Customer Insights Copilot
A running record of what customers keep saying, with the trend, the quotes, and the churn watch kept for you
The problem
The [Voice-of-Customer Miner](/toolbox/voice-of-customer-miner) reads a batch of feedback once and finds the themes. Then the batch is gone, next month's batch gets read in isolation, and nobody can say whether the delivery complaints are shrinking or growing, whether the quote you loved in March ever got permission, or which customer has now raised the same issue three times. Feedback in a small business is a stream, and reading it in puddles loses the one thing only a stream can show: direction. This skill keeps the stream.
The tool
Once installed, this skill turns your assistant into the keeper of everything customers tell you. Seed it with your last batch of feedback (run the Miner first if you like; its four sections paste straight in), and then feed it the signal as it arrives: a review, a support ticket, a survey answer, a line from a phone call, two lines each.
The standing loop is where the value lives. Every piece of feedback gets logged against a theme ledger with a running count per month, so you read "late delivery: 9 mentions in August, 4 in September" instead of "customers complain about delivery". Strong customer phrases go into a quote bank with a permission status (unasked, asked, approved), so marketing pulls from approved quotes only. Escalating language from the same customer lands on a churn watch list, with the Churn-Risk Detector rules applied consistently rather than by gut. And the monthly read tells you what a single batch never can: which themes are rising, which fix from last month worked, which quote to ask permission for this week, and the one thing to change next.
What's inside
- Seeding from a Miner output or any pile of raw feedback, messy is fine
- Two-line feedback logging mapped to a theme ledger with monthly counts
- A quote bank with permission status, so nothing unapproved reaches a website
- A churn watch list built from repeat and escalating signals per customer
- A monthly read: rising and falling themes, last month's fix checked, quote to request, one change
- A quarterly brief for whoever owns product, marketing, or the website
How to use it
- 1Click "Install skill" on this page and copy the install prompt.
- 2Paste it into your AI assistant (Claude, ChatGPT, anything that accepts pasted text).
- 3Seed it with your most recent feedback batch or Miner output, and answer its setup questions once.
- 4Log feedback as it arrives, two lines each: where it came from and what they said. That habit is the whole system.
- 5Monthly, read the report, act on the one change, and ask permission for the quote it picked.
Example
A meal-kit business seeds the skill with 25 tickets the Miner already read, then logs feedback through September. The ledger shows late-delivery mentions falling from 9 to 3 after the new courier started (the fix is confirmed, not assumed), while "portion size on the family plan" climbs from 4 to 7 and becomes the top rising theme. The quote bank holds "the only box my kids do not argue about", status approved after the owner asked, and it goes on the landing page the same week. The churn watch flags one customer whose third ticket in five weeks shifted from polite to "last box if it happens again"; the owner calls them. The monthly read closes with one change: relabel the family plan as "feeds 3" until the portion is fixed, because seven customers in a month cannot all be wrong.
Pro tip
Log the good feedback with the same discipline as the complaints. The rising theme in your praise is your next headline, and the skill can only show it rising if the five-star messages get logged too. Pair this with the [Feedback Survey Builder](/toolbox/feedback-survey-builder) once a quarter to fill the gaps the inbox does not cover on its own.
Related tools
Want this customized and automated for your business?
We take the tools in this toolbox and wire them into your business: your data, your brand voice, running on autopilot.
Talk to Vexlo