Mark Fedeli’s Post

Jane Overslaugh Rathbun continues to raise the bar in communicating how the DON can lead in digital delivery. Her notion of fusion teams with functional and technical experts side by side is spot on. Since it connects directly to frontiers we are pushing at Seekr Technologies in #GenAI application development, I wanted to amplify her point with a GenAI insight I've picked up recently that pokes at common biases about gaps b/w functional and technical knowledge work. Yes, it is an argument for more GenAI (like ours) to support fusion teams. My deeper point is that knowledge work itself works best when it follows key GenAI principles about document and team selection. The insight is subtle: GenAI generates question and answer pairs to critique itself, and both functional and technical experts need them to be robust and diverse enough to make the most of their analytical work products. The part that took me awhile to grasp is that since GenAI optimizes for outcomes, we can start with a specific end product in mind and reverse engineer the technical and functional elements that feed into it. From there, experts can use them to generate more robust and diverse Q&A pairs. The key is specificity. The more precise the better (if I'm not mistaken). To keep it interesting and tied to Ms. Rathbun's post, let's say the end product is a leadership brief summarizing a fusion team's review of sources sought proposals to this question from government to industry: "How can a GenAI agent explainably automate a vendor vetting workflow prompted by Garter Magic Quadrant (MQ) report findings?" Let's say leadership will use this brief to weigh whether adding agents to market research is useful now or too risky until more research is done. Let's also assume there is pressure for the agentic approach, given criticism (of course not in real life :-) that manual vendor analysis workflows are slow and error prone, limiting innovation. Let's also assume that vendor selection is susceptible to groupthink, given limited staff time to weigh all pros, cons, alternatives, unknowns. Tendency is to follow loudest or smartest-sounding voice (a bias I'm poking). Now, up for consideration is an agentic workflow using an LLM: it starts with Gartner MQ, pulls in internal engineering reviews, guidance docs, etc. and builds functional and technical summaries from source documents. Before long, it' clear that these workflows deliver compelling, accurate technical and functional insights through a chatbot with source documents. In that app, GenAI SMEs could pipeline functional and technical questions into an LLM optimized for these Q&A pairs. Back to the core insight: The more precise and diverse the fusion team's questions, the better these pairs, resulting in a more vetted final analytical product and validation of agentic workflows underlying the end product. Rinse repeat for fusion teams to modernize w/ GenAI moving forward. #PokeTheBias #BeyondGroupthink