18 Evaluation The evaluation phase outlines how an enterprise assesses which areas of the business are most ready for Gen-AI integration, helping to optimize further investments. Evaluation breaks down into the following steps: 1) Defining the North Star 2) Investigation, and 3) Triaging Tasks. We explain each of these phases below. Defining the North Star As a first step in any Gen-AI integration process, enterprises need to identify a desired end-state and its implied long-term impact on the organization. While this impact will be more profound and unpredictable than an initial Gen-AI strategy can predict, having a waypoint on where and how the organization wants to adopt Gen-AI tools, as well as why, is critical to getting started, as this sets the initial direction for investigative efforts in the evaluation phase. Organizations should at least incorporate the following core elements: i) defining the specific need and rationale for adoption, ii) defining how the organization believes adoption will take place in the near term (2-3 years), mid- term (3-5 years), and long-term (5-7 years), and iii) defining the full scope of investigation within the enterprise, and developing hypotheses on specific areas where Gen-AI adoption could have the strongest impact. These three elements—rationale, timeline, and location—form the basis of a Gen-AI adoption strategy and should be developed in a foundational ‘North Star’ strategy document as a starting point. As the organization begins the integration process, it should expect that the initial strategy will change as more information is collected. Therefore, it is critical throughout the process for leadership to revisit the document and assess which areas of their original hypothesis have been validated, expanded, or invalidated. Conducting regular reviews of the North Star against research findings, and formally discussing changes to the strategy between leadership, management, and staff can help prevent pitfalls such as scope-creep and rushed implementation. Investigation By following a clearly defined adoption strategy, the enterprise can target the initial areas it believes Gen-AI integration will yield benefits and start mapping relative processes. Though process mapping is a standard procedure in any software implementation, Gen-AI tools will require more nuanced forms of analysis given that the technology has more advanced cognitive abilities than many other software applications. In fact, while mapping out more conventional steps in software usage is still important, Gen-AI integration will require mapping additional steps that may be cognitive in nature, where users previously had no interaction with a physical system until after the cognitive steps are completed. Large Language Models, for example, can quickly synthesize documentation and structure logical outputs that reflect it. The specific act of reviewing and synthesizing documentation, and formulating a response all occurs cognitively, which traditional process mapping may not adequately capture. Gen-AI process mapping will need to identify previously overlooked areas that involve steps that precede physical interaction with a physical system. For exemplary purposes, and to demonstrate the practical effectiveness of our analytical framework, our study incorporates an existing process in the Department of Defense, involving end-to-end flight procedures for pilots within the Naval Aviation Community. In Figures 2.1-2.5 below, we provide a standard process map of these procedures and conduct an evaluation on how Gen-AI tools could support pilots at each step. Generative AI Adoption in the US Military