Kira: —and it booked the call. Right there. At midnight, Mexico City time, I'm lying in bed scrolling WhatsApp and this thing just... books the call.
Santi: Wait wait wait — back up. What triggered it?
Kira: Instagram ad. I clicked through to test someone's funnel, it opened WhatsApp, and within four questions — budget, timeline, use case, decision authority — it had qualified me and dropped a Calendly link. No human involved.
Santi: Four questions. That's it?
Kira: Four questions. And here's what actually happened — I said my budget was under five hundred, and it immediately pivoted to a forty-nine dollar workshop seat instead of the strategy call. The thing actually listened to my answers.
Santi: Okay but here's the thing — what's the per-message cost on that? Because we're talking about, what, like eight to twelve messages total including the greeting and booking confirmation?
Kira: This is where it gets interesting. They're using Gupshup — I checked the headers. So we're looking at the January 2026 rate card, marketing template messages in the US... point zero two four cents per message. Plus Gupshup's markup, so maybe three cents total.
Santi: Three cents times twelve messages. Thirty-six cents to qualify and book a lead.
Kira: Thirty-six cents. And if they started from an ad — which I did — Twilio says that's a seventy-two hour free conversation window. So actually, zero cents.
Santi: Zero cents to book a qualified lead.
Kira: If you structure it right. If you know the loopholes. If you can actually build the thing without it hallucinating appointment times at two AM.
Santi: By the end of this episode, you'll have a working WhatsApp AI concierge that captures leads, qualifies them with four targeted questions, and books paid calls automatically — running twenty-four seven while you sleep, travel, or focus on delivery.
Kira: We're talking exact costs per conversation, the Make or n8n blueprint to build it, and the compliance guardrails that keep you from getting throttled by Meta.
Kira: So here's the reality for most nomads trying to book calls. You're posting on LinkedIn, maybe running some ads, sending DMs when you have time. But you're also jumping between time zones, dealing with sketchy wifi, trying to have an actual life. And leads are slipping through because you can't respond fast enough.
Santi: Or you respond at the wrong time. I once lost a fifteen hundred dollar monthly retainer because I replied to their DM eight hours later — from Bali — and they'd already booked with someone else.
Kira: Eight hours.
Santi: Eight hours was too long. And that's the problem with email, LinkedIn, even Instagram DMs — they require you to be present. But WhatsApp? WhatsApp with an AI concierge? That thing never sleeps.
Kira: And the deliverability is completely different. Email open rates for cold outreach are what — fifteen, twenty percent if you're lucky?
Santi: WhatsApp messages get opened at ninety-eight percent. Ninety-eight! And responses happen in minutes, not days. But — and this is critical — only if you understand the cost structure. Because Meta changed everything in July 2025.
Kira: Right. So let me break this down for anyone who hasn't been following the pricing wars. Meta now charges per template message for business-initiated conversations. Marketing templates, utility templates, authentication — they all have different rates.
Santi: And then your BSP — that's Twilio, Gupshup, whoever — adds their own fee on top. So when you're modeling costs, you need both numbers.
Kira: Exactly. Gupshup's January 2026 rate card shows US marketing messages at point zero two four cents per message. Twilio adds point five cents as their platform fee. So you're looking at roughly three cents per outbound message all-in.
Santi: But here's where it gets interesting — if someone clicks your WhatsApp link from an ad, that starts a user-initiated conversation. And Twilio explicitly states those are free for seventy-two hours.
Kira: Seventy-two hours of free messaging.
Santi: Which means if you're driving traffic from Instagram or Facebook ads directly to WhatsApp, your entire qualification and booking flow costs nothing. Zero. As long as it happens within three days.
Kira: Okay but here's what you're not considering — you still have LLM costs. Every message back and forth is hitting GPT-4 or Claude or whatever you're using.
Santi: Sure, but look at the actual numbers here. Skywork's 2026 pricing guide estimates about three hundred tokens per turn — that's input plus output. A typical qualification conversation is what, six to eight turns? So we're talking eighteen hundred tokens total.
Kira: At current GPT-4o-mini prices... that's like three cents for the entire conversation.
Santi: Three cents. Even if you use the full GPT-4o, you're at maybe twelve cents. The LLM cost is basically negligible compared to what you're charging for the call.
