RescueSave playHuman review required

At-Risk Save Play

Turn a red account signal into a time-boxed recovery plan with owners, dates, and a stop rule, not a panic email.

A save play is not a promise to win, it is a disciplined bet with a stop rule. The plays that work name a root cause, staff two or three real moves against the capacity you actually have, and say in advance what would make you walk away.

The template

Copy it, swap the {{marked}} variables for your account, and run it.

prompt.txt

An account just went red and I need a save play, not a pep talk. Using only
what I give you, produce a recovery plan a manager would approve.

Account: {{account_name}}
ARR: {{arr}}
What went red (the trigger): {{red_signal}}
What I actually know about why: {{root_cause_known}}
What I can realistically commit (people, fixes, exec time): {{internal_capacity}}
Time to renewal: {{time_to_renewal}}

Produce:
1. A root-cause hypothesis, with a confidence label (low / medium / high).
2. The 2-3 highest-leverage moves. Each move names an owner and a date.
3. What "recovered" looks like in 30 days, in one measurable line.
4. Kill-criteria: the signal that means stop investing and let it go.

Rules:
- Every move has an owner and a date, or it is not a move, it is a wish.
- Do not promise a fix you cannot staff from the capacity I gave you.
- Separate the root cause from the symptom. Usage down is a symptom.
- Include the kill-criteria. A save with no stop rule becomes a money pit.

Variables

{{account_name}}: Account namee.g. Meridian Health
{{arr}}: Annual recurring revenuee.g. $140,000
{{red_signal}}: What triggered the red statuse.g. exec sponsor left; weekly active users down 30% in 6 weeks
{{root_cause_known}}: What you actually know about the causee.g. the sponsor drove the rollout; without her, two departments stalled
{{internal_capacity}}: What you can realistically commite.g. one eng fix, one exec intro, my time; no custom work
{{time_to_renewal}}: Time to renewale.g. 60 days

Worked example

A CSM opens Monday to a red account: the executive sponsor who drove the rollout has left, usage has fallen off a cliff, and renewal is 60 days out. She has goodwill for exactly one engineering fix and one executive introduction. She needs a plan she can defend, not a flurry of activity.

filled-prompt.txt

Account: Meridian Health
ARR: $140,000
What went red (the trigger): exec sponsor left; weekly active users down 30% in 6 weeks
What I actually know about why: the sponsor drove the rollout; without her, two departments stalled
What I can realistically commit (people, fixes, exec time): one eng fix, one exec intro, my time; no custom work
Time to renewal: 60 days

Good output vs bad output

GOODNames a root cause and labels its confidence, staffs three moves that fit the stated capacity, defines recovery in one measurable line, and says in advance what would make you walk away.

ROOT-CAUSE HYPOTHESIS (medium confidence): The value was tied to one person,
not built into the team's workflow. When the sponsor left, so did the reason
two departments logged in. The usage drop is the symptom; single-threading is
the cause.

MOVES:
1. Re-anchor the sponsor gap. Owner: me. By Friday. Get a warm intro to the
   sponsor's replacement or her manager. If none exists in 2 weeks, that is a
   kill signal (see below).
2. Ship the one fix that removes a live reason to leave. Owner: Eng (Dana),
   scoped call Thursday, fix by day 21. Spend the single fix on the blocker
   the stalled departments named, nothing else.
3. Rebuild a second thread. Owner: me. By day 14. Find one power user in each
   stalled department and make them the local reason to stay.

RECOVERED IN 30 DAYS = weekly active users back above 80% of the pre-drop
baseline in at least one stalled department, plus a named replacement sponsor
who has taken one meeting.

KILL-CRITERIA: no replacement sponsor identified in 2 weeks AND usage still
falling at day 21. If both hold, stop the fix, protect the eng goodwill for a
winnable account, and manage this to a soft landing.

BADLooks fine at a glance. Here is why it fails.

We can absolutely turn Meridian around. Here is the plan: schedule a QBR to
re-engage stakeholders, loop in our leadership to show commitment, build a
custom success roadmap, and deliver a value-realization deck. We will get
engineering to prioritize their requests and demonstrate the ROI they are
missing. With the right proactive outreach and a renewed focus on the
relationship, we can rebuild engagement and secure the renewal.

Over-promisingCommits engineering to "prioritize their requests" and a "custom success roadmap" when the stated capacity was one fix and no custom work; it promises what it cannot staff.
No next stepNot one action has an owner or a date, so nobody actually does anything Monday; it is a list of nouns, not a plan.
Generic filler"Proactive outreach and a renewed focus on the relationship" is the platitude a manager has read a hundred times; it does no recovery work.
Missing contextIgnores the 60-day clock and offers no kill-criteria, so a doomed save could burn the scarce eng goodwill for weeks.

Guardrails

When NOT to use this

Do not run a save play on an account that has already signed a cancellation; that is theater, and the team can smell it.
Not a substitute for the direct conversation with the customer about what actually went wrong; a plan built on a guess just fails faster.

Responsible use

The model cannot commit your colleagues' time. Confirm every named owner agreed before the plan leaves your desk.
A save play is a hypothesis under pressure. The account team owns the call to keep investing or let go, not the model.

A human must review the output before it reaches the customer.

Model + settings

Model: Claude Sonnet for most saves, Opus for a flagship account where the readout goes to leadership.
Temperature: 0.3 (low, you want a disciplined plan, not optimism)
Settings: Reject any output where a move lacks an owner or a date. That constraint is the whole value; an ownerless plan is a wish list.
Make it state kill-criteria every time, so a doomed save gets cut early instead of quietly draining your quarter.

Back to all prompts