Log inStart Free Trial
Meta Tech Provider

Notify your team when a WhatsApp chat needs a person

To notify your team when a WhatsApp chat needs a person, catch the signal in an automation, such as a customer writing "human", and add a Hand to a person step: it pauses the bot on that chat and emails the account owner. To reach Slack or a team inbox, post an alert from a webhook step. When a team member replies from the app, the automation pauses.

easyCoexistence connects the WhatsApp Business number your team keeps in the app to Meta's Cloud API through Coexistence, so your own AI and automations answer while people stay a reply away.

4 hours

Default pause on a chat after someone replies from the phone; set 0 to 168 hours under Advanced.

1,380 min

Longest Wait step, 23 hours; on resume the run is skipped if someone answered from the phone meanwhile.

30

Keywords a Has keyword trigger takes; any one is enough and capital letters do not matter.

6 hours

Shortest gap between two failure emails for the same automation, sent when a run fails on a customer's message.

{"text": ...}

Body of a Slack incoming webhook message: a JSON POST with a text field.

Which signals mean a chat needs a person?

Four signals cover most hand-offs, and each can start an alert. Two come from the customer's own words, one from your AI and one from easyCoexistence itself. Pick the ones that fit how your team works, then send them to the place the team already looks during the day:

  • The customer asks for a person: human, agent, atendente, falar com alguém, persona, asesor.
  • Words of a problem: refund, complaint, broken, reembolso, reclamação, devolución.
  • Your AI hands off: an agent's hand-off tool pauses the bot and emails you by itself, and a Reply with AI step can end with the fixed sentence you told it to use when it cannot help.
  • A run fails, for example your system did not answer: easyCoexistence emails you without any setup.

How do you alert the team when a customer asks for a person?

With a keyword automation. Trigger: A message arrives, Has keyword, Contains, with the words customers use in each language. Steps: Reply with text, so the customer knows someone is coming, then Hand to a person, which pauses the bot on that chat and emails the account owner with the customer's message; Notify me emails without pausing. For Slack or a team inbox, add Call my webhook to the workflow that posts the alert, or Call an API with a Slack incoming webhook URL. Keyword automations are checked before the others, so an AI assistant on the catch-all stays quiet on that message. Contains matches inside words: human also fires on humano, which helps, while person would fire on personal, which does not, so test the list in Try it. An assistant connected through MCP can create it with the JSON below.

create_automation: confirm to the customer, alert the team
{
  "connection_id": "3f2a9c1e-7b4d-4e8a-9c2f-1a2b3c4d5e6f",
  "name": "Ask for a person",
  "enabled": false,
  "trigger": {
    "type": "message_received",
    "match": "keyword",
    "keyword_mode": "contains",
    "keywords": ["human", "real person", "talk to someone", "atendente", "falar com alguém", "humano", "asesor", "hablar con alguien"]
  },
  "steps": [
    { "type": "reply_text", "text": "Got it, {{first_name}}. Someone from our team will reply here shortly." },
    { "type": "call_webhook", "url": "https://n8n.example.com/webhook/whatsapp-handoff" }
  ],
  "settings": { "takeover_hours": 12 }
}

How do you catch an AI hand-off?

With an Agent answers step there is nothing to build: when the agent calls its hand-off tool, the bot goes quiet on that chat for the agent's hand-off hours (24 by default) and you get an email with the reason. A Reply with AI step has no hand-off tool, but it already carries a rule, added under your instructions, to say a person from the business will answer soon when the customer asks for one or the model cannot help. Make that sentence exact: write in your Instructions "When you hand off, end with: Someone from our team will reply here shortly." Then add Call my webhook as the step after Reply with AI. The AI's reply is saved before the webhook runs, so it is the last entry of history, with direction out.

Your workflow checks that entry for the sentence and posts the alert only then; otherwise it answers 200 with no reply and does nothing. The webhook runs whenever this automation answers, so the filter belongs in the workflow.

How does the workflow post to Slack or send an email?

In n8n: a Webhook node with Respond set to Immediately, since no reply is needed; a Code node that verifies X-EC-Signature with the automation's Signing secret and builds the alert; an IF node that drops tests (dry_run true) and, for the AI route, messages without the hand-off sentence; then an HTTP Request node posting to a Slack incoming webhook, or a Send Email node. Put the customer's name, number and words in the alert, plus a link to the automation's runs, so whoever picks it up knows what was said before opening the chat on the phone.

n8n Code node after the signature check: build the Slack message
const b = $input.first().json.body; // verified Call my webhook body
const last = b.history[b.history.length - 1];
const handedOff = b.message.text && last && last.direction === 'out'
  && last.text.includes('Someone from our team will reply here shortly');

return [{ json: {
  skip: b.dry_run || !handedOff, // the keyword automation can drop the handedOff test
  slack: {
    text: ':raising_hand: WhatsApp chat needs a person\n'
      + '*' + (b.contact.name || 'Customer') + '* (+' + b.contact.wa_id + ') on ' + b.channel.phone + '\n'
      + '> ' + b.message.text + '\n'
      + 'Runs: https://easycoexistence.com/dashboard/automations/' + b.automation.id,
  },
} }];
// Next: IF skip is false -> HTTP Request, POST, JSON body {{ $json.slack }}, to your Slack incoming webhook URL

What happens when a team member replies?

