Weekly Customer Voice Digest for Product Teams

Product teams rarely have time to read every new G2 review, so recurring themes and emerging complaints get missed until they're already a pattern. Feedback also sits unstructured, with mixed praise, bugs, and feature asks in the same free-text fields, which makes it hard to act on. This blueprint gives a product team a consistent weekly read of what customers are actually saying, organized into the categories they care about, with trends flagged before they become problems.

1. The assistant asks for the product name and lookback window (default: last 7 days).
2. It queries the G2 MCP for reviews in that window, plus the prior equivalent period for comparison, pulling full review text for both.
3. If the window is too sparse for a meaningful trend read, it automatically widens (7 → 14 → 30 days) and says so.
4. It sorts qualitative feedback into three buckets: feature requests & gaps, bugs & friction, and what's working well along with a short representative quote per theme.
5. It compares the current window to the prior one: change in average rating, themes that are new/growing/fading, and a watch list of emerging issues.
6. It assembles a dated "Weekly Customer Voice Digest," then saves it to a connected document (Google Docs or Confluence), or outputs it in chat if no doc connector is available. Other preferred output formats can be requested via the LLM.

A product team gets a standing, low-effort pulse on customer voice that is structured the way they think (bugs, requests, wins) rather than as a raw review dump. Because it compares against the prior period every run, it surfaces emerging themes (e.g., a spike in a specific complaint) early, turning G2 reviews from a place you occasionally check into a weekly signal you can act on.

• Uses G2's Customer Voice / Verified Reviews data via the G2 MCP; access depends on your G2 subscription.
• Needs two connectors working together: the G2 MCP reads the review data, and a separate document connector (Google Docs or Confluence) writes the digest.
• Pulls full review text for both the current and prior period so trend comparison is theme-level, not just a rating change. This costs more tokens than a ratings-only pull, but produces a far more useful trend read.

• Effectiveness depends on review volume. Low-traffic products may have too few reviews in a 7-day window; the prompt widens the window automatically but will flag trend findings as low-confidence if even 30 days is sparse.
• Competitor reviews are intentionally out of scope — this blueprint analyzes a single product you have access to, to keep results focused and actionable.
• It reads and analyzes reviews only; it does not respond to reviews or edit your G2 profile (the G2 MCP does not currently have write capability).