Project Managers Navigate Shifting Responsibilities as LLM Automation Reshapes Development Workflows

A new study from software engineering researchers examines how the integration of large language models into development environments is reshaping the day-to-day work of project managers. Rather than fundamentally transforming management practices, the research shows that LLM-based automation is modifying existing activities — planning, estimation, coordination, monitoring, and governance — in ways that demand new skills and heightened oversight.

How Automation Is Changing Management Work#

Researchers conducted an exploratory case study with project managers at a large, multi-project software organization to understand their experience with LLM integration. The findings reveal that these tools are becoming woven into routine development work, rather than existing as separate tools. However, the effects are uneven: while some tasks become faster, the shift creates new challenges. Visibility into task execution has declined, making it harder for managers to track what's actually happening in their projects. At the same time, the workload for reviewing and validating LLM-generated code and artifacts has increased substantially.

The Hidden Cost: Review and Validation Burden#

One critical finding is that LLM-based automation, despite promises of productivity gains, has amplified the need for human review. Managers must now assess whether generated outputs meet quality standards, are free from errors, and align with project requirements. This validation function cannot be fully automated and falls largely to human team members, shifting the balance of work in unexpected ways.

Learning Curves and Managerial Judgment#

The research documents that managers are learning about LLM tools experientially and incrementally. Rather than formal training programs, learning centers on developing practical understanding of what these systems can and cannot do reliably, learning to critically assess their outputs, and establishing guardrails for responsible team use. These skills require judgment calls that can't be templated — managers must adapt as they go.

The study emphasizes that LLM integration is driving greater reliance on managerial coordination and decision-making, not less. As automation handles certain mechanics, human oversight of the broader system becomes more important. The research provides empirical grounding for how real software organizations are navigating this hybrid human-AI working model.