Post-Velocity Software Engineering: Assurance Saturation in AI-Enabled Socio-Technical Systems

Abstract

Automation and large language models are increasing the rate at which software changes can be produced in AI-enabled socio-technical systems. Yet responsible digital transformation is not only about generating code, models, or system modifications; it also requires understanding, validating, contesting, and approving changes before they can be trusted in operational, organizational, or public-interest settings. These assurance and governance activities depend on human attention, organizational processes, accountable oversight, audit evidence, and the participation of institutions and affected stakeholders, and they do not scale at the same pace as automated generation. We argue that software engineering is approaching a regime in which the rate of change can exceed the capacity of teams, institutions, and affected communities to review, contest, and govern it responsibly. We call this regime post-velocity software engineering. Beyond this threshold, increasing development speed may no longer improve outcomes and may instead weaken justified confidence, reduce oversight quality, and shift risk downstream into operations, users, communities, and public institutions. This article introduces post-velocity software engineering as a governance saturation problem in AI-enabled socio-technical systems. It describes signals that organizations may be approaching this threshold and outlines a research agenda for understanding how AI-accelerated development reshapes the balance between software production, assurance, accountability, contestability, and community-in-the-loop governance.