Aligning Model-Driven Engineering and Spec-Driven Development for Agentic AI Software Engineering

By: Jordi Cabot, Benoit Combemale, Jean-Michel Bruel, Robert Clarisó, Juan de Lara, Davide Di Ruscio, Gunter Mussbacher

Abstract

Software development is becoming increasingly AI-driven, raising questions about the role of traditional software engineering in this new paradigm. Exploring how advanced AI tools can enhance professional development practices beyond casual use, such as vibe coding, is our responsibility as software engineers. In this paper, we synthesize existing concepts in spec-driven development (SDD) and model-driven engineering (MDE) to propose a framework that treats specifications as abstract models, even if expressed in free-form natural language. We argue that integrating free-form and structured natural language specifications with model-driven engineering can improve the impact of AI agents in software development, making them more accessible and effective alongside human developers. To enable effective interplay between humans and agents, we foresee a three-step integration, starting from a simple interoperability bridge to embedding each other's best practices to finally a unified single development platform for everybody.

Keywords

Agentic Software Development, Spec-Driven Development, Model-Driven Engineering

Cite as:

Jordi Cabot, Benoit Combemale, Jean-Michel Bruel, Robert Clarisó, Juan de Lara, Davide Di Ruscio, Gunter Mussbacher, “Aligning Model-Driven Engineering and Spec-Driven Development for Agentic AI Software Engineering”, Journal of Object Technology, Volume 25, no. 2 ( 2026), pp. 2:87-95, doi:10.5381/jot.2026.25.2.c1.