The Stateless Founder

Ship a Weekend Voice Concierge: 5 Intents, Spend Guards, and Zero Surprises

Listen Later
Ship a Weekend Voice Concierge: 5 Intents, Spend Guards, and Zero Surprises
The Problem: Missing Leads Across Time Zones
  • The 2 AM lead story: Kira loses a $2,500 engagement because her phone had "business hours"
  • The math: One missed qualified call per night × $1,500 average engagement = $45,000/year in lost revenue
  • The nomad reality: Time zone chaos means prospects call when you're asleep
    • Architecture: Three Pieces, One Stream

    The Stack:

    1. Twilio Programmable Voice - Receives calls, routes to webhook
    2. OpenAI Realtime API - Handles conversation (listen, think, speak) in one stream
    3. Function calls - Books calendar slots, creates CRM leads, sends confirmation texts

      4. Caller Experience:

      • Phone rings → Twilio picks up instantly
      • Greeting within 1 second with consent request
      • Agent determines intent and handles or escalates
      • Sub-2-second response times throughout
        • The Five-Intent Scope

        What the agent does:

        1. Book appointments - Calendar integration with availability checking
        2. Reschedule existing appointments - Modify confirmed bookings
        3. Qualify new leads - Ask targeted questions, create CRM entries
        4. Deflect pricing inquiries - Text link to pricing page + callback offer
        5. Take voicemails - After-hours or escalation fallback

          6. What it doesn't do: Everything else gets handed to a human

          Cost Model: Realtime vs. Split Stack
          OpenAI Realtime API (March 2026 pricing)
          • gpt-realtime-1.5: ~9.6 cents/minute (audio only)
          • gpt-realtime-mini: ~3 cents/minute (audio only)
          • Typical call cost: 6-12 cents audio + telephony = ~35 cents total (US-to-US)
            • Alternative: Split Stack
            • Google Speech-to-Text: $0.016/minute
            • Amazon Polly Neural TTS: ~$0.0115/minute
            • Combined audio cost: Under 3 cents/minute
            • Tradeoff: More complexity, slower time-to-first-audio, latency management
              • Telephony Costs
              • Twilio rates vary by country - Use Pricing API to fetch real-time rates
              • Never assume flat global pricing - Pull country-specific rates at build time
                • Guardrails and Compliance
                Consent and Recording
                • First thing agent says: "This line uses AI to assist and is recorded for quality. Do I have your permission to continue?"
                • Federal vs. state law: One-party consent federally, but some states require all-party
                • Safe default: Announce, ask permission, pivot if declined
                  • Do-Not-Call Protection
                  • Immediate flag on "take me off your list"
                  • Agent response: "I've marked your number as do-not-contact"
                  • CRM integration: Tag number to prevent future calls
                    • Language Fallback
                    • Graceful pivot for non-English speakers
                    • No crashes or silence - Transfer or basic Spanish response
                      • Spend Guards and Fail-Safes
                      Cost Controls
                      • Per-call cap: Set maximum spend (e.g., 35 cents)
                      • Monthly cap: Prevent bill spirals
                      • Turn limit: Max 8 turns per call (later turns cost more)
                      • Auto-disconnect: Before costs exceed limits
                        • Quality Controls
                        • Intent confidence <0.6: Automatic human handoff
                        • Latency cap: ~1.8 seconds maximum response time
                        • Frustration detection: Escalate on repeated failures
                          • Proof It Works: Real Case Studies
                          The Melting Pot (Restaurant Chain)
                          • $250k revenue from after-hours bookings in 6 months
                          • 68% automation of reservation-related calls
                          • Four intents: Create, modify, cancel reservations + FAQs
                            • Small Business Example (Goodcall)
                            • 25% monthly revenue lift for junk removal company
                            • $2,500+ in bookings in first 30 days
                            • Simple scope: Answer calls, book appointments, basic Q&A
                              • Monitoring From Anywhere
                              Dashboard Metrics
                              • Booking rate: Successful appointments scheduled
                              • Escalation rate: Calls transferred to humans (target <20%)
                              • Average cost per call: Track spend trends
                              • Time-to-first-audio: Response speed monitoring
                                • Weekly Check-in (10 minutes)
                                • Monday morning review: 10 minutes from any café
                                • Red flags: Escalation rate >20%, rising costs, dropping booking rate
                                • Quick fixes: Tighten turn caps, simplify qualification, check integrations
                                  • The Weekend Voice Concierge Starter Pack

                                  Included Templates:

                                  • Call-flow JSON for five intents
                                  • Make.com/Zapier blueprint for calendar + CRM
                                  • Spend-guard script (per-call + monthly caps)
                                  • Consent and escalation copy snippets
                                  • Metrics dashboard template

                                    • Your Homework:

                                    1. Pull Twilio per-country rates using Pricing API
                                    2. Deploy the concierge with spend guards
                                    3. Check dashboard Monday morning
                                      4. Key Takeaways
                                      1. Cost-effective capture: 35 cents per call to book $1,500+ clients
                                      2. Bounded beats complex: Five intents work better than general-purpose agents
                                      3. Deflection converts: Pricing links + callbacks outperform agent quotes
                                      4. Proven ROI: Real businesses generating hundreds of thousands from after-hours automation

                                        5. Grab the Weekend Voice Concierge Starter Pack in the show notes and stop missing leads while you sleep.

                                        ...more
                                        View all episodesView all episodes
                                        Download on the App Store
                                        The Stateless FounderBy Santi, Kira