Emergencies and safety

A lead writes "I can smell gas in the kitchen" or "my husband has collapsed and isn't breathing". Those are not intake answers, and a friendly question about their budget is the wrong reply. So your bot does not decide what to say. A fixed message is sent by the product itself, before the AI runs, and your team is told at once.

The bot is not a substitute for the emergency services and never claims to be one. The message sends the lead to their local emergency number. Your team's job is to follow up.

What happens, step by step

When a lead's message contains an emergency phrase:

1. The lead gets one fixed message straight away. No AI writes or changes it.
2. An Urgent flag opens on the lead, so it shows under Needs you and on the bell. This happens even if you switched off Flag urgent leads in Settings: that switch is for the AI's own judgement of "urgent", not for a lead who has written that someone is in danger.
3. Your team is alerted the way it is for any lead waiting on a person: on every plan in the Inbox and the bell, and on Pro by email or WhatsApp at the addresses in Settings, with the lead's message quoted. One alert when the flag opens, and the one reminder you set under Remind your team after (minutes) if nobody has answered. The email and WhatsApp alert follows the Tell the team when a lead needs a person switch; the flag and the bell do not.
4. The intake stops for this enquiry. The bot asks no more questions, gives no verdict, does not sell, show options or book. Every later message the lead sends gets a short fixed follow-up that points them back to the emergency services, again with no AI. It is worded differently each time it repeats, so the lead is not sent the same paragraph twice. A second emergency message gets that follow-up too, rather than the first message word for word.
5. Nothing else changes. The lead keeps its stage, the owner of the conversation and the bot's on/off state, and nothing is scheduled afterwards: no follow-up, no finish-the-intake timer.

What lets the bot carry on. Only your team: press Take over on the lead, and press Hand back to your bot (it shows your bot's name) when it is right to continue. Nothing the lead writes lifts it, not even "it's all fine now, back to the quote", because the bot cannot check that and a wrong guess would ask a person in danger about their budget. An emergency belongs to one enquiry: a lead who comes back after the gap you set under Welcome a lead back after (hours) and says it is something new is welcomed back and starts fresh, and a new emergency message stops the intake again. A lead who asks the bot to stop is still opted out.

Handing back also closes the waits the hold opened (Urgent and Possibly a minor), so a lead who was choosing a time or an option can carry on and book. If the lead wrote that they are under 18 while your team had the conversation, handing back does not lift that: the bot reads what they wrote in the meantime, keeps the Possibly a minor flag open and stays on the one fixed message.

Three things follow from "before the AI runs":

• It costs no AI reply. A workspace that has used up its month's AI replies still sends it.
• It is sent even when your team owns the conversation. If a person took over the chat, or the lead was handed off, or was marked Lost or Not a fit, the message still goes. Life safety comes before who is handling the conversation.
• It reads the whole run of quick messages, not only the last one. "He has collapsed" followed a second later by "please hurry" is one emergency.

It does not go out in two cases: a workspace whose subscription is locked, and a lead who opted out by asking the bot to stop.

A lead who was closed for not replying and then writes an emergency is reopened first, like any message.

Every business has a basic safety net

Whatever you pick below, your bot answers a short list of life-threatening phrases:

• someone not breathing, unconscious or collapsed;
• a heart attack or cardiac arrest;
• severe bleeding, choking, an overdose;
• someone badly hurt, a fall from scaffolding, a roof or a height, "man down";
• someone who swallowed bleach, poison, a battery or a chemical;
• "he's dying", and a lead asking for an ambulance;
• a house, building or site on fire, and a shouted "fire!!";
• the smell of gas;
• a lead saying they want to harm or kill themselves.

The check reads common misspellings typed in a hurry ("wrker badly hurt, fir started") and has a list in each of twelve languages and scripts: English, Hindi (in Devanagari and in English letters), Marathi, Bengali, Tamil, Telugu, Arabic, Spanish, Portuguese, French and German. The lists beyond English and Hindi are shorter and have not yet been reviewed by native speakers, so they may miss an unusual way of putting it. The lists are kept short on purpose. It never reacts to a bare "fire", "gas", "accident" or "dying", because those words are ordinary in most trades ("our business collapsed", "a gas bill", "fire safety training", "I'm dying to try it").

Choose the kind of business

Open Profile and find Emergencies. Kind of business adds the hazards of your trade on top of the basic list.

Your own message is sent only for the kind of emergency it was written for (see What the lead is sent).

Your template chooses for you. Clinics start on Medical, HVAC and home services on Home safety, and every other template on General. Change it whenever you like.

What counts and what does not

The check looks for whole phrases, not meaning. Some examples from a clinic and an HVAC business:

A phrase list can miss an unusual way of putting it, and it does not read negatives: "no chest pain" is not understood as "no". The check leans towards sending the message when in doubt, because a miss is the failure this exists to prevent. A false alarm costs an extra line, a flag, and a paused intake until your team hands the conversation back, so the lists are kept tight.

