Email, Ratings & Agent Handoff
How visitors reach a human, what happens when your team is offline, and how Yaplet captures email addresses, sends transcripts, and collects conversation ratings.
Overview
Most conversations start with Yaplet. This article covers what happens when a visitor needs more than the AI can give — how they reach your team, how you stay reachable when nobody's online, and how the conversation wraps up.
Requesting an Agent
There are three ways a visitor asks for a human. Each has its own switch on Brand → AI agent → Human hand-off.
They ask from the header menu
The ⋮ menu in the chat header offers "Request agent" whenever nobody from your team has the conversation yet. This is the escape hatch, switched by “Request agent” in the chat menu. It stays there even with Visitor asks Yaplet for a person switched off, and then Yaplet points visitors here when they ask for a person.
The ⋮ button only appears when it has something to offer: "Request agent" while nobody from your team has the conversation and the visitor hasn't already asked, AI Mode whenever nobody is actively answering and Yaplet can answer (see Changing their mind below), and "Edit email" while the visitor hasn't given an email. While an agent is answering and the visitor's email is on file, the header shows no ⋮ at all.
A button appears automatically
Once a conversation has gone back and forth enough times without resolving, a "Talk to the support team" chip surfaces on the left under the last reply. It's a nudge for visitors who are clearly stuck but haven't thought to ask for a person. Its switch, “Talk to support team” button, also sets how many messages the visitor must send within five minutes before it appears.
They type it
A visitor can simply write that they want to talk to a person. This works while Visitor asks Yaplet for a person is on. With it off, Yaplet connects nobody and points the visitor to "Request agent" in the ⋮ menu instead.
Yaplet never offers a person on its own. The one automatic route is yours to choose: with When Yaplet has no answer set to Connect to a person, a question Yaplet can't answer from your knowledge connects the visitor straight away. When nobody from your team is online, Yaplet tells the visitor that nobody is available right now and the "Leave a message" card described below appears, so the question still reaches your team.
What the Visitor Sees After Requesting
Yaplet tries first
Let Yaplet try first is off unless you switch it on. With it on, a request for a person doesn't go straight to your team. Whichever of the three ways the visitor asked, Yaplet first makes sure it knows what they need and checks whether it can settle the issue itself. This only happens in the chat widget: on Messenger, Instagram and email the conversation is handed over directly.
What happens next depends on what has happened in the conversation so far. When Yaplet has to check the issue, it takes a few seconds: the composer is locked and a waiting row with animated dots reads "Preparing your request…", then "Just a moment…". If the check takes too long, the visitor is simply handed over.
If the visitor hasn't said what they need yet (only a greeting or small talk, just "I want to talk to a person", or Yaplet's last answer was more than 30 minutes ago), a "Request a person" card opens at once with the line "Tell us briefly what you need, so we can help faster.", one text box and a check-mark send button. Enter sends too; Shift+Enter starts a new line. The × closes the card, but nothing else skips it: pressing "Request agent" in the ⋮ menu or the button again only puts the cursor in the card's text box. What they send appears in the chat as their own message, and Yaplet checks it. One of the next two things then happens.
If Yaplet has just answered a question of theirs (in the last 30 minutes) and they ask for a person without writing anything new, from the ⋮ menu, with the button or by typing it, there is no card to fill in and no check: they go straight to a person, as listed below.
Yaplet answers right away in the chat, as a normal answer: typed out, with links and product cards where they fit. There is no promise sentence and no button for a person next to it, and the visitor doesn't have to write the issue again. If they ask for a person after that, they always reach one, as described under Yaplet just answered.
The visitor is handed to a person at once. That happens when the issue is a status the visitor wants changed or only your team can change (a stuck or late order, a held payout, a pending review, a refund), or when the only knowledge Yaplet finds says to contact support. Your team sees what Yaplet found as a private summary under the "Agent requested" line in the Inbox. If Yaplet starts to answer and finds nothing after all, it hands over at once too, and so does a typed request that Yaplet could help with but not settle. The hand-off goes ahead as described below. The one exception is a typed request where Yaplet judged that a person is needed: Yaplet first asks in its own words whether to connect them, with the plain "Yes, connect me" and "No, keep chatting" chips. When nobody is online, Yaplet instead says briefly that nobody from the team is available right now and asks whether they'd like to leave a message; the first chip then reads "Yes, leave a message" and brings up the "Leave a message" card. Only a yes hands the conversation over. The question is saved in the conversation, so your team sees it in the Inbox, and it stays in the thread if the visitor keeps chatting.
