MB SuiteMB Suite

Put the AI Team to work: the preconditions nobody validates for you

Updated 2026-10-13

Milestone 7 of 8 · The AI working (optional)

In two lines: the AI comes last, and for a very concrete reason: a position that's missing a piece gives no error, it simply does nothing. Going through the list below before turning it on saves you days of diagnosis.

When you finish you'll have: a position talking with real leads on a single channel, with its conversations already reviewed by you and no automation stepping on its work.

This is an optional step. The CRM works perfectly without it.

Which position suits me?

You add them in CRM > Settings > Team > AI Team. Three of them talk with leads and work them like one more salesperson — the Qualifier, the Setter and, in Beta, the Seller; this guide focuses on the first two. The rest score or advise in the background.

PositionWhat it doesDoes it talk?
QualifierWorks the lead until it's qualified and hands it to a salespersonYes
SetterWorks the lead until it books a meeting, disqualifies it or hands it offYes
Seller (Beta)Sells over WhatsApp from your catalog: recommends products and takes the orderYes (WhatsApp only)
Prospect scoringScores incoming prospectsNo
Deal scoringScores dealsNo
Deal coachSuggests next steps on a deal, when you ask it toNo

The Qualifier and the Setter become the owner of the prospect or the deal, just like a person.

How to choose. If your question is "is this lead worth it?", it's the Qualifier. If your bottleneck is "we can't get all the meetings booked", it's the Setter.

If your team doesn't take meetings through a public booking page yet, start with the Qualifier. The Setter requires the person in charge (or the calendar you choose in Assignment) to have their booking page enabled — the Booking column in Team —: without it, it won't even let you save it with the Booking channel on. And if that booking page gets deactivated later, the position stops working entirely, not just booking.

Start with a single one. The scoring positions and the coach are a different matter and are explained in Qualify and boost deals with AI (AI Score and AI Coach).

What has to be ready before turning it on?

A single list, and it's worth going through all of it. If something from the first part is missing, the position stays active and does nothing. If something from the second part is missing, it does work, but your lead receives two different messages from the same company.

For the position to be able to work:

  • The WhatsApp line is connected and active.
  • There's a lead source tied to that line. Without a source no prospect or deal gets created, and without that entity the position never wakes up.
  • There's an assignment rule that sends that source to the position. It's chosen as owner in A fixed person, or as Fallback if nobody matches; it doesn't enter a team rotation.
  • If the position converts prospects, the Target pipeline is chosen.
  • The position is Active in the table. When Paused, the incoming message is ignored: no reply and no turn in the AI Log.
  • Your subscription is up to date.

For it to be able to reach out on each channel, without colliding with anything already running:

  • WhatsApp. There are approved templates selected in the position: without them, every lead who didn't write to you first — the ones from forms and ads — is unreachable. And on that same number, switch off the WhatsApp Business app's automatic messages — away and greeting messages — and any external chatbot or platform connected. Those are what reach the lead stuck to the position's reply.
  • Email. The Email channel is on if your leads may not have a phone number: it comes off, and a position saved as-is only speaks over WhatsApp and hands every lead without a number to a human. The Person in charge has to have their mailbox connected — the form itself warns you if it isn't. And if the position is going to send the first email, switch off the automatic reply to the visitor of your web form.
  • The CRM's Sequences, on either channel. Their email and WhatsApp steps are sent by the engine on its own, and it doesn't know that lead is now being worked by the AI. Pause the sequence or unenroll those leads. If you keep it running, make sure it doesn't take them in again on its own: unlink its linked list and remove from its Start block the triggers that catch these leads (for example, Contact created or Deal created), or narrow them down with Who gets in; otherwise it enrolls the new leads by itself. See Make a sequence start on its own: triggers, entry filter and re-entry.
On MB Suite's side there's nothing else to switch off: there are no WhatsApp auto-replies of its own. And two things that are not duplication: two positions on the same lead — a lead has a single owner, so two positions never work it at the same time — and the meeting reminders that go out after the Setter books.
The Setter's booking page is validated when you save. What it doesn't warn you about on screen is if it gets deactivated later: then the position stops running and the person in charge gets an email alert, once a day, until it's resolved.

What do I configure inside the position?

The position opens in tabs. These are the ones that move the result:

