Spec-Graph Meta-framework for AI-Assisted Software Development: An Industry Case
1 — Tres Astronautas, Bogotá, Colombia
2 — Universidad de los Andes, Bogotá, Colombia
3 — Blend 360, Bogotá, Colombia
This is the preprint (author) version. It has been accepted for publication at ICSME 2026 but has not yet been through final copyediting or typesetting. The definitive version will appear in IEEE Xplore — please cite the published version once it is available.
AI-assisted software development increasingly relies on large language models to transform natural-language specifications into code. However, in most current workflows, specifications are treated as inline prompt content, limiting their structure, reuse, and traceability across development tasks. This paper introduces a meta-framework for spec-driven AI-assisted development in which specifications are represented as a typed, queryable graph consumed at runtime by software agents. The framework is organized around three principles: legibility, achieved through a typed hierarchy that distinguishes buildable artifacts from contextual constraints; addressability, enabled by a minimal set of edge types, substrate-agnostic graph materialization, and a cost-function-driven sequencing mechanism; and longevity, maintained through continuous curation of the specification substrate. We instantiate the framework in an industrial setting involving a multi-project software development organization with over fifty employees, using thirteen specification types, a multi-substrate architecture, and four cooperating AI agents. To explore the proposed framework’s behavior in practice we surveyed 19 practitioners in the organization after using the framework from one to five months.
- Spec-driven development
- AI-assisted software engineering
- knowledge graphs
- code-generation agents
- industrial study
D. Soto, D. González-Rueda, M. Linares-Vásquez, and A. Hurtado, “Spec-Graph Meta-framework for AI-Assisted Software Development: An Industry Case,” in Proc. 1st Int. Workshop on Agentic AI for Next-Generation Software Development (AgenticDev 2026), co-located with the IEEE/ACM Int. Conf. on Automated Software Engineering (ASE), Munich, Germany, Oct. 2026, to appear.
BibTeX
@inproceedings{soto2026specgraph,
author = {Soto, Daniel and
Gonz\'{a}lez-Rueda, Daniel and
Linares-V\'{a}squez, Mario and
Hurtado, Andr\'{e}s},
title = {{Spec-Graph} Meta-framework for {AI}-Assisted Software
Development: An Industry Case},
booktitle = {Proc. 1st Int. Workshop on Agentic AI for Next-Generation
Software Development (AgenticDev 2026), co-located with the
IEEE/ACM Int. Conf. on Automated Software Engineering (ASE)},
address = {Munich, Germany},
month = oct,
year = {2026},
note = {To appear}
}