These cases skip the cards and hand over at once:
- Yaplet can't answer this visitor right now: your plan's AI allowance is used up with no bought credit, the visitor's AI is paused, or they reached the daily answer limit.
- One of the brand's switched-on workflows that hand over to a person fits the issue.
- The visitor asks again while the Yes/No question is still open.
- The visitor asks for a person right after one of Yaplet's answers (within 30 minutes, nothing new written), including the answer Yaplet gave to their own request, by tapping or by typing. A typed request gets no Yes/No question here: Yaplet just tells them that someone from the team will join, or, with nobody online, that their request has been passed on. The ⋮ item and the button hand over without that line.
Every button the visitor taps on the Yes/No question ("Yes, connect me", "No, keep chatting") appears in the conversation as their own message, so your team sees in the Inbox what they chose.
The hand-off
The conversation now waits for your team. An "Agent requested" status event appears in the thread, the header's status line reads "Waiting for the team" (or "The team is away" when nobody is online), the composer's placeholder becomes "Write your message for the team…" and conversation starters are hidden. When someone from your team is online, Yaplet posts the handoff line, in the visitor's language:
"A team member will join this chat."
That sentence is Yaplet's own, and it never promises a reply time. It is left out when Yaplet has just said the same in its own words. Only a workflow's Request agent step, or an API tool that hands over with its own message when it fails, can say something else. With nobody online the sentence is left out, because nobody is there to join; a workflow's or API tool's own text still shows.
What comes next depends entirely on whether anyone is online.
The conversation sits in your Inbox waiting for someone to pick it up. If nobody from your team has taken it or written to the visitor 5 minutes after the request, the visitor gets the "Leave a message" card, titled "The team hasn't been able to reply yet".
The visitor gets the "Leave a message" card at once, titled "The team is away right now".
With team hours set (Brand → Brand settings → Team hours) and the request coming outside them, the card also says when your team is next usually online, in the visitor's language and time zone:
"Usually back Monday 09:00"
The time reads "today 17:00", "tomorrow 09:00" or a weekday and time. When the visitor's time zone is not known, it is given in your team hours' time zone, with that zone named. Inside your team hours, or with none set, no time is given — the card never suggests your team should be there now.
The "Leave a message" card
Under its title the card says "Write everything you need below — it goes straight to the team, and they'll reply in this chat." The chat itself is the message: the composer stays open under the card, everything the visitor writes reaches your Inbox, and nothing is lost.
- The email is optional. The field "Leave your email so they can also reach you by email" has a Save button and is already filled in when Yaplet knows the address. After saving, the card shrinks to one line, "The team can also reach you at …", with a change link. Closed with the × and no email given, it reads "You'll see the reply here next time you visit."
- "Or keep chatting with the AI" appears only when Yaplet can answer this visitor right now: Yaplet answers on the website chat, the visitor's AI isn't paused and they haven't reached the daily answer limit.
- The card stays up through a reload until the visitor saves or closes it. An older card turns into a plain gray line with its title.
The 5-minute card works the same for every brand and has no setting. It appears only in the website chat, never in email conversations or phone calls, and not when the visitor already got the "team is away" card for that request. If the brand has a switched-on workflow that starts on Nobody picked up (see How a workflow starts), that workflow speaks instead. At the same moment your workspace owners get an alert that nobody has answered.
When Yaplet doesn't have the visitor's email yet, the notice asks for it too. With team hours set and the request outside them, it adds "Our team is usually back Monday 09:00.""Sorry for the inconvenience, all our agents are currently offline. Please leave your message, and we will get back to you as soon as possible."
Who counts as "online"
An agent counts as online when their profile status is Online, they are still enabled in the workspace (a member you Disable never counts) and they have been granted access to this widget's brand. Access is granted per brand, not per widget — in the Brand access section of a member on the team page, which you open from your avatar menu at the top right under Manage team, or in the Inbox access list on the Chat widget's Chat tab. Owners and admins always count.
