The DORA AI Capabilities Model is a model introduced by Google’s DORA research programme in 2025 that identifies seven foundational capabilities — a clear and communicated AI stance, healthy data ecosystems, AI-accessible internal data, strong version control practices, working in small batches, user-centric focus, and quality internal platforms — which DORA’s research found to amplify the positive impact of AI adoption on outcomes such as individual effectiveness, code quality, team and product performance, throughput, reduced friction and organizational performance. It rests on DORA’s finding that AI’s primary role in software development is to amplify, magnifying both the strengths of high-performing organizations and the dysfunctions of struggling ones.
Usage
DORA presents the model as a companion to its long-standing DORA Core Model, which identifies the foundational drivers of software delivery and organizational performance; the AI model instead investigates the conditions under which AI assistance leads to better outcomes, and does not replace the Core Model. DORA calls it its first such model and says it will keep validating and refining it with further research. The model and practical guidance for each capability are set out in the report DORA AI Capabilities Model.
When It Applies
DORA frames the model as a guide for treating AI adoption as an organizational transformation rather than a tool rollout, and suggests using it to form hypotheses, run experiments and measure results in an organization’s own context, since every organization is different. It assumes teams will first establish a baseline — through DORA’s survey questions, the team profiles derived from its cluster analysis, and value stream mapping — before prioritizing capabilities. The report’s own caution about misapplication is the premise of the model: because AI amplifies whatever is already there, adopting AI tools without these capabilities can magnify dysfunction, and for teams lacking a user-centric focus it is associated with decreased team performance. The model’s standing is that of an empirical finding from one research programme: it is based on DORA’s 2025 survey of nearly 5,000 technology professionals and over 100 hours of qualitative data, which DORA describes as its inaugural version.