27 If implemented during a tool acquisition, as in Figure 8 above, RVA can serve as one possible framework (among existing procedures) to support the ‘Development’ portion of the SWP. Further, we clarify that leveraging RVA during the ‘Development’ phase assumes that this portion of the SWP is focused on tool integration—not the creation of the models themselves. We recognize that the creation of foundational models entails extensive capital investment and R&D that vendors likely will have completed by the time of DOD acquisition. Whether conducting tool-agnostic Gen-AI integration assessments or integrating a specific tool, RVA can serve as a useful framework to document, refine, and execute development cycles to narrow down use-cases that can materially benefit from Generative AI tools. We recommend documenting RVA outcomes in secured data networks accessible to other DOD teams with similar processes, needs, and clearance levels to expedite T&E. Treating RVA as an iterative process, maintaining documentation, and publishing it on highly secured networks can help accelerate the discovery and adoption of the most effective Gen-AI integration patterns across the DOD. Architecture involves 1) Defining the Future State; 2) Assessing Dependencies; and 3) Comparing Tools. We discuss each of these steps below: Architecture Defining the Future State The first stage in Architecture is translating the current-state processes into a future state, incorporating Gen-AI tools. This should occur in close coordination with stakeholders, and encompassing both future users as well as enterprise leadership. Analysts will need to first determine the role of Gen-AI integration for each step. Gen-AI tools may serve several roles for any given step, including no involvement, collaborative involvement, or automation. Understanding these roles is critical to identify how and where the workforce will interact with tools in each stage of a workflow process. Once a tool’s role is defined, analysts must identify the organization’s requirements. This includes envisioning user interactions, specifying prior actions the tools will support in the future, and outlining the necessary capabilities to do so. Once the future-state processes are sketched and the specific role of a Gen-AI tool at each step is defined, enterprises must determine if the organization has the capacity to integrate the technology in its current state. Assessing Technology Dependencies and Cost After mapping the future-state, analysts must identify if the chosen area of the organization is equipped with a sufficient technology stack to integrate specific Gen-AI tools. Conducting the future-state analysis allows teams to understand tool requirements, and determine which one is most appropriate (e.g., Language Model, Multi-Modal Model, Reasoning Model, Agent, or another form of Gen-AI). This also informs teams on how they will interact with the organization’s core technology infrastructure, which includes multiple layers—outlined in Figure 9 below. One critical consideration is assessing the organization’s current digital storage infrastructure, including existing Cloud providers, to determine their compatibility with the given model (e.g., running OpenAI GPT models on Microsoft Azure), as well as whether the security approval level (FedRAMP and IL levels) that the given Cloud provider holds is sufficient for the data involved in the anticipated use-cases. Considering that Agentic AI and RAG will need to navigate potentially complex webs of GUIs and data repositories to function, understanding the complexity and the number of enterprise resource systems is also critical. Another major consideration is the underlying infrastructure, including the source of the compute power to fine-tune models and run inferences with them. Across these areas, identifying where technical debt or technological obsolescence has accrued in the technology stack is critical to assess the organization’s readiness to integrate Gen-AI given its heavy reliance on enabling Generative AI Adoption in the US Military