Changing their mind
Whenever nobody from your team is actively answering, a "Talk to AI Bot" chip under the thread (and an AI Mode entry in the ⋮ menu) sends the visitor back to Yaplet. That is the case while the visitor waits with nobody in the conversation yet — the card's "Or keep chatting with the AI" does the same — and when the team member who has the conversation hasn't replied for 5 minutes since the visitor's last message; going back then takes the conversation off that team member. Either way the request is canceled and the visitor sees an "Agent request canceled" status event, so your team knows the visitor didn't need them after all. When the conversation was taken off someone, your team also sees the private line "The visitor went back to the AI".
These options only appear when Yaplet can answer: not when no AI answers on the website chat, nor when the visitor's AI is paused or they reached the daily answer limit. If the switch fails, the visitor reads "Couldn't switch back to the AI. Please try again." On Messenger and Instagram, the menu's "Talk to the AI bot" follows the same rules.
When the team member who has the chat is offline
A visitor sometimes writes into a conversation whose team member is no longer online: they switched to Offline, were disabled, or lost access to the brand. If that team member also hasn't written in the conversation for 5 minutes, the conversation doesn't wait for them:
- Yaplet can answer: Yaplet takes the conversation and answers the message. Your team sees the private line "<name> was offline — the AI took over".
- Yaplet can't answer: the conversation goes back to your team's queue as a normal request, exactly as if the visitor had just asked for a person. Your team sees "<name> was offline — the chat is waiting for the team".
A team member who wrote in the conversation less than 5 minutes ago keeps it whatever their switch says, because people often answer without switching to Online. This works in the website chat, on Messenger and on Instagram; email conversations always stay with their team member.
Email Capture
An email address is what turns an anonymous browser tab into someone you can actually follow up with. Yaplet collects it in four ways:
| Route | When it happens |
|---|---|
| Ask for an email switch | Turned on, the email form blocks the composer before the visitor's first message |
| "Leave a message" card | An optional email field, when nobody's online at the request or nobody replies within 5 minutes |
| The ⋮ menu | "Edit email" — visitor-initiated, and only shown while no email is on file |
| Workflow step | A custom workflow collects it as part of a guided flow |
The form is a card in the AI's place asking "What is your email address?", with one field on the page color and the submit check inside it; the format is validated before it is accepted.
Once an email is on file, the form never appears again.
Transcripts
Transcripts are sent by your team, not from the widget — the ⋮ menu has no "Send Transcript" entry. A visitor who wants the conversation by email asks your support team, and an agent sends it from the Inbox with Send transcript — see Inbox Sidebar Actions. It goes only to the email address the visitor gave.
This is genuinely useful after a long troubleshooting session — the visitor keeps the steps you walked them through. The visitor sees a status line in the chat — "Transcript was sent to …" — naming the address it went to.
Conversation Ratings
An agent can ask the visitor to rate the conversation at any time with Ask for rating in the Inbox sidebar — see Inbox Sidebar Actions. With Ask for a rating on, the widget also asks automatically when the agent leaves the conversation with Leave chat or Leave chats… — see Chat Actions. The agent may first send the visitor a short goodbye message; it arrives as their normal reply, just before the rating request, and doesn't change whether the request goes out. That automatic request only goes out when the leaving agent wrote at least 2 replies since they took the conversation and the visitor did not write last — a quick hand-over, or a visitor still waiting for an answer, is not asked. Once the visitor has rated an agent, that agent can't ask again in the same conversation for 7 days; if the visitor rates again within those 7 days, the new score replaces the earlier one.
The rating form shows five faces:
| 😡 | 😟 | 😐 | 😊 | 😍 |
|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 |
The card sits in the AI's place; the picked face is tinted and grows a little while the others dim. Below them is an optional free-text field, "Please provide additional feedback (optional)", with the submit check inside it. The question above the faces — "Please rate your experience" — is Yaplet's own, shown in the visitor's language.
Both the score and any written feedback land in the conversation as status events, and feed your Live Chat reports and the Ratings tab of Agent monitoring.
Putting It Together
A well-tuned setup usually looks like this:
- Ask for an email off — let visitors start instantly; the "Leave a message" card asks for the email when it's actually needed.
- Online means answering — with nobody online, or no reply within 5 minutes, visitors get the "Leave a message" card instead of a person. Switch to Online when you're really there, and pick up waiting conversations quickly.
- Ask for a rating on — it's one tap for the visitor and the only structured quality signal you get.
- Yaplet answering first — so your team only sees the conversations that genuinely need them.