Kira: So let's map out exactly how this funnel works. Someone sees your ad, clicks through to WhatsApp. What happens next?
Santi: First message is your greeting. Keep it short, personal, acknowledge where they came from. "Hey! Saw you clicked through from my Instagram post about AI automation. I help nomad founders automate their lead capture — mind if I ask you four quick questions to see if I can help?"
Kira: Wait — you're asking permission?
Santi: Always ask permission. It's not just polite, it's required by WhatsApp policy. Infobip's compliance docs are super clear — explicit opt-in before you start qualifying.
Kira: Right. And if they say yes?
Santi: Four questions. First: "What's your current monthly revenue range?" Give them brackets — under five K, five to fifteen K, fifteen K plus. This immediately tells you which offer to present.
Kira: And if they're under your minimum?
Santi: The bot pivots. Instead of booking a strategy call, it offers a workshop seat or a free resource. You're not wasting time on unqualified leads, but you're also not burning bridges.
Kira: Smart. What's question two?
Santi: Timeline. "When are you looking to implement automation?" Options: This month, next quarter, just researching. Anyone who says "just researching" gets nurture content, not a call.
Kira: Question three?
Santi: Use case. "What's the main thing you want to automate?" But here's the key — give them specific options. Lead capture, client onboarding, content repurposing, customer support. Don't leave it open-ended or the LLM has to interpret too much.
Kira: And question four is authority — can they actually make the buying decision.
Santi: Exactly. "Are you the decision maker for automation investments?" If no, the bot asks if they can loop in the decision maker or schedule a call when both can attend.
Kira: So in four questions, you've qualified budget, timeline, use case, and authority. The BANT framework, basically.
Santi: BANT in four messages. And based on their answers, the bot makes a decision. Qualified leads get a Calendly link. High-value leads get a Calendly link plus a Stripe checkout for a paid discovery call. Low-value leads get a workshop invitation.
Kira: How does the Calendly integration actually work?
Santi: Dead simple. Your automation platform — Make, n8n, Zapier — checks Calendly for available slots via API, generates a booking link with their info pre-filled, sends it in WhatsApp. They click, pick a time, done.
Kira: But here's what actually happens when you're on the ground — time zones are a nightmare. How do you handle someone in Singapore booking a call when you're in Mexico?
Santi: This killed me for months! The solution is to set your Calendly availability in blocks that work regardless of where you are. I do nine AM to noon in whatever timezone I'm in. The bot doesn't need to know where I am — Calendly handles the conversion.
Kira: Okay but what about the payment piece? You mentioned Stripe checkout.
Santi: This is optional but powerful. For high-ticket offers, you can require a deposit to book. The bot generates a Stripe Payment Link — not a full checkout, just a simple payment link — for forty-nine or ninety-seven dollars. "To secure your strategy session, there's a ninety-seven dollar deposit that's fully credited toward any package you choose."
Kira: And people actually pay?
Santi: The right people pay. Look, if someone won't put ninety-seven dollars down for a consultation that could transform their business, they're probably not buying your five thousand dollar package anyway.
Kira: Fair. But this whole flow assumes everything works perfectly. What about when it breaks?
Santi: So let's talk about failure modes. Because this will break. The question is whether it breaks gracefully or catastrophically.
Kira: First failure point — template rejections. Meta has to approve your message templates, and they reject about thirty percent on first submission.
Santi: Usually for the stupidest reasons too. You used the word "free" too many times. Or your call-to-action is too vague. The fix is simple — keep templates minimal, avoid promotional language, resubmit with tweaks.
Kira: Second failure — rate limits. WhatsApp has messaging limits based on your quality rating. Start too aggressive and you'll get throttled.
Santi: The October 2025 change actually made this better. It's now portfolio-based, not per-number. So one bad campaign doesn't tank your entire operation. But you still need to ramp slowly. Start with ten conversations a day, scale up over weeks.
Kira: What about LLM hallucinations? What if the bot starts making up appointment times?
Santi: This is why you never let the LLM freestyle. Every response is constrained by templates and function calls. The bot can only send pre-approved messages with variables filled in. It can't invent a Tuesday at 3 PM slot that doesn't exist.
