Insight Synthesis Prompt
Overview
Good decisions start with understanding what the data is really saying. That signal is usually buried across interviews, feedback, and reports in different shapes. Pulling it together by hand is slow, so the loudest input can drown out the pattern that matters.
Approach
I built a reusable prompt on an expert-and-apprentice model: the model drafts, I make the final call. The value came from iterating. My first version drifted between sections and smoothed over the data points that didn't fit, so I rewrote it to hold one consistent read, back every call with evidence, and surface the contradictions instead of hiding them. The lesson stuck: the more precisely you instruct, the better the output. Small changes in wording move the result a lot.
Solution
The prompt gives me a fast, rigorous first pass: clear themes, an evidence-backed read, and a recommendation I can trust because I can trace it. Because the domain and framework are editable, it flexes across questions. I've repurposed it more than once, from a learning needs analysis to making sense of ten real user interviews, reshaping the prompt each time for the new goal. I prototype in my model of choice, then adapt it to how our enterprise AI tool reasons, so it runs where people actually work.
Toolkit
-
Microsoft Copilot Cowork, Claude