SMS Reactivation Agent

• CF Worker
• D1
• Twilio (SMS)
• Claude
• CF Cron Triggers

A real prospect. A real demo. A real outbound message — all before 5pm.

1. 09:00–09:30 Pick a high-LTV business with a stale customer DB.

   Hair salon, MedSpa, dental practice, fitness studio. Pick one where average customer LTV is $500+ and you can credibly say "you have hundreds of customers who haven't booked in 90 days."

2. 09:30–10:30 Get sample data (or scrape Mindbody/Boulevard if accessible).

   Best case: they share an export. Realistic: build the demo with synthetic data that matches the shape they'd have.

3. 10:30–11:30 Build the segmentation query.

   D1 query: customers where last_visit_at < (today - 60 days) AND has_phone = true AND opted_out = false AND total_visits >= 2. The 2-visit minimum filters one-time customers.

4. 11:30–12:30 Build the message generator.

   For each customer: pull last service + last technician + last visit date. Generate a one-line SMS in the owner's tone: "Hey [name], it's been a while since we did your [service] — we've got Tuesday + Friday open this week. — [Owner first name]".

5. 12:30–13:30 Wire Twilio + the cron.

   Cron: Mon-Wed-Fri at 11am ET. Per-cron: pull next 20 customers from queue, generate personalized message, send via Twilio. Throttle. Log every send.

6. 13:30–14:30 Build the reply handler.

   Twilio webhook → CF Worker. Claude reads inbound + customer history + original outbound. Three outcomes: (1) clear booking → reply with Cal.com link, (2) question → reply naturally, (3) ambiguous/angry/unsubscribe → escalate to human inbox.

7. 14:30–15:30 Build the opt-out + compliance flow.

   STOP/UNSUBSCRIBE/QUIT/CANCEL detected → immediate confirmation reply ("Got it, you're unsubscribed. Reply START to opt back in.") + flag in D1 + remove from all future queues. Required by TCPA.

8. 15:30–16:30 Send 5 test messages to yourself + 2 friends.

   Test the full flow: receive → see personalization → reply YES → get Cal.com link → reply something off-topic → see natural response → reply STOP → confirm opt-out.

9. 16:30–17:00 Send the cold pitch with sample messages.

   Email the owner. Attach 3 sample messages your system would send to 3 of THEIR actual customers (anonymized initials). Subject: "I generated outreach for 3 of your stale customers."

• Always sign with the owner's actual first name. Never "the team."

  Why: Customers respond to a name, not an org. "From the Bella Salon team" has 3-5x lower reply rate than "From Maria at Bella." Customers know it's automated; they engage anyway when it feels personally addressed.

• Reference the last service or visit explicitly.

  Why: "Hey, we miss you" is a coupon mailer. "Hey, it's been 9 weeks since your highlights with Carla" is a friend texting. The specificity is the entire reactivation lever.

• Tuesday or Wednesday at 11am ET. No exceptions.

  Why: SMS engagement peaks mid-morning, mid-week. Friday afternoon = ignored. Saturday = reads as desperate. Monday = needy. Tuesday/Wednesday 11am = intentional.

• TCPA opt-out compliance is non-negotiable. Build it first, not last.

  Why: STOP/UNSUBSCRIBE/QUIT/CANCEL must trigger immediate opt-out + confirmation reply within 2 minutes. Carrier penalties for non-compliance are $500-$1500 per message. No exceptions.

• Max 1 outbound + 3 reply turns then human handoff.

  Why: AI texting feels great for 4 turns and uncanny by turn 6. Cap the conversation. Handoff to a human-staffed inbox at turn 4 or any sign of complexity.

• A/B test message tone (warm / casual / urgent) per customer segment.

  Why: 60-day-stale responds best to warm. 180-day-stale responds best to direct ("we're running a slot Friday — last chance"). Loyal-but-busy responds best to casual. Track reply rate by tone × segment.

• The booking link is a single tap, not a form.

  Why: A reply of "yes" should land them in a Cal.com page already pre-filled with their name + last service. Two taps to book. Friction at booking kills the reactivation.

Drop these into Claude Code. Replace the [BRACKETED] fields with the prospect's details.

Generate a D1 SQL query for the SMS reactivation agent at [BUSINESS NAME].

