DORA Finds AI Lifts Code Shipped 10%, Unveils ROI Framework for Verification and Pipeline Costs
Updated
Updated · O'Reilly Media · May 27
DORA Finds AI Lifts Code Shipped 10%, Unveils ROI Framework for Verification and Pipeline Costs
1 articles · Updated · O'Reilly Media · May 27
DORA’s 2025 data shows AI-assisted teams shipped about 10% more code to production, but the gain came with higher instability, including more rollbacks and fixes.
About 30% of survey respondents said they trust AI output little or not at all, while 46% trust it only somewhat, pushing verification work and pipeline upgrades into the real cost of AI adoption.
DORA’s new ROI framework and calculator are designed to model those hidden costs—learning time, verification overhead and delivery-pipeline changes—rather than promise a universal productivity number.
For 2026, DORA expects risks to shift toward agentic workflows, where heavier context switching could worsen burnout and add “cognitive debt” as developers lose shared understanding of systems.
The group argues the strongest response is platform investment: teams that pair AI with strong internal platforms, experimentation and organizational learning are more likely to turn output gains into durable performance.
As AI writes more code, are we creating developers who can no longer understand the complex systems they are building?
If AI can build anything, why is the new priority building platforms to constrain and control that same AI?