Command
Use the Peec MCP to pull our brand sentiment report for the last 7 days broken down by date. Compare it to the 7-day period before that. If our sentiment score has dropped, use the chat_id dimension to identify the specific AI conversations driving the lowest scores. For each low-sentiment conversation, use get_chat to retrieve the conversation details and identify the topics and source URLs associated with the lowest scores. Format the output as a Slack alert message that includes: the overall sentiment change, which AI models show the worst drop, and the 3 conversations with the lowest scores, the topics they cover, and the source URLs retrieved in those chats. Schedule this as a daily Claude task to keep Slack alerts firing automatically without manual pulls.
By the time you notice a sentiment problem in your dashboard, it has already been shaping AI answers for days. The issue is that sentiment scores are averages. They smooth out the signal. A single negative source being retrieved across hundreds of conversations can drag your number down without ever becoming obvious in a weekly review.
This workflow catches the shift the moment it happens, identifies the specific AI conversations behind the decline, and surfaces the exact verbatim responses driving it. Your PR team gets the raw material: not a vague metric, but the actual phrases AI is using, which models are affected, and what sources appear in those conversations.
That is the difference between reacting to a problem and being able to address its root cause.
