Can Large-Language Models Replace Humans in Agile Effort Estimation? Lessons from a Controlled Experiment
A controlled experiment on generative AI for agile effort estimation, examining estimation accuracy and support for planning.
Luka Pavlič’s talk in the Programi track at Evolucija.AI 2026 on AI throughout the software development lifecycle.
The talk presents the evolution of AI-supported software development, from traditional development assistants to Copilot and modern AI agents. Its focus is on how artificial intelligence participates throughout the development lifecycle: from eliciting and refining requirements, planning and programming to testing, documentation, deployment and maintenance. Particular attention is given to input quality, traceability, verification of results and the developer’s role in oversight and decision-making.
A controlled experiment on generative AI for agile effort estimation, examining estimation accuracy and support for planning.
A review and evaluation of intelligent code completion assistants, including functionality and developer usability.