Build a weekly voice-of-customer report
Cluster the week's tickets, calls and Slack mentions into themes with quotes and trends.
A voice of customer report answers the question every product meeting circles: what are customers actually telling us? This workflow reads the week's support tickets, sales and success calls, and Slack mentions, clusters them into themes, and delivers a report with counts, trends and verbatim quotes. No survey, no manual tagging.
What does this workflow do?
Each week the agent gathers customer signal from everywhere it already lives: support conversations in Gmail and Slack, call transcripts from Fireflies or Granola, feature requests filed in Linear, and offhand customer quotes pasted into team channels.
It clusters the raw signal into themes ("confusion about the new billing page", "requests for API webhooks"), counts mentions, tracks how each theme moved against previous weeks, and attaches the two or three quotes that make the theme concrete. The report lands in Notion and a summary posts to Slack, so the loudest voice in the room stops being the only data point.
How does it work?
- The agent collects the week's signal. Tickets, transcripts, Linear requests and Slack mentions, respecting each source's permissions.
- It clusters mentions into themes. Deduplicated across channels, so a complaint that arrived by email and again on a call counts as one customer, two touches.
- It quantifies and trends. Mentions per theme, week-over-week movement, affected customer segments if your CRM is connected.
- It fills gaps by asking. If a theme is ambiguous ("three customers mentioned performance, but the tickets lack detail"), the agent can ask the support or success person who handled them for a sentence of context.
- It delivers the report. A structured page in Notion with themes, counts, trends, quotes and links to every underlying source, plus a five-line summary in Slack.
Why automate the VoC report?
Feedback is scattered by default. Support sees the bugs, sales hears the objections, success hears the churn risks, and nobody sees the whole picture. Assembling it by hand takes an afternoon, so it happens once a quarter, and recency bias fills the space between.
The workflow's memory keeps themes stable across weeks. "Billing confusion" stays one theme rather than becoming a new label every run, which makes the trend lines real. When product ships a fix, you watch the theme decay in the data.
Works with
Slack, Gmail, Fireflies, Granola, Linear, Notion. Runs on a weekly schedule; monthly and quarterly rollups reuse the same theme history.
Frequently asked questions
Which sources does the report draw from?
Support conversations in Gmail and Slack, call transcripts from Fireflies or Granola, feature requests in Linear, and customer mentions in team channels. You choose the sources when you create the workflow.
How are themes kept consistent between weeks?
The workflow remembers its theme taxonomy. New feedback maps to existing themes when it fits and creates a new theme only when it doesn't. You can rename or merge themes, and the memory keeps your edits.
Does it count the same customer twice?
No. Mentions are deduplicated across channels, so one customer raising the same issue by email and on a call is counted once, with both touches linked.
Can it break results down by customer segment?
Yes, when a CRM like Attio or HubSpot is connected. Themes can be split by plan, deal size or lifecycle stage, which shows whether a complaint comes from trial users or your biggest accounts.
How is this different from survey tools?
Surveys measure what you ask; this measures what customers say unprompted. The two work well together, but the unprompted signal arrives every week without response-rate problems.