Paper

Software Delivery as a Service
How AI could turn software delivery into an on-demand service

Renan Deves and Marcelo Scheidt, Independent Researchers. Published on Zenodo on 22 September 2026 as a preprint, version v1.

Read the paper on Zenodo

Abstract

AI raises a strategic question for software companies: if understanding, designing, building, and validating software require less effort, what should customers be able to buy? We propose Software Delivery as a Service, an operational, economic, and commercial model in which customers purchase useful stages of delivery and decide how far to continue. A request can begin with discovery, become requirements and a prototype, acquire an implementation price informed by architecture and technical investigation, and proceed to development, testing, and deployment when authorized. The provider maintains the engineering capability to run this process across multiple demands. Our central hypothesis is that AI can reduce the cost of learning, producing, and coordinating enough to make this relationship viable for a wider range of work. Progressive purchasing, shared specialist pools, and delegated decisions follow from that change. We explain the model through a familiar development pipeline, connect it to established engineering and economic ideas, and identify the conditions under which it could become a sustainable service.

How to cite

The Zenodo record and its DOI are the citable reference.

Deves, R., & Scheidt, M. (2026). Software Delivery as a Service (Version v1) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.22904797

@misc{deves2026software,
  author    = {Deves, Renan and Scheidt, Marcelo},
  title     = {Software Delivery as a Service},
  year      = {2026},
  month     = sep,
  version   = {v1},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.22904797},
  url       = {https://doi.org/10.5281/zenodo.22904797}
}

About this publication

The paper is a preprint. It has not been through peer review, and it does not report an operating implementation or measured results: it defines the model, explains why AI could make it viable, and states the conditions that would have to hold in practice.

The paper is licensed under Creative Commons Attribution 4.0 International.

The paper is the stable public statement of the model. This site explains the same model in more depth: the model follows a single demand from request to decision, and the glossary defines each term with a link to its explanation. The paper remains the formulation to cite.