DORA Metrics in the AI Era

Narrative

Financial and business coverage of DORA-style delivery metrics in the AI era clusters around three storylines: capital flooding into AI coding tools, contested productivity evidence, and enterprise caution about code quality. On capital flows, Bloomberg's reporting (relayed via TechCrunch and Crunchbase News) tracked Anysphere, maker of the Cursor editor, raising 900 million dollars at a 9.9 billion dollar valuation in June 2025, with Bloomberg describing the three-year-old firm as "the fastest growing startup ever" after its annualised revenue crossed 500 million dollars. Subsequent reporting citing Bloomberg and CNBC shows Cursor's annualised revenue run rate climbing past 1 billion dollars in November 2025 and past 2 billion by February 2026, with the company forecasting more than 6 billion dollars by end-2026, a trajectory investors are pricing even as the underlying productivity case remains disputed.

On the productivity question, Reuters technology correspondent Anna Tong reported on the METR randomised controlled trial, describing how AI tools "slowed down experienced software developers when they were working in codebases familiar to them, rather than supercharging their work". The METR study itself, published as an arXiv preprint by Joel Becker, Nate Rush, Elizabeth Barnes and David Rein, recruited 16 experienced open-source developers across 246 real tasks and found actual completion time rose 19 percent under AI assistance even though developers forecast a 24 percent speedup and, after finishing, still believed they had been sped up by 20 percent. METR's own February 2026 update flagged that its second-wave experiment produced an unreliable signal because too many developers refused to work without AI, biasing estimates downward, a methodological caveat that financial-press summaries of the "AI productivity paradox" (echoed by Moody's chief economist Mark Zandi's comments to Business Insider that AI's productivity lift will be gradual) have generally omitted.

DORA's own Google Cloud-sponsored research provides the enterprise-scale counterpart. The 2024 report found a 25 percent rise in AI adoption associated with roughly a 1.5 percent drop in delivery throughput and a 7.2 percent drop in delivery stability, prompting the DORA team's own reclassification of failed-deployment recovery time from a stability to a throughput measure and the addition of rework rate as a stability signal, as tracked by industry analyst Rachel Stephens (RedMonk). The 2025 DORA report, based on survey responses from nearly 5,000 technology professionals and over 100 hours of qualitative interviews, reversed the throughput sign to positive while instability persisted, and introduced the "AI Capabilities Model" describing AI as an amplifier that "magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones," per Google Cloud's own announcement.

Code-quality telemetry from GitClear, drawn from analysis of 211 million lines of changed code between 2020 and 2024, found that copy-pasted code exceeded refactored ("moved") code for the first time in 2024, that duplicated code blocks rose roughly eightfold that year, and that short-term code churn rose from roughly 3.1 percent in 2020 to 5.7 percent in 2024. Wall Street Journal reporting cited in trade press quoted MIT's Armando Solar-Lezama comparing AI-assisted coding to "a brand new credit card...going to allow us to accumulate technical debt," while the same coverage noted TD Bank telling the Journal it now screens for abstract coding and prompt-engineering skills as AI absorbs routine coding tasks. These vendor and press accounts remain claims requiring independent replication rather than settled findings, particularly where productivity percentages diverge by 30 to 50 points between vendor marketing, DORA's Bayesian survey modelling, and METR's controlled trial.