Build a workflow that shows sales and CS teams what prospects and customers are talking about, broken down by country — a live world map of themes discussed on recent sales calls, designed to live on an office screen so people can compare what's hot in different regions at a glance. Trigger the data pipeline on an hourly cron plus an on-demand refresh route. It should: Pull recently completed external calls and their transcripts from your call-recording/conversation-intelligence tool (I used Gong — or whatever your team uses), skipping calls already processed and forwarding only new transcripts. Use an AI model (I used Anthropic's Claude — or your preferred LLM) to extract normalized discussion themes from each transcript and infer the customer's country. Append one record per call to a persistent named volume. Expose a data endpoint that aggregates the last rolling 24 hours of records per country and returns per-country call counts and top themes. Render a webpage dashboard: a world map with countries shaded by call volume, top themes labelled per country, and a click-through detail panel. The page should fetch its data client-side so it never blocks on the pipeline. Gate the webpage, data API, and refresh routes to authenticated tenant members. Keep everything read-only against external systems — the only write is appending records to the volume. When the volume is empty, have the data endpoint serve a built-in dummy per-country dataset so the map renders before any live data exists, with real records automatically taking over once present. Tools used: Gong, Anthropic Claude