AI
What happens when companies hand out AI coding tools: developers see gains but also risks
A qualitative study of a major telecom's coordinated AI-devtools rollout reveals that developers appreciate the productivity help but worry about reliability and face mismatched expectations from management.
A new case study from researchers examining how software developers actually experience AI-powered coding tools in real organizational settings found a more nuanced picture than the hype suggests. When a large Swedish telecommunications company rolled out AI development tools across three sites, developers reported appreciating the assistance—but also surfaced significant tensions between what management expected and what workers felt comfortable relying on.
What the research covered#
The study involved semi-structured interviews with 12 software professionals working across the telecom company's operations. Researchers used process coding and thematic analysis to understand how developers perceived the tools, then applied the extended Technology Acceptance Model (a framework used in technology adoption research) to interpret their findings.
Key findings on experience and concern#
The interviews confirmed what earlier surveys had suggested: developers do find AI-devtools useful in specific contexts, they do see productivity benefits in some workflows, and they encounter real frustrations with the tools' limitations. However, the in-depth accounts revealed something earlier metrics had largely missed—a significant gap between what management expected from the rollout and what developers believed was realistic or safe.
Developers weighted perceived risk heavily in their adoption decisions. This concern about fallibility and reliability isn't adequately captured by standard technology acceptance frameworks. The research team found that organizational expectations, the scale at which tools are deployed, and developers' own skill levels all shape how valuable and trustworthy an AI assistant actually becomes in practice.
Why this matters#
Most evidence about AI-devtools comes from tool evaluations, isolated productivity metrics, and surveys. Few studies document what happens in coordinated, organization-wide rollouts over time. Because companies are making significant budget commitments and changing working practices based on incomplete information, understanding the real-world developer perspective—especially the skepticism and safety concerns—matters for how these tools get implemented and governed.
