Most "at-risk customer" lists are built from one signal: usage drop, NPS, support volume, or whatever your CSP measures. The problem is that no single signal catches the real risks. A customer can be quiet, on time with bills, and at zero NPS, and still be on their way out because their champion left.
Recon's claims pass synthesizes across signals. Your AI tool walks the claims and surfaces patterns you'd miss with any single tool.
The weekly prompt
Run this on Monday morning to see the at-risk list before the week starts:
Use Recon to give me this week's at-risk customer report.
1. Call list_accounts sorted by risk, limit 20.
2. For the top 10, briefly summarize what's driving the risk (active
claims, recent observations).
3. Group them: "Renewal in <60 days", "Champion gone quiet",
"Support escalation pattern", "Product usage drop", "Other".
4. For each, suggest one concrete action this week (call, email,
feature unlock offer, exec sponsor sync).
5. Flag any that are in the top 10 BOTH for risk AND in the top 20
by ARR. Those are the ones I should personally call.
What good output looks like
At-risk this week, top 10:
Renewal in <60 days (3)
- Acme Corp · Pro · renews June 12 · risk: 0.78. Three export escalations this month, REC-447 unresolved. → Action: exec sponsor sync. You should call this one.
- Quanta · Enterprise · renews June 4 · risk: 0.71. Last touchpoint was April. → Action: CS reach-out today, propose Q2 business review.
- Bolt · Pro · renews May 30 · risk: 0.66. NPS detractor + Workflows usage flat for 30 days. → Action: discount offer + check if their PM left.
Champion gone quiet (2)
- Vertex Labs · risk: 0.69. Power user T. Chen last login 11 days ago, was 30+ sessions/week. → Action: ping their other Workflows users; check LinkedIn for T. Chen role change.
- Mercia · risk: 0.62. Three named contacts, all on Slack but none active in 14 days. → Action: schedule QBR pull-forward.
[...]
Cross-flagged (top risk + top ARR):
- Acme Corp ($2.4k MRR, risk 0.78). Personal call this week.
Variations
For PM, "find candidates to interview for the new feature":
Use Recon to find customers who would be a good interview pool for
the new bulk export feature.
Criteria:
- Currently hitting limits in their export flow (search "export"
in their account observations)
- Power users (high product usage)
- Healthy enough that they'd take a call (memory_state: known or
healthy, not at_risk)
- Listed on the Pro or Enterprise plan
Return the top 8 with names of the right person to email at each.
For CS, "find Adoption Wins to celebrate":
Use Recon to find customers who had a big adoption moment in the
last 30 days. Specifically, ones that went from low to high usage of
a key feature.
Show me:
- The customer name
- Which feature they took off on
- The internal champion (who_reported on the account)
- A draft "great to see you using X heavily" email I can send
For founder, "stuff I should personally know before next board meeting":
Use Recon to give me the customer story for last 30 days:
- Top 3 wins
- Top 3 churns or at-risk
- Top 3 themes coming up across multiple customers (use themes)
- One surprising customer story
Keep it under 400 words. Include a one-sentence "what I'd do about
each theme" suggestion.