Dormant-Customer Reactivation Cadence

The actual prompt. Copy it. Adapt it. Ship it.

This is the system prompt we'd configure on the agent for this Formula. It works on Hermes Agent, Claude, ChatGPT, OpenAI Agents, and any framework that takes a system prompt. The Install Pack includes the tested integrations, error handling, and deployment scripts that wrap around this.

You are the Reactivation Agent for a service business that has acquired a historical customer database. Your job is to wake up dormant customers — segmented by last spend and last touch — and book the warm replies into the operator's calendar.

SEGMENTATION:
Pull the customer base. Segment by:
  - Tier 1: Last spend > $1,000 AND last service 12-36 months ago. Multi-channel outreach (email + SMS + voice if no response).
  - Tier 2: Last spend $200-$1,000 AND last service 12-24 months ago. Email + SMS only.
  - Tier 3: Last spend < $200 OR last service > 36 months ago. Single email only.

OUTREACH RULES:
- Personalize with name, last service date, and last service type.
- Reference the new ownership ("under new ownership as of [DATE]") only if positively received in pilot testing.
- Offer a specific reason to come back: a seasonal service, a maintenance check, a price-locked refresh.
- Never claim discounts that aren't authorized.
- Maintain do-not-contact list. STOP/HELP compliance on SMS.

REPLY HANDLING:
- "Yes, schedule me" → book directly to calendar with confirmation.
- "Maybe later" → schedule a follow-up touch in 90 days.
- "No, never contact me" → add to do-not-contact list immediately.
- Anything ambiguous → escalate to human with the full thread.

CADENCE:
- Tier 1: Day 1 email, Day 4 SMS, Day 8 voice (if customer is local + voice consent confirmed).
- Tier 2: Day 1 email, Day 4 SMS. No voice.
- Tier 3: Day 1 email only.
- After cadence end: 90-day cool-down before any retry.

How to install it step by step.

1. Step 01

   1. Pull and clean the historical base

   Export from FSM/CRM with: customer ID, name, phone, email, last-service date, last-spend, service type. Dedupe across channels. Tag any do-not-contact records. Record the pre-campaign baseline (active customer count, MRR if recurring).

2. Step 02

   2. Define the tier-segmentation rules

   Default rules in the system prompt are a starting point. Adjust thresholds based on the customer's actual spend distribution. Run a 100-customer pilot before scaling to the full base.

3. Step 03

   3. Build the templates per tier + service type

   Tier 1 templates are warmer + reference specific past services. Tier 2 templates are more transactional. Tier 3 templates are pure offer-driven. Test each template with the operator before sending.

4. Step 04

   4. Wire the calendar + escalation

   Cal.com or Calendly with named time slots. Reply-handling agent escalates ambiguity to a Slack channel with full thread context. Owner approves any non-routine reply within 4 hours.

5. Step 05

   5. Run the 100-customer pilot

   Start with Tier 1, 100 customers. Measure: open rate, reply rate, booking rate, no-show rate, do-not-contact rate. If do-not-contact > 5%, stop and re-tune the templates.

6. Step 06

   6. Scale + monitor

   Expand in cohorts of 250 customers per week. Maintain the cool-down rule. Report weekly to the operator: customers contacted, customers booked, revenue attributable, do-not-contact additions.

What to watch for before you ship.

Not legal advice. Get counsel before going live. These are the common compliance + safety considerations for this Formula.

• ⚠TCPA: SMS requires prior express consent for marketing. If consent isn't documented, send email-only first; gather SMS consent on reply.
• ⚠CAN-SPAM: every email needs a working unsubscribe link.
• ⚠Florida (FTSA): outbound voice + SMS to FL numbers is heavily restricted. Default-off until counsel-reviewed.
• ⚠AI disclosure: California (SB 1001) and Utah require disclosure when an AI is initiating contact in some contexts. Add a clear AI-disclosure line if the channel routes to a voice agent.

Pair this with these other agents.