What your bot finds out

Your bot is a salesperson, not a form. It does not read your questions out in order. It knows what it needs to find out, why each thing matters and when it applies, and it picks the one question worth asking next from what the lead has already said. When a lead cannot answer, it helps them work it out instead of asking again.

Open it at Business setup → What your bot finds out (the page that used to be called Intake questions). Adding a question, choosing its type and its choices are in Intake fields; this article is about how your bot is led by what you write there.

The page

Your questions sit in three groups. Every question is in exactly one, so nothing is hidden.

Each row leads with three things you decide, and the rest waits behind More:

1. The question, in your words (with its type beside it, and arrows to move it up or down in its group).
2. Must know before your team calls, a switch. On means the conversation cannot reach a verdict until it is answered or settled. Beside the question's name a badge says Required (switch on) or Optional (switch off), the same two counts the page heading gives.
3. Why you ask, a sentence you can edit in place.

More opens the rest: If they're not sure, How much it matters, Only ask when…, Sensitive, and for amounts, Money basis. The placeholder (example text in the box where your team types this answer on a lead's page; your bot never reads it) and, where they apply, the minimum, maximum and maximum length also live there. Those limits apply where your team types an answer on a lead's page; they do not check what a lead tells your bot. Most owners never open it.

The panel beside the list, What leads will see, previews the first questions, what your bot does when a lead says "I'm not sure", and the read-back. It is built from fixed wording around your own words, so it is the same every time. The real reply is worded fresh in each conversation.

Why you ask

Write the reason, not an instruction. "So we can check what's available when they need it" is a reason; "Ask when they want to travel" is a script, and a script makes the bot repeat itself. Given the reason, your bot words the question itself, follows what the lead just said, and shapes any help it gives around what you are trying to learn.

Your template, and AI setup, write a reason for the questions they create. Leave it blank and your bot falls back to the older AI prompt hint (still in the question's edit form); without either it has only the question itself to go on.

The reason is for your bot. It is never read out to the lead, and your bot never copies it.

How much it matters

Behind the scenes your bot also weighs a question that feeds a hard rule or your matching, and one it has just asked, so it does not circle back to the same thing. You do not set that.

Every applicable optional question is asked of every lead, with a few exceptions: an optional question marked Sensitive is never asked, Safe mode asks only the must-knows, and a lead who asks to see the options or to book a time, or who signs off, is not kept waiting on optional questions (one that a hard rule reads is still asked). Your bot spends few messages doing it: it pairs two quick questions (a number, or a choice of four or fewer), puts a "nice to have" beside another question, and does not repeat something the lead already said in a long message. A lead who answers everything in one message gets few follow-ups.

Whatever your bot plans to ask, it answers what the lead asked first. A lead who asks the same thing twice gets that answered and nothing else.

Only ask when…

Some questions only apply to some leads. Under More, Only ask when… takes a sentence like Only when Mode is one of In person, Either. Tutoring uses it so an online student is never asked for a location.

• Until the question it depends on is answered, the question waits.
• Once the answer rules it out, the question is skipped for good and nothing waits for it, even if you marked it must-know.
• A question cannot depend on itself.

The core questions

Most businesses need the same answers, so the platform keeps one list of them. The list below is the whole set the platform knows; your template may use only some of them, so your page can show fewer. A question that is one of them is tagged core under More. Templates and AI setup fill in a reason and an "If they're not sure" line for it, which you can change. Each core question can be used once.

You can rename any of them in your own words. The tag stays with the question, not the label, so renaming changes nothing about how your bot treats it.

You can also switch one off from the row's menu. A switched-off question is dimmed and marked Switched off:

• your bot never asks it, and never waits for it, even if it was a must-know;
• a rule or a matching comparison that uses it pauses with it;
• answers already collected are kept, and your team still sees them;
• Switch on from the same menu brings it back, with nothing lost.

Switch off, rather than delete, when you might want it back. The Notes question is a closing line your bot adds after the last real question ("anything else I should pass on?"); it never takes a turn of its own.

When a lead is not sure

Not every lead knows which specialist they need, which site is nearest, or what a term means. For such a lead your bot does what a person would, in five steps:

1. Asks about their situation instead of repeating the question.
2. Suggests the kind of option that fits, in plain words, and checks.
3. Offers a simple choice in words when an open question is hard, from what is described.
4. Falls back to your safe default when it is still unclear.
5. Explains a term in a sentence when they may not know it.

