Every began by mapping the investment team's research and memo-drafting workflows to identify where institutional knowledge was hardest to access under time pressure. Rather than building new infrastructure, the team operated entirely within the firm's existing ChatGPT Enterprise subscription — eliminating security and IT approval hurdles from the start. The core work was ingesting and structuring historical deal data: past investment memos, thesis documentation, and comparable transaction records were organized so they could be queried in natural language rather than searched manually through SharePoint folders. Prompt engineering created workflows that analysts could run at deal intake — pulling relevant precedents, surfacing comparable transactions, and generating a V1 investment memo already structured to match how the firm thinks. The entire build took place over two to four months, leveraging only existing subscriptions and document repositories. No engineers or custom development were required.