Skip to content

Inbox

Every conversation that needs a person, in one inbox.

When the bot reaches its limit, your team picks up on any channel, with the customer’s records beside the chat, replies drafted from your own documents, and translation into the customer’s language.

A customer's Swahili message in Buni.ai's Inbox is translated into English. The agent asks for a draft, which arrives in Swahili from the utility's English documents with its two sources, and sends it; the customer reads it.

Six chat channels, one queue

WhatsApp, web chat, Telegram, Messenger, Instagram and Slack, with USSD sessions alongside.

Drafts from your documents

Written from your knowledge base, with the passages they came from.

Their language and yours

Read their message in yours. Send your reply in theirs.

Nothing sends itself

Every reply is your agent’s to check, change and send.

Reply drafts

Answers from your own documents, ready to check.

Press Draft and the Inbox writes a reply from the knowledge base your bot uses, plus any documents you keep for agents only. Your agent sees exactly what it was written from, then decides.

  • Shows its sources: the passages each draft is based on, and how closely they match.
  • Says when it doesn’t know. If nothing in your documents covers the question, it tells the agent instead of guessing.
  • In the customer’s language, even when your documents are in English.
  • Shorter, warmer or more formal: regenerate it in another style with one click.
  • Never sends itself. The agent inserts it, edits it and sends it.
A reply drafted in Swahili from a water utility’s English knowledge base, with the two passages it was based on and buttons to insert it, regenerate it or discard it.
An agent’s English reply translated into Swahili before sending, shown above her own words, with a choice to use the translation or keep hers.

Translation

Speak every customer’s language.

Customers write the way they speak. