Layer 1 of the AI-First SDLC: An agent-driven product and designer layer for autonomous discovery, requirement refinement, experience design and backlog generation

The rising of Agentic AI has a very positive impact on software development and delivery, and organizations are improvising their processes very fast. In our earlier work, we introduced a thirteen-phase AI-First Software Development Lifecycle (SDLC) framework to make the process autonomous with mandatory and optional human-in-the-loop checkpoints, and in this paper, we are going to deep dive into the important entry point which we call the Product and Designer Layer.
Many organizations still follow a manual process to convert simple, high-level business ideas into clear requirements and usable designs in the form of stories and tasks, which leads to misunderstandings and also increases the chance of human error. Moreover, product managers or analysts spend days or weeks refining those high-level and ad-hoc epics before they become ready for engineering.
In this paper, we are going to explain the first module of the AI-First framework that we introduced in our earlier paper. This Product and Designer layer consists of four agents: Product Intent Agent, Research Agent, Design Agent, and Backlog Agent. These agents interact with each other and are managed by an orchestrator.
Starting from a natural-language epic, this layer enriches the initial idea with historical context and generates machine-readable design artifacts aligned with the organization’s design system, and finally produces an engineering-ready backlog. The intention is not to replace product owners or designers, but to reduce the translation effort between business, design, and engineering. By doing so, the time required to move from idea to backlog can be reduced from weeks to minutes, while keeping humans firmly in control of the key decisions.