DORA

DORA researches software delivery performance inside Google Cloud, along a publication line that runs 2020 ROI of DevOps Transformation, 2025 State of AI-assisted Software Development, then The ROI of AI-assisted Software Development, a 60-page framework published in April 2026 under CC BY-NC-SA 4.0 with an interactive calculator at dora.dev/ai/roi/calculator. Nathen Harvey leads the group.

The central thesis is "AI is an amplifier": AI magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones at the same time. Buying licenses changes nothing on its own. Injected into a system with manual testing, bureaucracy or fragmented data, AI accelerates technical debt. Hence the measurement rule the authors state: count the bottlenecks cleared, not the code written.

The report also puts a number on the trough before the gain. Its J-Curve of value realization covers learning curve, verification tax and pipeline adaptation, a cost to budget explicitly rather than panic over. The sample calculation (500 FTE, $176k salary, 12.5% time saved per developer) returns $11.6M of value against $8.4M of investment: 39% ROI, payback at 8 months. The normative line is blunt. Do not adopt a headcount-reduction strategy, reinvest the freed productivity, treat the Internal Developer Platform as a product.

That 12.5% is the floor practitioners keep arguing with. Dorra Bartaguiz opens her talk Quand la vélocité explose avec l'IA on DORA's 2024 and 2025 findings, over 80% of developers reporting higher individual productivity, then spends the rest of it on what the figure hides: code review is now the bottleneck.