Surface churn risk from customer communications with Gainsight's Risk Analyst
When customer emails, calls or tickets suggest an account might leave, this agent studies similar past cases and suggests specific next steps. Customer success teams get a plan, not just a risk warning.
Paste this into ChatGPT, Claude or your agent tool. See the prompt
- CS / Expansion
- beginner
- CS
- Category
- CS / Expansion
- Difficulty
- Beginner
- Est. cost
- Not stated by source
Works with
Problem
Churn signals hide in emails, calls and tickets that no CSM has time to read in full. A one-paragraph risk summary gives no guidance on what to do next.
Setup
When Staircase AI detects a churn-risk event in customer communications, Gainsight's Risk Analyst runs several sub-agents: one builds an experience library from six months of past churn-risk events and their outcomes, one scans current emails, calls, tickets and meetings, and others synthesize a risk score, trajectory and reasons and generate a playbook of recommended actions and stakeholders. The synopsis refreshes incrementally about every eight hours while the risk is active.
Prompt
Starter prompt by The GTM Agent Weekly, written from the setup described on this page. It is not the source's own prompt; see the source link below for their write-up.
# ROLE
You are a churn risk analyst. When a risk event appears in customer communications, you
assess it against past cases and recommend actions.
# INPUTS
- {{RISK_EVENT}}: what was detected, where (email, call, ticket, meeting) and when.
- {{RECENT_COMMS}}: the last few months of emails, call notes, tickets and meetings with the account.
- {{PAST_CASES}}: earlier churn-risk events at other accounts, what was done and the outcome.
- {{STAKEHOLDERS}}: known contacts and their roles.
# TASK
1. Summarize the risk in two lines, with the quote that triggered it.
2. Scan {{RECENT_COMMS}} for supporting or contradicting signals.
3. Compare with {{PAST_CASES}}: which past cases looked similar and what worked.
4. Give a risk level (low, medium, high), a trajectory (improving, stable, worsening) and the reasons.
5. Recommend a short playbook: actions, who should own each and which stakeholders to involve.
# OUTPUT FORMAT
RISK: level, trajectory. WHY: bullets with dated evidence.
SIMILAR CASES: what happened. PLAYBOOK: action, owner, stakeholder.
# GUARDRAILS
- Every reason cites a dated communication. No speculation about people's motives.
- Re-run as new communications arrive while the risk is open; note what changed since last time. Replace every {{PLACEHOLDER}} before running. Works with any capable LLM and runner.
Tools
Steps
- Connect the communication sources Staircase AI scans: email, calls, tickets and meetings.
- Add the Risk Analyst insight card to the account page.
- Review the risk synopsis: score, trajectory and reasons.
- Work the recommended playbook steps and stakeholder guidance.
- Feed outcomes back so the experience library learns which actions worked.
Results
Vendor documentation; no outcome data.
What goes wrong
Gotchas
- Notifications are suppressed after the first alert per account and event type for a long window, so later escalations may not ping anyone.
- Recommendations depend on past churn events; a young customer base gives the playbook little to learn from (editor note).
- A risk score without a CSM's validation can trigger the wrong save play (editor note).
Get 5 like this every week, free.
Real agent setups with sources and failure modes, in one short weekly email.
Free. One email a week. Unsubscribe anytime.
Source and credit
Gainsight documentation · vendor
Source dated 2026-10-11 · last reviewed 2026-10-11. Summarized in our own words; see the source for details.