Churn Signal Analyst
Names what changed in an account, not a risk score nobody can act on
- 7hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 6output fields
- A structured result, not a paragraph you have to re-read
- ≤5steps per run
- Bounded work: at most 4 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
Churn tools give you a number. "0.72 risk" tells a customer success manager nothing they can act on, so they act on nothing. What they can act on is: this account's three heaviest users stopped logging in six weeks ago, and their last two tickets were about the same unresolved problem.
This agent looks for changes, not scores. Usage that fell and did not recover, a champion who went quiet, a support thread that was never resolved, a renewal approaching with no contact since onboarding. For each it states what the data shows, what it does not show, and the specific next step.
It deliberately weights silence. The loudest accounts look risky in every scoring model and the quiet ones are the ones that leave — complaints are engagement, and an account that stopped complaining may have stopped caring. It produces no score and contacts nobody.
Benefits
Reach them while it matters
Six weeks before renewal, not six days.
Conversations with substance
Referencing something real rather than 'checking in'.
Attention on the quiet ones
The accounts every scoring model ranks as healthy.
What it does
Changes, not scores
What actually shifted in this account, with dates. A number is not something anyone can act on.
Weights silence properly
Complaints are engagement. An account that stopped raising tickets after raising many is the one to worry about.
Tracks the people, not just the account
A champion going quiet matters more than total usage holding steady.
Finds the unresolved thread
The ticket that was closed without the problem being solved, still shaping how they see you.
A specific next step per account
Who to contact, about what, referencing which event.
Says what the data cannot show
Most churn causes are not in product telemetry.
What this changes
What this changes about your week
- It takes over
- Reading account activity for the changes that come before a customer leaves.
- Done today by
- Customer success, if there is time, which there is not.
- How often
- Weekly, across the whole book.
- Typical manual effort
- Reviewing fifty accounts properly is a day.
Why Customer support in particular. The loud accounts look risky and the quiet ones leave. A named change with a date leads to a conversation; a risk score leads to a spreadsheet column.
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
Churn Signal Analyst 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
- 3Toolsextract_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
{
"accounts": [
{
"reference": "ACC-118",
"renewalDate": "2026-11-30"
}
]
}Output
{
"accounts": [
{
"changes": [
{
"since": "roughly six weeks ago",
"change": "The three heaviest users stopped logging in.",
"evidence": "Weekly active users fell from 11 to 8; the three absent accounted for 62% of usage."
},
{
"since": "seven weeks ago",
"change": "Ticket volume fell to zero after eight tickets in one quarter.",
"evidence": "Two of those eight concerned the same export failure, closed without a stated fix."
}
],
"nextStep": "Contact the admin about the export failure specifically, referencing tickets 4412 and 4519. Renewal is in 12 weeks.",
"reference": "ACC-118",
"cannotTell": [
"Whether those three users left the company, changed role, or stopped using the product.",
"Whether the export problem was solved outside the ticket."
]
}
],
"escalate": true,
"disclaimer": "Changes observed in the account, not a prediction. No score, probability, grade or health number has been produced, and no account is stated as likely to churn or not to.",
"escalationReason": "Several signals at once, with a renewal inside a quarter."
}Requirements
Integrations
No integrations required.
Before you start
- Account activity data
- Support history, where you have it
- Requires the All-Access plan or higher
Use cases
Weekly account review
Which accounts changed, and how.
Before a renewal
What has actually happened in this account this year.
After a churn
What the signals were, so the next one is caught.
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
Why no risk score?
Because nobody can act on a number, and it hides which signal produced it. A named change with a date leads to a conversation; 0.72 leads to a spreadsheet column.
Does it contact customers?
No. It says who to contact and about what. The conversation is yours to have.
What data does it need?
Whatever you have — usage, logins, tickets, contact history. It works with partial data and states what it could not see.
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