Where do developers want AI to support their work? (rdel.substack.com)

A large mixed-methods study from Oregon State University and Microsoft surveying 860 developers mapped where engineers want AI to help—and where they don’t. The researchers found that developers’ cognitive appraisals of tasks (value, identity, accountability, demands) strongly predict AI adoption: AI is welcomed for high-demand, high-value, and high-accountability tasks but resisted for identity-laden work (mentoring, strategic decisions). Three task clusters emerged: “Build/Improve” (coding, testing, debugging) shows high AI demand; “Sustain/De-prioritize” (mentoring, AI-building) resists automation for relational and identity reasons; ops/coordination (DevOps, docs, stakeholder comms) has moderate demand but lags due to trust and capability gaps. Practically, developers want AI as a collaborator, not a replacement—favoring boilerplate reduction and augmentation with human oversight. Trust prerequisites are non-negotiable for systems-facing work: 95% cite Reliability & Safety, 89% Privacy & Security, 84% Transparency, 82% Goal Maintenance, and 79% Steerability. More experienced devs use AI less but demand stronger reliability and steerability; risk-tolerant or technophilic devs adopt AI more readily. For engineering leaders the takeaway is clear: map tasks to fit AI, prioritize tooling that reduces operational toil and augments core technical work, and implement suggest-only modes, reversible changes, traceable outputs, and explicit approval checkpoints to preserve agency and avoid deskilling.