DORA AI Capabilities Model - Review of the Report (Category Productivity)

I read it this weekend. report about this companion model report 2025 year, which is called "State of AI-assisted Software Development" The main point of the DORA AI Capabilities Model is that AI itself does not guarantee improvement, but it strengthens the already existing sociotechnical system – good practices reinforce, bad too. The model 7 Capabilities that increase the chance that AI will have a noticeable positive effect. I would like to note separately what I have already told you. How Dora reports are collected, How to announce AI Capabilities and What happened to Dora 2025 - State of AI-assisted Software Development

But if you go back to the new report, here are the great seven capabilities and ~~goalkeeper~~ AI 1. Clear and communicated AI stance The organization has a clear and communicated position on the use of AI: what is allowed, what is expected, what tools can be used. 2. Healthy data ecosystems Internal data is high-quality, accessible and not disparate. 3. AI-accessible internal data AI tools can securely obtain context from internal codebases, wikis, systems, and documents. 4. Strong version control practices Mature version control practices, including frequent commits and rollback. 5. Working in small batches Small changes, small releases, short feedback loop. 6. User-centric focus The team keeps the user’s value at the center. 7. Quality internal platforms A high-quality internal platform helps to scale the effect of AI at the organization level.

Researchers at DORA found the following links between these families: capabilities and interested outcomes 1. Clear and communicated AI stance ~ individual effectiveness and organizational performance 2. Healthy data ecosystems ~ organizational performance 3. AI-accessible internal data ~ individual effectiveness and code quality 4. Strong version control practices ~ individual effectiveness and team performance 5. Working in small batches ~ product performance downsizing friction 6. User-centric focus ~ team performanceBut without that focus, AI can even worsen. team performace (||The team will move faster in the wrong direction.||) 7. Quality internal platforms ~ organizational performance

It should be noted that these new capabilities are directly related to the classic DORA Core Model - new capabilities complement the core model. - Core Model This is the more conservative core of DORA: capabilities, metrics and outcomes, which have been repeatedly confirmed in studies over the years. - AI Capabilities Model It’s an add-on on top of Core that answers another question: under what conditions does AI-assisted development really improve results? DORA emphasizes that many AI-capabilities are the same basic engineering and organizational abilities that were previously associated with high performance.

It is interesting to note that 7 The factors did not quite come out of my head - according to the description of DORA, the process was as follows: At first, the researchers formed a broad list of capability hypotheses that could influence the success of AI-assisted development, relying on AI-assisted development. 78 In-depth interviews, opinions of domain experts and past DORA studies

• Then through prioritization, they selected 15 candidates in capabilities for inclusion in the survey Then in the quantitative analysis, they left the ones that found significant evidence of interaction with AI use. In general, the final model included those 7 Capabilities that statistically enhanced the effect of AI adoption on significant outcomes

#DevEx #Metrics #DevOps #Engineering #Software #Management #Leadership