Every booking product gets sold the same way. Four autonomy levels, start cautious, turn it up as the client learns to trust the AI. That framing is wrong, and it costs agencies bookings in the first month. The levels and the setup steps below come from the Academy page on booking and calendar. The argument about how to choose one is mine. What has to exist before the bot can book anything Two things, and both are content work rather than configuration: Services. Name, duration, and optionally price and description. The docs put it well: a line like "Deep-tissue massage, 60 min, $80" beats a bare "Massage". Specialists. Each one needs assigned services and a weekly schedule. A specialist with no service or no schedule never shows up in a booking conversation, and nothing warns you about it. Both go in through the Co-Worker chat in plain language. "Add two specialists: Ana does manicures and pedicures, Tue–Sat 10:00–19:00; Marta does haircuts, Mon–Fri 9:00–18:00." That is the whole setup. The four levels Who creates the booking, and whether a human can move it afterwards: Intent Collection. The AI only collects what the lead wants. A human creates every booking. No human reschedule. Booking with Human Approval. The AI collects everything, then a human confirms and creates the booking. No human reschedule. Booking with Human Review. The AI creates the booking, a human reviews it afterwards and notifies. Human reschedule works. Full. The AI handles it end to end, and a human steps in only when the customer explicitly asks for one. Human reschedule works. You change the level by telling the Co-Worker: "Set booking autonomy to Full", or "I want human approval before bookings are created." It never touches existing bookings, so the dial can move mid-campaign without breaking next week's calendar. Notice what the cautious levels cost Read that list again with the reschedule column in mind. The two levels people pick when they are nervous are also the two where a human cannot move an existing appointment. The client who chose Approval because they wanted more control ends up with less of it the moment somebody calls to shift a Tuesday slot.

TikTok is the one channel where a client with a single viral video gets more raw engagement in a day than their WhatsApp sees in a month. That is the pitch, and it holds up. It is also why agencies put TikTok automation into the proposal before checking two things that can kill the project before it starts. Both restrictions come straight from the Academy page on connecting TikTok. What they mean for how you sell is mine. Deal-killer one: US accounts cannot connect If the client's TikTok account was signed up in the United States, it cannot be connected to Chatfuel at all. This is about where the account was registered, and that is not something you or the client can change afterwards. Deal-killer two: EU leads are invisible If a lead's own TikTok account is from the EU, their messages do not appear in Leads, Contacts or Chats. The automation simply does not see them. The client's account can be connected and working perfectly, and those people still land nowhere. The two are not the same kind of problem This is the part worth slowing down on, because the docs state both restrictions in one breath and they behave completely differently once you are live. Read them again and notice whose account each one is about: The US rule is about the client's account. One question, one answer, and you know before you sign. If it fails, nothing gets built and nobody loses money. The EU rule is about each individual lead's account. You cannot check it in advance, it throws no error, and it stays silent for every lead it affects. The second one is the one that costs you a client, and it costs you the client three months in rather than on day one, because a slow trickle of DMs looks exactly like a targeting problem or a creative problem, and you will spend two reporting cycles adjusting audiences and rewriting hooks for a channel that is quietly dropping some unknown share of the people it reaches before you think to ask where those people are sitting. A restriction that returns no error is worse than one that blocks you outright.

Access is where most builds stall. The client says yes in the meeting, and then two weeks disappear into screenshots, offers of a password, and one Facebook account nobody can log into anymore. This one comes from the Academy page on partner access. The click paths, the permission levels and the WhatsApp warning are theirs. The ordering advice and the war story are mine. People access and partner access are not the same door The docs draw the line clearly, and getting it wrong is what creates the two-week version: - People access is for individuals. The client's own employees. Use it for your team and the client has to invite each of your teammates one by one, and when someone leaves your agency the client has to remember to remove them. - Partner access is for businesses. Agencies and freelancers with a portfolio. One grant from the client, and you assign your own people to their assets yourself. Someone leaves, you remove them on your side and the client does nothing. - Ownership does not move either way. Partner access never transfers ownership, and the client keeps everything. Worth saying out loud on the call, because it is the thing they are quietly worried about. You need your own portfolio before any of this Partner access is business to business, so you need a Business Portfolio of your own at business.facebook.com. Your ID is the 16-digit number under your business name in Business Settings → Business Info. Keep it somewhere you can paste from. You read it out on almost every onboarding call. Route A: the client grants access The client does the work, from their own settings: 1. Business Settings → Partners → Add → Give a partner access to your assets. 2. They enter your 16-digit Business Portfolio ID. 3. They pick the assets to share. 4. They set the permission levels and click Assign. 5. The assets appear in your portfolio within a minute. This route is fine when the client is technical, or when you are sitting next to them anyway.