  • Goal — the Fields to capture it has to get from the conversation, and in which case it considers the lead qualified or disqualified. Write them with concrete examples: "has a defined budget and is the decision-maker" works; "is a good lead" doesn't. Don't overdo the required fields: each one is one more question, and a position with too many mandatory ones interrogates instead of conversing. In the Setter, the Email field comes first and decides whether it can book without it: it starts on Required, and you can switch it to Optional or Don't ask (what changes for the lead, further down).
  • Knowledge — it's the only thing the position knows about your business. This is where your FAQs, your price ranges and your shipping, warranty or cancellation policies go. The tab itself shows the maximum number and size of files. It only appears after saving the position for the first time.
  • Assignment — the Person in charge is mandatory, and without it you can't save. When the position hands the lead to your team, the deal goes to whoever the assignment rule of the lead's source picks; the person in charge receives it when that rule doesn't pick anyone on the team (for example, because the source sends its leads to the position itself). This is also where you pick the Target pipeline.
  • Channels — which channels it uses and from which of your team's accounts. Check that Email is on if you need it. The WhatsApp card carries its own two decisions: Lines and templates — which lines it may use and which approved templates of each one; with more than one line they start collapsed — and Outside WhatsApp's 24h window, which is what it does once free text is no longer allowed. Below the cards, as soon as one channel is on, the Contact strategy block appears: Follow-up pace — how long it waits before each reminder, per active channel — is always there, and with WhatsApp and Email both on it also brings the two settings that cross channels, Channel to start with and the alternation between channels. What each option changes for the lead is in the next section.
  • Schedule — when it replies and when it initiates, with its time zone. It's the tab that most decides the experience of whoever contacts you: the next section takes it apart control by control.
  • Limits — the Guardrails block, and nothing else. The two caps that decide when the position stops insisting and passes the conversation to a person: how many consecutive follow-ups with no reply, and how many turns without progress. Those two caps are also what bounds the spend: for the first week, lower them to around half of what they come with.
  • Test — it appears once the position is saved, and it's what saves you the grief: you load a test lead, pick the channel and actually talk with the position, with its same brain and its same data, without sending anything to anyone. Use it until you like the conversation.
The position doesn't invent prices by design: if the knowledge file doesn't have them, it hands off. No need to upload a huge document just in case.

Details of every tab: The CRM's AI Team: Qualifier, Setter and Seller.

What does your lead experience with the position on?

None of what follows changes what you see: it changes what happens to the person on the other side. It's one behavior worth knowing and eight decisions worth making consciously, instead of inheriting them.

Do they know they're talking to an AI? The position doesn't introduce itself as AI or open messages describing itself, but if asked directly whether it's a person, it tells the truth: that it's an AI. It never lies, and it doesn't announce it on its own either. That leaves the decision on your side: if in your industry it's better to say it from the start, write it into your first-contact template, which is the only text of that conversation you write word for word.

If they write to you at nine at night, do they get a reply? That's decided by Behavior on incoming messages, in Schedule: Always reactive (replies outside business hours too) or Business hours only. The first is the main reason an AI position gets set up — replying when your team isn't there —; the second makes the lead wait until tomorrow. Neither is wrong, but it's your choice, not a detail.

If you write to them first, what time does it arrive? That's Behavior on outgoing messages: Always, Business hours only or Custom window. And if you sell in more than one country, turn on Respect the lead's time zone: the position postpones outgoing messages until it's business hours where that person is, detected from their number's country. It's what avoids the sales message at three in the morning.

Double-check the time zone in Schedule. Set wrong, the lead writes and the reply goes out the next day, without anything failing or warning you.

Who does the email come from? In Channels, the position writes from the mailbox of a real person on your team — the screen tells you in so many words, "Sends from … mailbox" — and with Add my signature to every email it goes out with their signature. For the lead, that email arrives from a colleague of yours, with their name and their address. It's a decision, not a configuration detail: choose the mailbox of someone who can carry the conversation if that person replies.

Which channel does the first message arrive on? When the position has WhatsApp and Email both on, that's decided by Channel to start with, in the Contact strategy block of Channels. With Automatic by country (LATAM → WhatsApp, US → email) — the default — leads from the United States and Canada start by email and everyone else by WhatsApp. With AI decides case by case, before the first message the position looks at where the lead came from, whether the phone is mobile or landline, how reliable the email is, how urgent what they wrote sounds, and whether they already answered a form; it picks once, at the start, and leaves the reason as a note in the lead's activity. The other two options, WhatsApp and Email, force the channel. In every case, if the lead wrote to you first, the position replies on that same channel — and if the lead is missing a phone or an email, there's nothing to decide: it goes out on the channel that exists.

AI decides case by case thinks once per lead, with the cheapest model, and that thinking is charged to your AI credits like any other turn. The note it leaves in the activity is what lets you review its judgement: if it starts choosing badly for your business, go back to Automatic by country or force the channel.

Are they going to get hammered on the same channel? Also in Contact strategy, Alternate with email makes it so that, if the lead doesn't answer several WhatsApps in a row, the next follow-up goes out by email the next day and then goes back to WhatsApp. You choose how many unanswered WhatsApps it waits for. It's the difference between insisting and chasing.

How often does it insist? That's Follow-up pace, the last sub-section of Contact strategy: the ladder of minimum waiting time before each reminder when the lead goes quiet. You set it per active channel and step by step — the first reminder, the second, the third, and from the fourth onward —, in minutes and hours for WhatsApp and in days for email, and it widens out as the silence grows. Replies to a message the lead did send never go through that ladder: those are immediate. If what you hear back is "they write to me too much", this is the control to lower, not the follow-up cap.

And if more than 24 hours went by since their last message? WhatsApp no longer allows free text there: only a template approved by Meta. What the position does at that point you choose in Outside WhatsApp's 24h window, inside the WhatsApp card, and the screen explains the option you picked as you pick it. Let the AI decide (recommended) — the default — weighs each case: it tries sooner if the conversation is still warm, waits if it cooled, and outside the window picks between template, waiting or escalating. Force HSM template always sends an approved template. Hand the conversation off to a human and Wait (don't contact until the lead messages again) both mean it doesn't write: the first passes it to the person in charge, the second holds until the lead comes back.

Leave it on Let the AI decide unless you have a reason not to: it's the only option that keeps working when the conversation is still open. If you do choose Force HSM template, two selectors appear right below to pin which approved template it uses for the first contact and which one to re-engage — and until you set at least one, the position won't let you save. Templates that MB Suite can't send yet (an image, video, document or location in the header, or variables in a text header) show up disabled, with the reason.

Do they have to give you their email to get the meeting? That's decided by the Email field of Fields to capture, in Goal, and it only exists in the Setter. Required —the default— asks for it and won't book without it. Optional asks for it once and, if the person would rather not share it, books anyway. Don't ask books with the name and phone number. If the lead's record already has an email, the booking uses it in all three cases. Without an email, the person gets no calendar invitation and no booking emails, and the Setter doesn't promise them: before booking, it confirms the day, the time and where it is right there in the chat. If the event it books has the WhatsApp confirmation and reminders on, those do reach them, so turn them on: Automatic booking reminders on WhatsApp. It's meant for in-person appointments and for people who won't leave an email; the self-booking link the Setter can send does ask for the email on its form.

Two things to check if you choose Optional or Don't ask. If the event it books —the one in Event it books, in Channels— is a video call without the Immediate WhatsApp confirmation to the booker, no automatic message gets the link to a person without an email: the field itself warns you, and its Go to Team link takes you to where you edit the event. Turn on the confirmation there —or, if the appointment is actually in person, change its Location to In-person meeting—. And make sure neither the Criteria nor the knowledge —the position's or the agent's— tell it to ask for the email: that text competes with this setting.
The cheapest way to know what your lead will experience is to experience it yourself: the Test tab lets you actually talk with the position, with its same brain, its same data and its same booking rules —with Email on Required, it won't book a test lead without an email either—, without sending anything to anyone.

How do I turn it on without surprises?

  • Activate it on a single source and a single line.
  • For the first three days, read all the real conversations.
  • Adjust the qualification criteria, the required fields and the knowledge files.
  • Only then expand.
Tell your team before turning it on, and explain what happens if someone steps into a thread the AI is working. Inside MB Suite there's no risk of stepping on it by accident: the text field is locked with the notice "Conversation handled by AI — tap to take control", and when you tap it a modal explains that you become the deal's owner and the AI stops working it. From the WhatsApp Business app on the phone it's different: if you reply there, the AI steps away from that conversation for a day and resumes on its own if you stop replying. And if you just want to stop it for a while, use Pause AI in that prospect's or deal's AI Log, instead of taking control.

About cost, plainly. Consumption is counted per turn — each time the position thinks and acts —, not per message. The only thing that cuts a lead's spend is the position reaching an end — booked, disqualified or handed off —, and that's governed by the two caps in Limits. If something gets out of hand, the switch in the AI Team table sets it to Paused instantly: that's the brake. To follow consumption: See your AI usage and control your credits.

The included credits, charges and terms of your subscription are governed by the Terms and Conditions, which always carry the current version.

How do I check it's set up right?

Send a new message to the line the position works, from a number that isn't in your database.

  • The position replied.
  • The reply is correct and sounds like your company.
  • It became the owner of the prospect or the deal.
If it didn't reply, open the prospect or the deal and look at its AI Log: there you see whether the position ran and why it didn't act. Without that, "it didn't reply" is indistinguishable from "it will reply later". If, under Next Run, it says "reason: incoming message to handle" or "reason: first turn pending", it's the second case: the message is already saved —or the lead was just assigned— and the position acts on its next pass, within a few minutes.

What if it does nothing?

What you seeWhat's missing
It does absolutely nothingThe source is missing, the assignment rule that sends it the lead is missing, the position is Paused, or billing is blocked
It hands everything to a human right awayIt has no templates to open a conversation, it has too many required fields, or it gets leads without a phone with the Email channel off
The Setter doesn't book because the lead won't give their emailThe Email field, in Goal, is on Required. If your appointments don't need it, switch it to Optional or Don't ask
It replies the next dayThe time zone in Schedule is wrong, or Behavior on incoming messages was left on Business hours only
The customer gets two different repliesA Sequence was left running; if it takes in contacts on its own (a linked list or a trigger in its Start block), cut that off
It stopped working a leadSomeone took control of that thread from the inbox, or it was left Paused in the AI Log. Reassign the deal to the AI, or use Resume AI

What's next?

Everything is configured now. What's left is checking it end to end: Check that your CRM was set up correctly: the final tests.

Still have questions?

Ask MIA, the MB Suite AI assistant: open it with the MB AI button (⌘J) in the top bar. MIA knows the section you're in, so you can ask it about what you see on screen — and it can also answer how-to questions about using MB Suite.