Schema:
- customers: id, first_name, last_name, phone, opted_out, total_visits, last_visit_at, lifetime_value
- visits: id, customer_id, service, technician, visit_date, total
- outreach_log: id, customer_id, sent_at, message, response_at, response, outcome

Query requirements:
- Last visit between [N1] and [N2] days ago (warm 30-60d, mid 60-120d, cold 120-180d).
- total_visits >= 2.
- opted_out = false.
- No outreach sent in last [M] days (cooldown, default 30).
- Has non-null phone.
- Order by lifetime_value DESC.
- LIMIT 20.

Also output the cron-driver function: takes the query, sends each customer to message generator, sends via Twilio, logs to outreach_log. Throttle 1/sec.

Output: SQL + cron function in ESM JS.

Generate a personalized reactivation SMS for [BUSINESS NAME].

Customer:
- First name: [name]
- Last service: [service]
- Last technician/provider: [name]
- Days since last visit: [N]
- Total visits: [N]
- Lifetime value: $[N]

Business:
- Owner first name: [name]
- Tone: [warm / casual / direct]
- Available slots: [pull from Cal.com or static]
- Booking link: [Cal.com URL pre-filled]

Constraints:
- Single SMS, MAX 160 characters.
- Sign with owner first name + business first letter (e.g. "— Maria, B.")
- Reference specific last service AND last technician.
- Suggest a specific window ("Tuesday or Friday").
- No emojis. No "miss you." No "valued customer." No exclamation points.
- Include the Cal.com link only if 160 chars allow.

Output: full message text + tone (warm/casual/direct) + confidence (high/medium/low) on personalization data.

Handle this inbound SMS reply for [BUSINESS NAME].

History:
- Outbound message: [original SMS]
- Reply #[N] in thread (max 4 before human handoff).
- Customer name: [first name]
- Last service: [service]
- Booking link: [Cal.com URL]

Inbound: [reply text]

Classify and respond:
1. **Clear booking** ("yes," "Tuesday works") → "Great, [first name] — here's the link to grab Tuesday at [time]: [URL]. Take 30 seconds."
2. **Easy question** ("how much," "weekend slots") → answer in one SMS, propose a slot, include link.
3. **Complex question** (medical/dental/billing) → "Got it — [Owner first] will follow up directly today. What's the best window?"
4. **Negative** ("no thanks") → respect immediately. "All good, [first name] — we'll stop. Reply YES anytime."
5. **STOP / UNSUBSCRIBE / QUIT / CANCEL** → immediately fire opt-out workflow. Reply: "Got it, you're unsubscribed. Reply START to opt back in."
6. **Ambiguous** → ask one clarifying question. If reply count at 3, escalate.
7. **At reply #4** (regardless) → "Let me get [Owner first] in here directly — sec." + escalation flag.

Output JSON:
{
  "category": "...",
  "reply_text": "...",
  "escalate_to_human": true|false,
  "escalation_reason": "...",
  "include_booking_link": true|false,
  "log_outcome": "booked|conversation_continuing|negative|opted_out|escalated"
}

Generate the weekly performance digest email sent to [OWNER NAME] at [BUSINESS NAME] every Friday at 4pm.

Inputs:
- N customers messaged this week
- N replies received (split by category)
- N bookings closed
- $ revenue from re-bookings
- N escalations awaiting owner response
- Comparison vs last week + 4-week average

Structure:
- Subject: "[BUSINESS] · Reactivation week of [date]: $[revenue] from [N] re-bookings"
- Top: headline number (revenue) + cost ($300 + SMS this week).
- Stats grid: 4 KPIs side by side.
- Highlight box: BEST reactivation of the week ("Sarah J., last visit Aug 2025, booked highlights for Friday — $145").
- Action items: list of escalations awaiting owner reply.
- Mini-recommendation: one tone or timing change to test next week.
- Footer: "Reply with questions or schedule a 15-min review: [Cal.com]."

Tone: factual, mildly conversational. Owner should look forward to this email.

Output: complete HTML email template.

Anchor on customer LTV recovery. A $1k LTV customer reactivated covers 3+ years of monthly fees. Per-SMS cost: $0.0079 in the US (Twilio). At 60 outbounds/week = ~$2/mo pass-through. Setup includes: customer DB integration, voice training, message templates per segment, reply handler, opt-out compliance, weekly digest. Does NOT include: Mindbody/Boulevard/Vagaro live integration ($1k each), multi-location ($500 each), custom personality per technician ($500). Upsell: combine with Block 01 (voice receptionist) for customers who reply by calling, Block 04 (review responses) for post-visit review request → "Customer Loop" at $700/mo.