Agentic Software Engineering: Foundational Pillars and a Research Roadmap

Researchers propose Structured Agentic Software Engineering (SASE), a framework reimagining software development where AI agents autonomously pursue complex goals rather than simply generating code. The approach introduces two complementary environments—one for human oversight and one for agent execution—establishing a human-AI partnership model that demands fundamental changes to traditional software engineering processes, tools, and artifacts.

This academic research paper addresses a critical transition in software engineering where autonomous AI agents move beyond code generation into goal-oriented problem-solving. The distinction matters because traditional SE practices, designed around human developers, require restructuring when agents become primary actors capable of making architectural decisions, managing trade-offs, and even requesting human intervention strategically. The SASE framework acknowledges this shift by proposing dual modalities: SE for Humans (oversight, mentoring, decision-making) and SE for Agents (autonomous execution with callback mechanisms). This represents a fundamental architectural rethinking rather than incremental tooling improvement.

The research emerges from convergence of multiple trends: advances in large language models with planning capabilities, growing recognition that AI coding assistants require governance structures, and industry realization that coordination between human and machine workers demands new processes. The framework's emphasis on Merge-Readiness Packs and Consultation Request Packs suggests formalizing handoff protocols between agents and humans, transforming ad-hoc AI assistance into disciplined engineering practice.

For software development organizations and enterprise AI adoption, this roadmap legitimizes investment in agentic SE infrastructure. Rather than treating AI as a coding plugin, the framework positions it as a team member requiring oversight, communication protocols, and quality gates. The emphasis on trustworthiness and structured collaboration addresses enterprise concerns about autonomous systems.

The research community now faces implementation challenges: defining how agents escalate decisions, establishing quality metrics for agentic work, and redesigning educational curricula. Organizations monitoring AI's role in software delivery should track frameworks moving from theoretical to practical implementation.

• →Agentic SE requires reconceptualizing foundational SE pillars (actors, processes, tools, artifacts) across human and agent modalities.
• →Structured human-AI partnership with agent-initiated callbacks creates new engineering workflows beyond traditional code generation.
• →The framework addresses governance and trustworthiness as core challenges in autonomous software engineering systems.
• →Enterprise adoption demands formalized protocols like Merge-Readiness Packs for managing agent-produced outputs at scale.
• →Software engineering education must evolve to prepare developers for collaborative human-agent engineering practices.