Lead: My knee hurts when I run. Not sure who I should see. Bot: Sorry to hear that. It sounds like a job for a physiotherapist, who looks after joint and muscle problems. Shall I look for one? Lead: I don't really know if that's right. Bot: That's fine. A GP can do a first assessment and refer you to the right specialist. Would you like to book that instead?

That is the clinics template's Specialty question: its description of Physiotherapy says "muscle, joint and back pain", and its If they're not sure text says to suggest a kind of practitioner, never name a condition, and offer General Practice as the fallback. Your own questions work the same way once you write their If they're not sure line.

In Safe mode your bot suggests nothing: it says in one sentence what it is asking, asks again simply, and carries on if the lead still cannot say. Two rules hold in every business:

• A suggestion is not an answer. Your bot records it only after the lead agrees. Until then it is held: your team can see it on the lead as Suggested, waiting for their yes. "Ok" to something else, or carrying on with a different subject, is not agreement.
• Your bot never invents a figure. It will not offer a price, a range, a band or a quantity that you did not give it and the lead did not say. When a lead is unsure about money, it asks whether they have a rough figure in mind or what the amount has to cover.

Your bot asks the question plainly first. It uses If they're not sure only once the lead has said they are not sure.

Describing your options

A choice question is only as helpful as the descriptions of its choices. Under Business setup → Categories, Edit beside each item and add a Description, up to 300 characters:

• for an option, what it means to a customer in everyday words ("Muscle, joint and back pain, sports injuries, recovery after surgery");
• for a place, the areas it serves, so your bot can suggest the nearest.

Your bot uses descriptions to explain an option and to suggest the right one. The templates write them for you. Check anything marked as a sample address, and write real ones. A description is a guide for the bot, not text it pastes into a message.

Sensitive questions

Mark a question Sensitive when the answer is private, such as the reason for a medical visit.

• Your bot never reads it back, puts it in the running summary or the notes it keeps for your team, or repeats it to the lead later, with one exception: when the lead asks what they told the bot, it says their own answer back on that reply (how). Your team still reads it on the lead's page.
• If the question is optional, your bot does not ask it at all; it is recorded only if the lead brings it up.
• If you made it must-know, your bot does ask, gently and on its own. In its reply to the message that gives the answer it may acknowledge it briefly, because it needs it to suggest the right practitioner. After that one reply it never mentions it again.
• A team alert for a qualified lead leaves the answer out and says "Some details are private and are shown only on the lead's page".

Money basis

For an amount, Money basis says what it covers: per person, per unit, in total, per hour, per day or per month. The read-back uses it ("about 25,000 each", not just "25,000"). The templates set it for you.

Where each answer came from

Every answer on a lead remembers how your bot got it. On the lead's page, and in the inbox panel, a small badge sits beside it.

Under the answers, Also mentioned lists things the lead said that none of your questions asks about: an anniversary, a dog, "my boss decides". A fact the lead repeats in other words is kept once, in the newest wording. "Asking for someone else" appears there when the lead is writing on another person's behalf. The legend appears on the lead only while it has a badge to explain.

The read-back

Before it decides, your bot reads the key details back and asks if anything needs changing, but only when something was guessed or suggested: an answer is Inferred or Suggested, agreed, or a hard rule is failing. When every answer is stated, a read-back would only repeat what the lead just said, so it is skipped and costs no message.

It happens once per enquiry. A correction changes the answer and is not read back again. Sensitive answers are never in it. If the lead carries on without answering it, your bot does not ask again and the guessed answers stay Inferred.

Guesses never turn a lead away

A lead is never turned away on something your bot only guessed. If an inferred answer fails a hard rule, the read-back goes first. If the lead confirms it, it counts like any answer. If they carry on without answering, it is treated like a negotiable miss: your team sees it and the lead is not marked Lost. See qualification rules.

What it does not do

• It does not follow the order of the list. The order on the page only settles ties between questions that are equally worth asking. Use Must know, How much it matters and Only ask when… to steer it.
• It does not answer from general knowledge. It explains an option from your descriptions.
• It does not decide for the lead. The rules still decide the outcome; the questions only decide what the conversation is about.

To see all of this at work, open the playground, say you are not sure, and read the lead's page afterwards. The rest of the conversation is in How the bot builds its conversation.