Their reply from the WhatsApp Business app reaches easyCoexistence as an echo, and one from the Conversations screen is known at once; either way automations step back on that chat for the hours set under Advanced, in "Pause on a chat after I answer it (hours)", 4 by default. The pause is read from the automation that would answer next, so raise it on your AI assistant on the catch-all, not only on the hand-off automation; 12 or 24 hours suits chats that a person takes over.

Until someone replies, the AI keeps answering new messages, unless a Hand to a person step or an agent's hand-off has paused the chat: the bot then stays quiet for its hours, 24 by default, or until you resume it in Conversations. A short first line from the phone, such as "Hi, it's Marta, I'm on it", claims the chat at once. The customer sees one conversation, from the same number, with no change of app.

How do you escalate only when nobody answers?

Add a Wait step and a second webhook. The steps become Reply with text, Call my webhook (first alert), Wait 15 minutes, Call my webhook (escalation). When the run resumes, easyCoexistence checks whether someone answered the chat from the phone since the run started; if so, the rest is skipped and the run shows "You answered this chat from your phone recently". Only unanswered chats reach the second alert.

Give the two steps different URLs, such as the same workflow path with ?stage=1 and ?stage=2, because the body does not say which step called. If the customer writes a hand-off word again, the waiting run of the same automation is superseded and a fresh first alert goes out.

Which alerts does easyCoexistence send by itself?

Three, by email to the account owner, with nothing to configure. When a run fails on a customer's message, for example because your webhook did not answer in 20 seconds, an email names the customer and the message so someone can answer from the app; at most one per automation every 6 hours, and the Runs tab keeps every failure. When an AI provider refuses your key, one email says which automations paused. And the number itself is monitored around the clock: disconnection, the 14-day rule for opening the Business app, and quality changes. Steps you add send email too: Hand to a person, and Notify me, at most 20 an hour per automation.

How to set it up

  1. Build the alert workflowIn n8n, Webhook node (POST, Respond Immediately), Code node with the signature check and the alert text, IF node on skip, then HTTP Request to your Slack incoming webhook or Send Email. Activate it and copy the Production URL.
  2. Create the hand-off automation, switched OffNew automation, Keyword reply under Basics, Contains, your hand-off words. Keep Reply with text, add Hand to a person, then Call my webhook with the Production URL, set the switch to Off and click Save.
  3. Copy the Signing secretSelect the Call my webhook step of the saved automation and paste its Signing secret into the Code node. New secret makes another one if it ever leaks.
  4. Raise the pause on the AI assistantOpen your catch-all AI automation, Advanced, and set Pause on a chat after I answer it (hours) to the time your team needs.
  5. Test, then switch OnClick Test and run "can I talk to a human?": the confirmation shows and the workflow runs with dry_run true. Switch On, save, and try it from another phone.

Common mistakes

  • Using fragments such as "person" or "help" in Contains mode without testing. They fire inside personal, personalized or helpful, and the team gets alerts for ordinary questions.
  • Leaving the hand-off sentence to the model's wording. Without an exact sentence in the Instructions, the workflow cannot tell a hand-off from a normal answer and misses alerts.
  • Pointing the alert and the escalation at the same URL. The body does not say which step called, so the second alert looks like a duplicate of the first.
  • Raising the pause only on the hand-off automation. The AI catch-all applies its own setting, so after 4 hours it answers the customer again while the team is still on the case.
Doing this with EasyCoexistence

Hand to a person and Notify me steps email the owner, a Call my webhook step sends hand-offs to n8n, Slack or email, a reply from the phone pauses automations on that chat, and failed runs and number problems are emailed to you. US$ 9 down to US$ 0.75 per number per month, falling as you add numbers, with the first 3 days free.

easyCoexistence is a multichannel API and MCP server that connects AI agents such as Claude and ChatGPT to WhatsApp, Instagram, TikTok, Facebook and more. On WhatsApp it uses Coexistence: the business keeps using its number in the WhatsApp Business app while the AI works through Meta's official API.

Frequently asked questions

Can the alert arrive on WhatsApp instead of Slack?

Your workflow can send an approved utility template to a staff member's own phone through Meta's Cloud API, with the send endpoint and token from get_api_credentials. Staff must agree to receive these messages, and Meta bills each template to your WhatsApp Business Account.

Can several people receive the same alert?

Yes. Post to a shared Slack channel or an email group, or add more nodes after the IF in the workflow. Whoever replies first from the WhatsApp Business app takes the chat, and the reply is visible to anyone else using the app on that number.

Can the team answer the customer from Slack?

Not from Slack: it only gets the alert. Replies go from the WhatsApp Business app on the phone, from the Conversations screen in the dashboard, where teammates you invite to the account can answer too, or from your own system through the Cloud API. Messages sent through the API also show in the app's chat thread.

Do alerts add to the bill?

No. Automations have no per-run charge and no cap. The confirmation text the customer receives is a WhatsApp message, billed by Meta to the business's own WhatsApp Business Account by category and country when Meta charges for it, never marked up by easyCoexistence.

Keep reading

Ready to get started?

Set up WhatsApp Coexistence in minutes, not months. The app keeps working on the phone.

Start Free Trial

Verified on