The past and figures of speech are not emergencies. A sentence that puts the danger well in the past ("we had a gas leak at the office last year", "my father had a heart attack two years ago", "my brother fell off a roof in 2019", "recovering from a stroke"), or that has a laugh in it ("heart attack grill reviews lol"), is not read by the medical, fire and home lists. Nor is a common figure of speech ("this price gave me a heart attack", "our old agency collapsed", "my feet are on fire after the hike", "I'm dying to visit Goa", "I want to die of boredom"). Three things to know:

• If the same sentence also says it is happening now ("and now there is smoke", "today", "again", "urgent"), it fires.
• "Yesterday", "last night" and "last week" do not count as the past, because that person may still be in danger. A clinic's list still fires on "I fainted last week".
• Nothing about someone harming themselves is skipped for being in the past or a joke. "I tried to end my life years ago" still gets the crisis line. Only a few exaggerations that name what the lead is dying of ("dying of embarrassment") are set aside.

The AI keeps its own second net: if it judges that a lead describes an emergency, the reply it wrote is replaced by the standard general message, the Urgent flag opens, and the intake stops just as it does for a listed phrase.

What the lead is sent

The reply follows what the lead wrote, not your trade. A heating company whose lead writes "he collapsed and isn't breathing" does not send its gas-leak message, and a clinic whose lead writes "I can smell gas" does not send its clinic message: your own message is used only for its own kind of emergency. When one message is about two kinds, the person comes first, and a risk of self-harm comes before both: "a worker is badly hurt and there's a fire at our site" gets the message for a person in danger, not the gas message.

The standard messages follow the language the lead writes in (the twelve above; any other language gets English). They always point to "your local emergency number" and never to a specific phone number, because the right number depends on where the lead is, and each ends by saying the message is passed to the team. Your own message is sent exactly as you wrote it, in the language you wrote it.

A lead who may be under 18

A lead who says they are under 18 ("I'm 15", "I'm in class 10", "tengo 15 años") is not asked anything more. The product notices it itself, before the AI runs, and the AI is a second net for the ways of saying it the product does not read. From that message on:

• The lead gets one fixed, friendly message: thank you, the bot can't take their details or sign them up, and could a parent or guardian get in touch. No AI writes it, and it costs no AI reply. Every later message from them gets a short fixed acknowledgement of the same kind, worded differently each time.
• The bot asks nothing, decides nothing and sells nothing. No question, no verdict, no options, no booking. What the lead already said stays in the conversation for your team.
• A Possibly a minor flag opens under Needs you, and your team is alerted once.

The product reads the same twelve languages for this as for emergencies. Only the person writing counts. A parent writing about a child ("my son is 12", "she's in class 8") is an adult, so a tutoring centre or a children's clinic talks to parents exactly as before. A number that is a length of time, a distance or a count is not an age: "I'm 15 minutes away", "I'm 15 years into running this restaurant" and "I'm 15 people short" never start it. An age followed by a comma ("I'm 15, next Tuesday works") does.

When a parent or guardian takes it from there, press Take over, talk to them, and press Hand back to your bot if it should carry on with the parent. Nothing the young lead writes lifts it, and a new enquiry from the same lead keeps it. A lead the bot stopped this way is never followed up by the bot on its own, and is never given a verdict when they go quiet.

Your own emergency message

Under Kind of business, a Medical or Home safety profile shows an Emergency message box (up to 500 characters). The other kinds do not show it. Write it in your voice, for your trade:

If you can smell gas, hear a carbon monoxide alarm, or see sparks or smoke, leave the property now and call your local emergency number or gas emergency line from outside. Please don't switch anything on or off. I'm letting our team know straight away.

• It is sent word for word, in any language, to every lead whose message is about the kind of emergency it covers: a gas, fire or home hazard for Home safety, a person in danger for Medical. Write it so it makes sense to someone in a hurry.
• Leave it blank and the standard message for your kind of business is sent instead.
• Put no phone number in it unless every lead can use it. A number that works in one country fails in another, and a wrong number at that moment is worse than none.
• Say the team has been told only if that is true. Your team is alerted, but nobody promised the lead a call, so promise none.
• A message you saved stays saved if you later switch to a kind of business that hides the box, and it is still sent. Clear it first if you do not want that.

A workspace on Medical or Home safety (the clinics, HVAC and home-services templates start there) is asked to confirm the message in Quick start, on the Your rules step. It is shown in a box with a Use this message for emergencies tick, and setup does not finish until you have ticked it. Ticking saves the text on the screen, blank included (blank confirms that the standard message is the one to send). Check that the number people should call where you are is the right one. Change the text afterwards and the confirmation is cleared until you give it again. After setup the same tick sits under the Emergency message box on the Business profile page, in red until you tick it, and an owner or admin ticks it there without leaving the page. (You do not need to press Save first: the tick saves the text in the box.) The message you saved is the one sent in the meantime; the tick records that you have read it.

Every template that ships one writes it in its voice. On the Business profile page the box shows Not confirmed yet — check it beside it until you tick, whoever wrote the words: editing the message clears the tick too. Read it once and change it if it does not sound like you.

Try it

Emergency replies are on for every workspace, so there is nothing to switch on. Open the playground and write a phrase from your trade, such as "I can smell gas" on a home-safety workspace. You see the fixed message, and the lead gets an Urgent flag, but nobody on your team is messaged, because the playground never alerts anyone. Then try a near miss ("my gas bill is high") and check the bot carries on with the intake.

For more on the Urgent flag and the other reasons a lead waits for your team, see How the bot builds its conversation.