Kira: But here's what actually happens — someone asks an edge case question. "Can you help me automate my dropshipping store but only for products from sustainable suppliers in Southeast Asia?" The bot has no template for that.
Santi: Human handoff trigger! Any message that doesn't match your expected patterns triggers an alert. The bot says "That's a great question — let me connect you with Santi directly." Then it pings you on Telegram.
Kira: And if you're sleeping? Or on a flight?
Santi: The bot sets expectations. "I've flagged this for Santi and he'll respond within twelve hours." Then it offers to schedule a call instead of continuing the chat.
Kira: What about payment failures? Someone's card declines on the Stripe link?
Santi: Retry logic with backoff. First retry after one hour, second after six hours, then send a different payment method option — PayPal, Wise, whatever you accept. After three failures, human handoff.
Kira: Timezone misses are brutal though. The bot books someone for nine AM your time, but you forgot to update Calendly when you flew from Lisbon to Bangkok.
Santi: Been there. The fix is redundancy. Calendly sends confirmation emails with timezone clearly stated. The bot sends a WhatsApp reminder twenty-four hours before. And you set up a final manual check each evening for next-day calls.
Kira: Double bookings?
Santi: Shouldn't happen if your Calendly API integration is solid. But as backup, implement a mutex — a booking lock — in your automation. Once someone starts booking a slot, lock it for five minutes until the booking completes.
Kira: Okay, so we've covered the happy path and the failure modes. Let's talk real numbers. What does this actually cost to run?
Santi: Let me pull up my spreadsheet... Okay, assuming a hundred leads per month hit your WhatsApp. Thirty percent respond and engage — that's thirty conversations. Each conversation averages eight turns, so two hundred forty messages total.
Kira: At three cents per message if they're not from ads...
Santi: Seven dollars twenty cents in WhatsApp costs. Add LLM costs — thirty conversations times eighteen hundred tokens each at GPT-4o-mini rates... another ninety cents. Call it eight dollars total.
Kira: Eight dollars to process a hundred leads.
Santi: But here's where the math gets interesting. If your qualification is good, maybe forty percent of those engaged leads book calls. That's twelve booked calls from a hundred leads. If just three of those convert to clients at fifteen hundred dollars each...
Kira: Forty-five hundred in revenue from eight dollars in costs.
Santi: Five hundred sixty-two X return. And that's conservative — assumes zero leads from ads, lowest tier pricing, minimal conversion.
Kira: Okay but you're not counting the tools. Calendly, Make or n8n, the WhatsApp Business API setup...
Santi: Fair. Calendly is twenty dollars a month. Make.com starts at nine dollars. WhatsApp Business API through Twilio or Gupshup has no monthly fee — just usage. So fixed costs of about thirty dollars monthly.
Kira: Plus your time to set it up.
Santi: Which is why we're giving away the blueprint! The complete Make.com scenario, the prompt templates, the error handling logic — it's all in our Slack. Probably saves you twenty hours of trial and error.
Kira: Speaking of which — let's talk about the actual build. What does the Lisbon Test say about this?
Santi: Can you build it from a café with terrible wifi? Actually, yes. The whole thing runs on Make or n8n, which are visual builders. No coding required. You can literally set this up on your phone if you're desperate.
Kira: But should you?
Santi: God no. Use a real computer. But the point is, it's resilient. Once it's running, it doesn't need you. I've had mine running for six months through three visa runs and probably fifty different wifi networks.
Kira: Let's talk about the three offer archetypes you could run through this funnel. Because the beauty is, same tech, different positioning.
Santi: Archetype one: The audit tripwire. Ninety-seven dollar automation audit that leads to a fifteen hundred dollar implementation package. You're basically getting paid to pitch.
Kira: How does that flow?
Santi: Someone books the audit through WhatsApp, pays ninety-seven via Stripe, you do a thirty-minute screen share showing them their automation gaps. At the end, you present three packages — basic at fifteen hundred, standard at three thousand, premium at five thousand. The ninety-seven credits toward any package.
Kira: Conversion rate?
Santi: In my experience? About sixty percent take at least the basic package. So you're effectively acquiring customers at negative CAC — they pay you to sell to them.
Kira: Archetype two?
