Automate a weekly roundup of customer feature requests

Customers ask for things in every channel: a support reply, a sales call, a chat message. Each ask looks small alone. Together, five people asking for the same export is a signal. Without a habit of collecting them, the loudest or most recent request wins, not the most common one. Dedicated tools exist for this, and Linear's Customer Requests docs describe flagging a request as important for a given customer, which shows why who asked matters as much as how many.

You add the job and connect read access to your support inbox or chat through a remote MCP server (Beta) you set up. Each week a scheduled agent lists conversations from the previous seven days; Intercom, for example, documents a list conversations endpoint with cursor pagination. It picks out messages that ask for something the product does not do, groups them into themes, and counts distinct customers per theme rather than messages. Each theme gets one quote and a link to the conversation. The agent reads and reports. It does not reply, tag or change anything in your inbox.

The CPO agent, as department head, reviews the roundup before it goes out: is each request real rather than a bug report, are counts by customer, does each theme have a quote that supports it. Then you approve, edit or skip it, and nothing is sent until you do. Read an early roundup next to your inbox to see what it missed or merged wrongly, then adjust the brief.

A count is evidence, not a verdict. A request from one important customer may outweigh five from trial users, so keep the customer column. Compare the roundup with last week's to see which themes persist, since a theme that returns three weeks running deserves a decision: build it, schedule it, or tell customers it is not planned.