Voice of Customer Agent
Finds the complaint that shows up in every channel at once
- 8hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 9output fields
- A structured result, not a paragraph you have to re-read
- ≤7steps per run
- Bounded work: at most 5 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Every company collects far more feedback than it uses. Reviews, survey comments, support tickets, sales call notes, churn reasons, app store ratings — each source owned by a different team who reads only their own.
That fragmentation hides the only thing worth knowing. A complaint in one channel is noise. The same complaint in reviews, in churn reasons and in sales objections is a company priority, and no single team can see it because no single team sees all three.
The two standard approaches fail differently. Sentiment scoring produces a number that moves and never tells anyone what to do. Manual tagging is accurate and stops within a month because nobody has time.
What helps is specific and unglamorous: group everything by the problem being described, count how many independent sources raise it, weight by the revenue attached, and separate what people say from what they ask for. Customers describe symptoms and propose solutions — the symptom is the useful half, and taking the proposed solution literally is how products end up with a hundred features nobody wanted.
This agent does that reading across every source you have, and reports what would change if each theme were fixed. It replies to nothing and never quotes a customer's identifying detail.
Benefits
One priority list instead of six
Which is the only version a leadership team can act on.
Evidence for the roadmap
Themes with sources and revenue attached, not anecdotes.
The strengths protected
Because you know which ones people actually named.
What it does
Counts independent sources, not mentions
One theme raised in reviews, churn reasons and sales objections is a priority. Fifty mentions in one channel is a channel.
Separates the symptom from the request
Customers describe a problem and propose a fix. Taking the proposal literally is how you ship features nobody wanted.
Weights by revenue, not volume
The loudest segment is rarely the most valuable one, and volume alone always points at the loudest.
Tracks whether a theme is growing
A stable complaint and a spreading one need very different responses.
Finds what people praise
What you must not break in the redesign. Nobody ever compiles this and it is genuinely useful.
Quotes nobody
It reproduces no customer name, handle or identifying detail. Examples are paraphrased.
What this changes
What this changes about your week
- It takes over
- Reading reviews, surveys, tickets, churn reasons and call notes together.
- Done today by
- Five different teams, each reading only their own source.
- How often
- Quarterly at best, and only per channel.
- Typical manual effort
- Thousands of pieces of text that have to be grouped by problem rather than by origin.
Why Research in particular. A complaint in one channel is noise; the same one in three is a company priority — and no single team can see it. Sentiment scores move without telling anyone what to do.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Voice of Customer Agent has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolssearch_knowledge, extract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output, abridged.
Input
{
"period": "Q3 2026",
"feedback": []
}Output
{
"themes": [
{
"theme": "Setting up a second team is unclear and often done wrong",
"trend": "growing",
"examples": [
"A customer described creating three teams before understanding what a team was for."
],
"customers": 74,
"sourceList": [
"Reviews",
"NPS comments",
"Support tickets",
"Churn reasons",
"Sales call notes"
],
"revenueTouched": 1240000,
"whatTheyAskFor": "A duplicate-team button. Building that would not solve it — the request is a workaround for not understanding what a team is.",
"describedProblem": "People cannot tell whether a second team shares data with the first, so they either avoid creating one or create it wrongly and discover it weeks later.",
"independentSources": 5
}
],
"praised": [
"Speed of the reporting export is named unprompted in 41 pieces of feedback. Whatever the redesign does, it must not slow this down."
],
"escalate": true,
"disclaimer": "An analysis of feedback you already hold. Nothing has been replied to or published, no customer contacted, no identifying detail reproduced, and examples are paraphrased. Themes appearing in churn feedback are correlated with churn, not shown to have caused it.",
"channelBias": "Support tickets are 61% of the volume. Theme rankings are reported by independent sources to correct for it.",
"sourcesRead": [
"Reviews",
"NPS comments",
"Support tickets",
"Churn reasons",
"Sales call notes"
],
"thinEvidence": [
"Two themes rest on 3 customers each and are reported as thin rather than ranked."
],
"escalationReason": "One theme appears in all five sources at once, touches 1.24m of revenue, and is growing across three periods.",
"volumeVersusRevenue": "By volume, the top theme is mobile layout. By revenue it is eleventh. The teams raising it are on the smallest plans, and volume alone would have pointed the roadmap at it."
}Requirements
Integrations
No integrations required.
Before you start
- Feedback from whichever sources you have: reviews, survey comments, tickets, churn reasons, call notes
- Account value per customer, where you want revenue weighting
- Requires the All-Access plan or higher
Use cases
Quarterly roadmap input
What customers actually raised, ranked and evidenced.
After a bad review period
Whether it is one theme or many.
Churn analysis
What leavers said, against what current customers are saying.
Before a redesign
What to fix and what not to touch.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Is this sentiment analysis?
No. A sentiment score moves and tells nobody what to do. This reports named problems, how many independent sources raise each, and what revenue sits behind them.
Does it reply to reviews or customers?
No. It never posts, replies or contacts anyone. It is an analyst.
How does it handle personal data?
It reads feedback that may contain personal detail and reproduces none of it. Examples in the output are paraphrased, never quoted with identifying information.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent