Agentic Software Engineering: Foundational Pillars and a Research Roadmap

Abstract: Agentic Software Engineering (SE 3.0) represents a new era where intelligent agents are tasked not with simple code generation, but with achieving complex, goal-oriented SE objectives. To harness these new capabilities while ensuring trustworthiness, we must recognize a fundamental duality within the SE field in the Agentic SE era, comprising two symbiotic modalities: SE for Humans and SE for Agents. This duality demands a radical reimagining of the foundational pillars of SE (actors, processes, tools, and artifacts) which manifest differently across each modality. We propose two purpose-built workbenches to support this vision. The Agent Command Environment (ACE) serves as a command center where humans orchestrate and mentor agent teams, handling outputs such as Merge-Readiness Packs (MRPs) and Consultation Request Packs (CRPs). The Agent Execution Environment (AEE) is a digital workspace where agents perform tasks while invoking human expertise when facing ambiguity or complex trade-offs. This bi-directional partnership, which supports agent-initiated human callbacks and handovers, gives rise to new, structured engineering activities (i.e., processes) that redefine human-AI collaboration, elevating the practice from agentic coding to true agentic software engineering. This paper presents the Structured Agentic Software Engineering (SASE) vision, outlining several of the foundational pillars for the future of SE. The paper culminates in a research roadmap that identifies a few key challenges and opportunities while briefly discussing the resulting impact of this future on SE education. Our goal is not to offer a definitive solution, but to provide a conceptual scaffold with structured vocabulary to catalyze a community-wide dialogue, pushing the SE community to think beyond its classic, human-centric tenets toward a disciplined, scalable, and trustworthy agentic future.

• The paper introduces a dual modality framework that delineates structured collaboration between human 'agent coaches' and autonomous agents.
• The paper presents detailed methodologies, including BriefingEng, Agentic Loop Engineering, and related practices, to guide complex software tasks.
• The paper emphasizes practical implications in SE education and industry, advocating a shift from routine execution to strategic oversight.

Agentic Software Engineering: Foundational Pillars and a Research Roadmap

Agentic Software Engineering (SE 3.0) marks a significant evolution in the field of software engineering, transitioning from AI-Augmented development (SE 2.0) to a paradigm where intelligent agents are deployed not just for code generation but for achieving complex, goal-oriented SE objectives. This paper introduces a comprehensive framework that delineates this new era, driven by structured methodologies and dual operational modalities that focus on fostering collaboration between humans and agents. The interplay of SE for Humans and SE for Agents necessitates a profound reimagining of SE's foundational aspects—actors, processes, tools, and artifacts—in manifesting the practices required for leveraging intelligent agents effectively while ensuring trust and scalability.

Dual Modalities: SE for Humans and SE for Agents

The central thesis of the paper posits two symbiotic modalities in the Agentic SE era: SE for Humans (SE4H) and SE for Agents (SE4A). SE4H focuses on redefining the human's role as an "Agent Coach," whose responsibilities include high-level intent specification, strategic planning, and mentoring. Conversely, SE4A establishes a structured environment where agents execute tasks and participate in complex decision-making processes while maintaining transparency.

These modalities require specialized workbenches: the Agent Command Environment (ACE) for human oversight and orchestration, and the Agent Execution Environment (AEE) for task execution by agents. This dual framework supports agent-initiated human callbacks, contributing to a bi-directional partnership that elevates agentic software engineering from simple automation to structured collaboration.

Structured Agentic Software Engineering (SASE)

SASE is proposed as a vision defining structured engagement between humans and agents, characterized by carefully crafted processes and supportive environments. The SASE framework promotes a set of foundational activities that include:

1. Briefing Engineering (BriefingEng): Crafting comprehensive briefs that outline scope, objectives, strategic advice, and acceptance criteria in a structured format for agent consumption.
2. Agentic Loop Engineering (ALE): Defining and orchestrating workflows that guide agents through task execution, facilitating iterative refinement and convergence on solutions.
3. AI Teammate Mentorship Engineering (ATME): Codifying team norms and best practices into structured rules that guide agent behavior, transforming feedback from a transient activity into a durable, impactful component of the development process.
4. Agentic Guidance Engineering (AGE): Optimizing human involvement in the loop by ensuring efficient escalation and leveraging human expertise through structured consultation request packs (CRPs).
5. AI Teammate Lifecycle Engineering (ATLE): Enabling agents to become lifelong partners with memory and proactive maintenance capabilities, transitioning from stateless beings to informed collaborators.
6. AI Teammate Infrastructure Engineering (ATIE): Building the essential infrastructure that supports agent-centric practices, facilitating execution environments and toolchains optimized for agent performance.

Industrial Implications and Impact on SE Education

The implications of agentic SE extend far beyond technical implementation; they necessitate a re-evaluation of SE education and practice. As agents take on more autonomous roles within development processes, there's a pressing need to train future engineers in skills that prioritize strategic thinking, complex problem-solving, and effective collaboration with AI. Furthermore, as AI systems continue to reshape industries, the role of human engineers is elevated from implementers to orchestrators and stewards of AI teams, guiding them to achieve more sophisticated outcomes.

Conclusion

This paper offers a structured vision for Agentic Software Engineering, pushing the boundaries of traditional SE paradigms to embrace the potentials offered by intelligent agents. By establishing clear guidelines and structured methodologies, SASE aims to catalyze discussion and innovation within the SE community, advocating for a disciplined, scalable, and trustworthy approach to agent-enabled software development. Through these efforts, the SE field can build the foundations necessary for the next generation of intelligent, collaborative, and highly productive software engineering practices.