals included zero. METR also said the results were unreliable because participants withheld some tasks they did not want to attempt without AI.
Individual company examples suggest benefits can be concentrated in particular work. A staff engineer at a SaaS company with about 600 engineers told the Pragmatic Engineer that customer-issue triage had fallen from as much as a day to minutes of agent work and up to an hour of checking. The engineer also described code-history research taking minutes rather than a day or two, and prototypes taking hours rather than one to two weeks. These are reported examples, not controlled measurements of company-wide productivity.
At Ix Technologies, a two-person agency using a uniform C# and Blazor stack, co-founder Rogier van der Hee said AI helped the team work through a long-standing backlog. He described the result as feeling like two people could do the work of five to seven before AI. That assessment is his characterization, not an independently verified productivity measure. Meyer’s source says simpler, consistent stacks and bounded tasks appear more suited to such gains than complex legacy systems.
“A question that few companies seem able to answer – at least for now.”
— Gergely Orosz, summarizing findings from The Pragmatic Engineer report
As an affiliate, we earn on qualifying purchases.
What the Productivity Figures Miss
The available figures do not settle how much AI coding increases valuable output across the industry. Vendor data reflects users of those products, while company examples can rely on self-reports or distinct internal measures. Pull-request counts, stories closed and task-time estimates capture different aspects of work and cannot be treated as interchangeable.
The source does not provide a common baseline or methodology for every spending and productivity figure. It also does not establish whether reported gains persist after subscription discounts end, whether extra review and security work offsets faster coding, or how results differ by team and task. The cited vulnerability result is a test finding; it does not show that 44% of all AI-written production code is vulnerable.
As an affiliate, we earn on qualifying purchases.
Track Results Through 2027 Budgets
Companies setting 2027 budgets can compare AI expenses with a defined set of delivery and quality measures before renewing or expanding subscriptions. The report’s practical approach is to establish a baseline, track accepted changes, rework, change failures and backlog movement, then review the results by task and team rather than relying on one company-wide productivity estimate.
As billing plans and model choices change, engineering leaders will also need to monitor how those changes affect costs. The source does not identify a shared industry standard or a single upcoming study that will settle the return-on-investment question. For now, the next evidence will come from organizations comparing their own costs and outcomes over time.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The source reports budgets ranging from $150 to about $3,000 per developer per month across organizations discussed in the Pragmatic Engineer survey. The figures describe reported spending and caps, not a universal industry average.
Do the reported productivity figures prove AI coding pays off?
No. Reports of more pull requests or faster tasks are evidence of activity or perceived time savings, but do not by themselves measure delivered value. The source recommends pairing cost data with accepted changes, rework, failure rates and backlog outcomes.
Where does the report say AI coding appears most useful?
The examples point to tasks such as issue triage, code archaeology, prototyping and backlog cleanup, particularly on relatively simple and consistent technology stacks. The source presents these as observed cases, not guaranteed results for every team.
What is still unknown about AI coding’s return on investment?
There is no common measurement method across the cited studies, and it remains unclear how reported gains compare with additional review, rework, security work and changing subscription costs. Results also vary by task, organization and study design.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
