Collaborator or Assistant? How AI Coding Agents Partition Work Across Pull Request Lifecycles

TL;DR

This paper analyzes how AI coding tools are integrated into pull request workflows, focusing on their roles in initiating, overseeing, and executing merges, and introduces a taxonomy and data resources for further research.

Contribution

It characterizes AI tools along a spectrum from collaborator to assistant, provides a taxonomy and state machines, and offers a dataset for studying automation and oversight in PR processes.

Findings

Collaborator tools initiate over 96% of PRs, but humans retain merge authority.

AI tools mainly support bounded tasks within human-led workflows.

Logs record executor identity but not decision-makers in automated merges.

Abstract

When AI coding agents open branches and submit pull requests (PRs), two questions co-determine oversight design: who starts the work (operational agency) and who authorizes its completion (merge governance). We characterize tools along a Collaborator-Assistant spectrum in how they redistribute initiative, oversight, and endorsement, while merge governance remains predominantly human across five tools (OpenAI, Copilot, Devin, Cursor, Claude Code). We analyze 29,585 PR lifecycles using an Initiator x Approver taxonomy with six interaction scenarios; lifecycle reconstruction supplies the how behind those roles. Collaborator tools (Cursor, Devin, Copilot) concentrate operational initiative in agents that open and carry PR work forward, with humans retaining review and endorsement on the path to merge; Assistant tools (OpenAI, Claude) leave task direction primarily with humans and supply bounded support within human-led workflows. Across the spectrum, agency and governance decouple: Collaborator workflows are >=96% agent initiated, yet terminal merge authority remains almost exclusively human, with agent-classified approvers confined to a small fraction of PRs. Where automation executes a merge, logs record the executor but not the decision-maker, marking a boundary of observation. We contribute the taxonomy, per-tool state machines, and a replication package for research on automation, oversight, and governance in PR workflows.