The proposal meeting is the one hour where the build gets decided, and most agencies walk into it with the wrong plan. They bring a deck about the platform. The client already believes WhatsApp works, which is why the meeting exists at all. What they do not believe yet is that you should be the one holding their number. The meeting is not a demo Two rules I keep: - Do not open the platform. The second a screen goes up the conversation becomes about features, and features invite comparison shopping. - Stop presenting the moment they say yes. I sat in a meeting where the founder agreed twenty minutes in and the agency kept going for another forty, hit a slide about AI, and reopened a question that had already closed. You are in the room to answer five objections. Those five are what the hour is for. "We already reply fast" Do not argue with this and do not reach for a stopwatch. Ask one question instead: fast between which hours. Almost nobody has an answer that covers Saturday evening, and that is when a good share of the enquiries land, because people shop for apartments and clinics and courses when they are not at work. Then ask what happens to the ones that arrive outside those hours. Usually someone gets to them in the morning, which means the lead has had a night to message two competitors. "My customers will know it's a bot" They will. Agree with it immediately, because the alternative is defending a claim you cannot keep, and the first time someone types something unusual you lose the argument in front of the client. The move is to change what is being bought. What you are selling is an intake that never sleeps and hands over to a named human with the details already collected. Nobody is pretending it is a person. Then read the first message out loud, the exact words the client's leads would see. Most robot objections die when the owner hears a first line that sounds like their own receptionist. "It's too expensive" Never defend the number. Reprice the unit. Ask what they pay today for a lead that turns into a booking, put your monthly figure over the volume you expect, and compare those two numbers instead of comparing your retainer against zero.

Real estate is a Special Ad Category. Housing sits there next to credit, employment and financial products, and you have to declare it when you create the campaign. Failing to declare is itself a violation, so this is not a box you skip and sort out later. What most people miss is what the declaration does to the ad set. What Meta switches off Once the campaign is declared, targeting is deliberately blunted: - Age locks to 18–65+. No narrowing, whatever the property is. - All genders are included. No exclusion. - No ZIP-code targeting. Location goes broad, and the tight geography you built the media plan around is gone. - No Lookalike audiences. Your best seed list stops being usable as a source. Everything you would normally do to make a housing ad efficient is switched off, and it is switched off for legal reasons, so it is not coming back. The expensive mistake Mine, and it is the first thing I look for in a housing account: people try to rebuild the excluded audience out of interests. Interested in mortgages, interested in home improvement, engaged shoppers, income proxies. That fails twice. Delivery gets worse because you shrank the pool the algorithm needs, and you have just built a housing audience by demographic proxy, which is the exact thing the category exists to prevent. I sat in an account last year with six saved audiences still attached to declared housing ad sets. Meta had been ignoring every one of them for months, and the team had a whole theory about which ones worked. Angle over audience The playbook that survives is angle over audience: separate ads per intent, buyers, sellers, investors, with copy that self-selects, and let broad delivery find the people. Delivery is good at finding who responds to a message. Give it five distinct messages instead of one message and five audiences. Five doors Mine, the set I would run for a residential client. One ad each, first line doing the filtering: 1. Ready buyer. "Three-bedroom units in [area], delivery 2027, floor plans and prices in WhatsApp." 2. Budget-first buyer. "What a 400k budget buys in [area] this month." 3. Seller. "What apartments on [street] actually closed for this quarter." 4. Investor. "Rental yields by neighbourhood, updated monthly." 5. Renter. "Available this week in [area], with the fees listed up front."