Santi: The workshop funnel. Forty-nine dollar seat in a weekly group workshop. "Build Your First AI Agent" or whatever your expertise is. During the workshop, you pitch your done-for-you service or your course.
Kira: This is smart because it's lower commitment than a one-on-one call.
Santi: Exactly. And you can run it async if needed. Record the workshop once, deliver it automated, then do live Q&A via WhatsApp. Scales way better than individual calls.
Kira: What's the upsell rate?
Santi: Lower than the audit — maybe twenty percent. But you're processing more volume. If you get fifty people through a workshop monthly and ten convert at a thousand dollars each...
Kira: Ten thousand MRR.
Santi: From a forty-nine dollar workshop. The key is the WhatsApp follow-up. The bot continues the conversation after the workshop, sending case studies, answering questions, eventually offering a "workshop special" discount.
Kira: And archetype three?
Santi: Free lead magnet to paid strategy call. This is the classic funnel but supercharged with WhatsApp. They get a free template or checklist, the bot nurtures them for three days with tips, then offers a three hundred dollar strategy session.
Kira: Three hundred seems high for a cold lead.
Santi: It is. But here's what actually happens — the three-day nurture sequence is basically a mini-course. By the time they're offered the call, they've gotten real value. They know you're legit. And three hundred dollars filters out tire kickers.
Kira: What's the booking rate on that?
Santi: About eight percent of lead magnet downloads book a call. But those eight percent are serious. Conversion to high-ticket offers is around forty percent.
Kira: So let's talk about compliance. Because Meta will shut you down if you screw this up.
Santi: First rule — explicit opt-in. Every single person in your WhatsApp funnel must have explicitly agreed to receive messages. No buying lists, no scraping, no "I thought they'd be interested."
Kira: And you need to document the opt-in.
Santi: Timestamp, source, exact language they agreed to. Infobip's compliance docs are crystal clear on this. Store it all, because Meta can ask for proof.
Kira: Second rule?
Santi: Honor opt-outs immediately. Someone types "STOP" or "unsubscribe"? The bot must recognize that and cease all messages. No "are you sure?" No "but wait there's more." Stop means stop.
Kira: What about data retention?
Santi: This is where it gets tricky. GDPR says minimize data collection and delete when no longer needed. But Meta wants opt-in proof. The compromise? Keep opt-in records and conversation logs for six months, then auto-delete.
Kira: Template approvals are a pain though.
Santi: The worst. Meta reviews every template, and their standards are... inconsistent. One day "Book your free call" gets approved, next day it's rejected for being too promotional.
Kira: Tips for getting approved?
Santi: Keep it conversational, avoid marketing speak, include clear opt-out instructions. And always have backup templates ready. If one gets rejected, you need alternatives queued up.
Kira: What about the new messaging limits?
Santi: Since October 2025, it's portfolio-based quality rating. Start slow — ten conversations per day for the first week. If your quality score stays high, it auto-increases to a hundred, then a thousand, then ten thousand daily.
Kira: And if your quality drops?
Santi: You get throttled back down. Too many blocks, reports, or opt-outs and you're back to ten per day. That's why the qualification questions are so important — you want engaged leads, not annoyed strangers.
Kira: So we've covered the tech, the economics, the compliance. But here's what I'm wondering — is this actually better than just hiring a VA?
Santi: Look, for eight dollars a month in variable costs, can you hire someone to respond to a hundred leads within minutes, twenty-four seven, never missing a message?
Kira: No, but a VA can handle edge cases the bot can't.
Santi: Which is why you need both! The bot handles the ninety percent — standard qualifications, bookings, FAQs. The VA handles the ten percent that need human touch. It's not bot versus human, it's bot plus human.
Kira: That's actually the key insight here. You're not replacing human interaction, you're—
Santi: You're eliminating the repetitive parts so humans can focus on high-value conversations. My VA used to spend three hours a day on initial lead qualification. Now she spends those three hours on client success.
Kira: And the clients get better service because of it.
Santi: Exactly. Plus — and this is huge for nomads — the bot doesn't care what timezone you're in. It doesn't get confused when you fly from Berlin to Bali. It just works.
Kira: Okay, so let's talk about your build-in-public experiment. You're actually going to run this funnel for a week?
