🔍 Read the full analysis: AI Coding Costs And Returns: A Clearer Way To Measure The 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
A review of AI coding budgets and workplace data published by The Pragmatic Engineer found spending ranges from $150 to about $3,000 per developer per month, while measured output varies widely. The evidence points to gains in tasks such as triage and backlog cleanup, but more pull requests alone do not show whether the investment pays off.
Engineering teams are spending from $150 to about $3,000 per developer per month on AI coding tools, but many cannot say what that spending has delivered, according to a report published by The Pragmatic Engineer on October 20. Interviews across 27 workplaces and vendor data show sharply different changes in output, prompting a focus on accepted work and quality rather than tool usage alone.
The report, by Gergely Orosz, combines interviews with engineering leaders and senior engineers at 27 workplaces with aggregate information from code-review platforms CodeRabbit and Gitar. Spending ranged from a $150 monthly cap at one publicly traded company with 1,500 engineers to about $3,000 per developer at Ix Technologies, a two-person European .NET agency. The source says company size did not predict spending levels.
Output measures do not tell a consistent story. CodeRabbit reported that developers on its platform created 113% more pull requests per week than a year earlier. Microsoft research covering 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. The source notes that vendor-platform data reflects teams already using an AI review tool, and Microsoft’s researchers cautioned that a merged pull request is not the same as delivered value.
Among the 27 workplaces, some reported flat pull-request frequency, while others described increases ranging from doubling to five- or tenfold. One company saw pull-request frequency stay flat as the requests became three times larger. The source also cites a Veracode test in which 44% of AI coding tasks produced a known vulnerability. These figures use different methods and settings; they do not establish a single expected productivity gain or a general rate of security failure across all AI-assisted development.
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 Outcome Measures
The differences matter as companies set 2027 budgets and some discounted subscriptions near their end. The source reports that several companies expect their bills to rise five- to tenfold if discounts expire and they do not adjust usage. GitHub also changed Copilot billing on June 1 to charge usage at each model’s API rate, making the cost of intensive, agent-based use more visible to budget owners.
A higher number of generated or merged changes can create more work if those changes require extensive review, rework or security fixes. The report argues that teams should compare spending with cost per accepted change, rework, change failure rate and whether completed stories outpace incoming work. Those measures connect tool costs to engineering outcomes more directly than token consumption or raw pull-request counts. They still require teams to define what counts as accepted, valuable work and to account for differences in project complexity.
The source describes clear uses in customer-issue triage, understanding unfamiliar code, rapid prototypes and clearing backlogs on consistent technology stacks. A staff engineer at a SaaS company said triage that could take a day sometimes became minutes of agent work plus up to an hour of human checking. That is an attributed workplace account, not a controlled estimate of typical savings, but it illustrates how task-level evidence can reveal benefits that an organization-wide output metric may miss.
As an affiliate, we earn on qualifying purchases.
Mixed Evidence Behind AI Gains
The report’s timing reflects changing costs as well as uncertainty about productivity. It says spending on API plans peaked in June and July and had roughly halved by the time of its reporting, a trend Orosz attributed to the arrival of Opus 5.5 and a shift toward open models. The account also says spending caps are common and can be bypassed, which can make a nominal budget limit a poor measure of actual use.
Independent research cited in the source complicates claims of consistent speedups. 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 in part because 30% to 50% of participants withheld tasks they did not want to do without AI. In a survey of 349 technical workers, the median respondent reported a 1.4-to-2-times change in the value of their work; METR also said participants in its 2025 study overestimated AI’s effect by 40 percentage points.
These results are not directly interchangeable: they cover different workers, tasks, periods and measurement methods. The source’s overall interpretation is that individual output may be rising, but how much of that increase becomes delivered value depends on the team and its work. Its examples suggest simple, uniform stacks may be easier to measure and automate than complex legacy systems.
“a question that few companies seem able to answer – at least for now”
— Gergely Orosz, summarizing the companies covered by The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
ROI Still Varies by Team
The report does not establish a single return on investment for AI coding tools. Its workplace interviews and vendor data are not a controlled comparison of all engineering teams, and the reported output measures differ. The source also says part of Orosz’s article is paywalled and that the figures cited come from its freely readable portion.
It remains unclear how much of the added output reported by some teams translates into higher-quality software, faster customer delivery or lower total engineering costs. The source gives examples of vulnerability findings and increased pull requests, but does not provide a consistent, comparable measure of review burden, defects or total cost across the 27 workplaces. The exact effects of expiring discounts and changing model choices on future bills are also not quantified for the companies discussed.
AI-powered developer productivity tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Track Results as Bills Change
Engineering leaders setting 2027 budgets will need to compare actual spending with work that teams accept and maintain, while tracking rework, failures and incoming demand. Companies facing subscription changes or usage-based charges will also need to monitor whether shifts in models and agent use change costs without weakening review and security practices.
The evidence described in the report is likely to become more useful as teams collect results over time using consistent definitions. For now, the source’s examples support measuring specific workflows rather than treating a broad increase in pull requests as proof of return. No single next industry-wide benchmark or follow-up result is identified in the material.
security vulnerability detection AI
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 report describes budgets ranging from a $150 monthly cap per developer at one large company to about $3,000 per developer per month at a small European agency. These are examples from 27 workplaces, not an industry-wide average.
Does the report show that AI coding increases productivity?
It reports gains in some settings, including more pull requests in CodeRabbit’s platform data and about 24% more merged pull requests among adopters in a Microsoft study. Other workplaces reported flat pull-request frequency, and the studies do not show that every increase became greater business value.
Which AI coding tasks appear to offer clearer gains?
The source identifies customer-issue triage, code archaeology, prototypes and backlog cleanup on simple, consistent technology stacks. The examples are company accounts and should not be treated as guaranteed results for other teams.
How should a company measure the payoff?
The source recommends tracking cost per accepted change, rework, change failure rate and whether completed stories exceed new incoming stories. Those measures can help connect tool spending to outcomes rather than usage alone.
What is still unknown about AI coding returns?
The available figures do not establish a typical return across companies or show consistently how added output affects software quality, delivery time and total costs. Results vary by task, team and measurement method.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
