🔍 Read the full analysis: Is AI Coding Saving Your Team Money? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A report based on interviews at 27 workplaces finds AI coding spending ranges from about $150 to $3,000 per developer per month, while reported productivity results vary widely. Evidence points to gains in tasks such as triage and prototyping, but many companies cannot yet show whether higher output translates into savings or delivered value.
Engineering teams are spending from about $150 to $3,000 per developer per month on AI coding tools, but many cannot say what return that spending produces, according to a report published October 20 by The Pragmatic Engineer. Its interviews with people at 27 workplaces, alongside vendor and company data, describe uneven productivity results as businesses prepare 2027 budgets.
The report by Gergely Orosz combines interviews with engineering leaders and senior engineers at 27 workplaces with aggregate information from AI code-review vendors CodeRabbit and Gitar. The publicly readable portion shows spending caps and actual costs varying widely, with company size alone not explaining the difference. Some interviewed companies cap spending at $150 per developer monthly; others report typical costs around $1,000, while a two-person European .NET agency spends about $3,000 per developer.
Reported output measures do not line up. CodeRabbit says developers on its platform created 113% more pull requests per week than a year earlier. Microsoft’s study of 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. At other surveyed companies, pull-request counts remained flat, or grew substantially. These figures measure activity, not necessarily useful work or financial savings.
The report also describes areas where engineering staff see practical gains: customer-issue triage, understanding why existing code was written a certain way, prototyping and clearing backlogs on uniform technology stacks. At one SaaS company of about 600 engineers, merged-pull-request velocity barely changed, but a staff engineer said triage that could take a day sometimes became minutes of agent work plus up to an hour of review. The report also cites a multinational team that closed three times as many user stories per month, and a small agency working through a long-standing backlog.
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
Why Budget Owners Need Better Measures
The spread in spending and output matters as companies decide whether to expand, limit or restructure tool access. A higher number of generated or merged changes does not establish that a team has saved money: work may require more review, create rework, or fail to improve delivery. The report says some companies are also approaching the end of discounted subscriptions, making the eventual cost of continued use less certain.
It points budget owners toward outcome measures, rather than token use alone: cost per accepted change, rework, change-failure rates, and whether teams close more stories than arrive. Those measures connect tool spending to work shipped and maintained. They also allow comparisons across teams with different codebases and workflows, rather than assuming a single productivity figure applies throughout a company.
There are risks alongside the potential savings. The source cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability. That result is a warning about the tested tasks, not a measure of vulnerability rates across all AI-assisted software. It does underscore why security checks and human review costs should be counted when companies calculate returns.
As an affiliate, we earn on qualifying purchases.
How AI Coding Costs Are Shifting
The report arrives as companies set budgets for 2027 and some AI plans move toward usage-based charges. Orosz says spending at companies on API plans peaked in June and July and had roughly halved by the time of his report, which he attributes to the release of Opus 5.5 and a shift to open models. That is his account of the companies surveyed, not a universal trend across the industry.
Billing changes add pressure to make costs visible. The source says GitHub moved all Copilot plans to usage-based billing on June 1, charging for tokens at each model’s API rate and removing the cheaper-model fallback. GitHub explained the change by saying, “Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.” The report also says spending caps are common and can be bypassed, while some employers said discounts are due to expire.
Productivity research cited in the report is mixed. 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 ones, but both confidence intervals included zero; METR also described the data as unreliable because developers held back some tasks. The studies have different participants and methods, so their results do not settle how AI changes output across workplaces.
“A question that few companies seem able to answer – at least for now.”
— Gergely Orosz, The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
What the ROI Evidence Cannot Show Yet
The report does not establish an industry-wide return on investment. Its workplace interviews are not a representative survey of all engineering teams, vendor figures describe users of those vendors’ products, and reported changes in pull-request volume do not show the cost or quality of the resulting software. The full Pragmatic Engineer article is partly paywalled, and the cited material comes from its freely readable portion.
It also remains unclear how much observed output is attributable to AI tools rather than changes in staffing, task selection, processes or other factors. METR’s estimates have uncertainty, and the report says its participants sometimes withheld tasks. The 44% vulnerability finding applies to Veracode’s tested AI coding tasks; the source does not establish that the same rate applies to other models, teams or codebases.
For individual businesses, the essential missing figures are often their own: total tool and review costs, rework, quality outcomes and the value of work delivered. Without those, an increase in activity cannot by itself confirm savings.
AI productivity tools for engineering teams
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Companies Should Track for 2027
As 2027 budgets are finalized, companies will need to compare tool costs with accepted, maintained work and account for review and security effort. The report recommends tracking cost per accepted change, rework, change-failure rates, and stories completed relative to stories coming in. Those measures can show whether particular tasks or teams benefit, rather than treating adoption as proof of savings.
Further evidence will depend on companies publishing comparable results and on studies that track quality and delivery, not just coding speed or pull-request counts. Subscription discounts ending and usage-based billing will also make actual costs more visible. The source does not provide a timetable for a common industry measure, so whether firms can answer the ROI question in time for their budgets remains open.
AI coding budget management software
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 Pragmatic Engineer report says interviewed workplaces range from caps of about $150 per developer per month to spending around $3,000. These are examples from 27 workplaces, not an industry-wide average.
Does the report show that AI coding saves money?
No. It documents reported gains and costs, but says many companies cannot yet quantify their return. Higher pull-request counts alone do not establish savings or better outcomes.
Where do teams report the clearest benefits?
The report describes gains in customer-issue triage, code archaeology, prototyping and backlog work on simple, uniform technology stacks. Results vary by team and task.
What should a company measure?
The report points to cost per accepted change, rework, change-failure rates and stories closed compared with incoming work. It also cites security findings as a reason to account for review and testing.
Are the productivity numbers conclusive?
No. The figures come from different settings and methods, and some are vendor data or company reports. The cited METR estimates include uncertainty, while pull-request volume does not directly measure delivered value.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
