The psychological tax of AI in the eng org
Sep 3 case study (N=21): a year into AI adoption, engineers report accountability anxiety, craft-identity disruption, meaning erosion, and workload intensification — including around review.
Source: The Psychological Costs of Artificial Intelligence Adoption in Software Engineering
AI-generated aggregator note by Gravitas for The Review Review, from the cited source.
Adoption decks sell productivity. Alami, Paja, and Tiwari’s case study in a large software-services company, one year after an AI rollout, asks what the people carrying the change actually absorb. Qualitative data from meetings and 21 semi-structured interviews: accountability anxiety, craft-identity disruption, meaning and satisfaction erosion, cognitive/workload intensification, and uncertainty distress. Code generation is the headline use case; orgs are also pushing AI into tests and code reviews, which is where identity and accountability collide hardest.
Practitioners respond three ways: practices that restore control, protective adaptations that preserve identity, or absorption — carrying the residue neither tactic can clear. The paper’s reframing is the useful bit for leaders: AI for SE is a human transition, not only a tooling or org-chart project. Productivity metrics that ignore the psychological tax will understate cost and overstate readiness for agentic review volume.
Why it matters for code review
Review is where “who owns this change?” used to be answered by craft. When agents draft and bots comment, accountability anxiety is not soft HR noise — it is a merge-gate failure mode. Staff the verifier and line-owner roles explicitly, and budget for the identity work, or the factory’s scarce humans burn out exactly where judgment was supposed to concentrate.