Impact of Generative AI in Software Development

Alexander Polomodov and Igor Kurochkin move from DORA methodology to findings about generative AI. The report connects tool use with individual experience, organizational practices, and delivery. The discussion separates measured responses from causal explanations that remain hypotheses.

At the individual level, AI improves flow, job satisfaction, and perceived productivity, yet the modeled share of productive work unexpectedly falls. The vacuum hypothesis says freed time needs more meaningful work. Without new product hypotheses, small batches, and complexity controls, faster code generation may create legacy faster.

A value matrix captures different motivations: output, authorship, recognition, enjoyment, and learning. The same rollout therefore produces different adoption. Trust in AI changes both perceived benefit and behavior; safe access, clear rules, learning, and gentle incentives work better than mandatory use for everyone.

At organization level, DORA connects capabilities to delivery performance and orders adoption from utilization through feedback loops to impact and cost. Not every team can obtain the metrics, so platform support matters. Start with a goal and data, verify quality, and scale only demonstrated use cases.