Santi: Starting tomorrow! I'm setting up a ninety-seven dollar automation audit offer, driving traffic from my Twitter and LinkedIn, tracking everything. Cost per lead, cost per qualified lead, cost per booked call, show rate, conversion rate.
Kira: What's your hypothesis?
Santi: I think I can book five qualified calls for under fifty dollars total spend, with at least a sixty percent show rate. And at least two of those should convert to the fifteen hundred dollar package.
Kira: That's aggressive. What if the WhatsApp costs are higher than expected?
Santi: Then we'll know! That's the point. Real numbers from a real funnel, not theoretical calculations. I'll share everything in next week's episode — the good, the bad, the broken.
Kira: What about the tech stack? What are you using?
Santi: Make.com for the automation — I already have an account. Gupshup for WhatsApp BSP because their rates are slightly better than Twilio. GPT-4o-mini for the LLM because it's dirt cheap and good enough for qualification. Calendly for booking, Stripe for payments.
Kira: And for monitoring?
Santi: After my Thailand disaster? Everything has monitoring. Make.com webhooks to Telegram for errors. Daily summary of conversations and costs. Alert if qualification rate drops below twenty percent.
Kira: Smart. What templates are you starting with?
Santi: Super simple. Greeting, four qualification questions, three different offer presentations based on their answers, booking confirmation, reminder sequence. Maybe twelve templates total.
Kira: And if Meta rejects them?
Santi: I've got backups ready. Learned that lesson the hard way. Always submit three versions of each critical template. One will get through.
Kira: What's your biggest concern about this experiment?
Santi: Honestly? That it works too well and I get flooded with calls I can't handle. But seriously, the risk is always human bandwidth. The bot can book unlimited calls, but I can only do so many per day.
Kira: That's a good problem to have though.
Santi: It's the best problem. And it's solvable — raise prices, add a junior consultant, productize more of the delivery. But first, let's see if the funnel actually works.
Kira: I'm excited to see the results. And for anyone who wants to build this themselves?
Santi: Join our Slack. The complete blueprint is there — the Make.com scenario, all the prompt templates, the error handling logic, even my spreadsheet for calculating ROI. Link is in the show notes.
Kira: Plus we'll be answering questions all week as people build their own versions.
Santi: And here's the thing — this isn't just for consultants. This same funnel works for course creators, coaches, service providers, even product demos. The structure is universal.
Kira: The four questions change but the flow stays the same.
Santi: Exactly. Budget, timeline, use case, authority — that framework applies to almost any B2B sale. You just adjust the specifics.
Kira: You know what strikes me about this whole system? It's exactly the kind of thing that would have seemed impossible five years ago. AI that can actually qualify leads, WhatsApp as a business channel, payments built into chat...
Santi: The tools have finally caught up to the vision. For years we talked about conversational commerce, but the tech wasn't there. Now? Now a solo founder in Canggu can build something that used to require a whole sales team.
Kira: And that's the real opportunity for nomads. You don't need an office, you don't need employees, you don't even need to be awake. Just solid systems and good monitoring.
Santi: The Lisbon Test in action. Can you run it from a café? Yes. Should you? Also yes, because that's the whole point of this lifestyle.
Kira: Build once, run anywhere.
Santi: That's the dream. And with WhatsApp AI concierges, it's actually achievable. Not someday — today. With tools that exist right now.
Kira: So here's what we covered — a WhatsApp AI concierge that qualifies leads with four questions, books calls automatically, and costs less than ten dollars a month to run at scale. The math is undeniable.
Santi: Five hundred X returns if you structure it right. And remember — if you're driving traffic from ads, those first seventy-two hours of messaging are completely free. Zero cost to qualify and book.
Kira: The blueprint is waiting for you in our Slack — the complete Make.com scenario, all twelve message templates, the error handling logic, even Santi's ROI calculator. Everything you need to build this yourself.
Santi: And next week I'll share the real numbers from my build-in-public experiment. How many leads, how many booked, how many showed up, how many bought. No theory — actual results.
Kira: Plus we'll be in the Slack all week answering questions as you build your own version. Because here's the thing — this works for any offer. Consulting, coaching, courses, even SaaS demos.
Santi: The structure stays the same. Four questions, smart routing, automated booking. That's it.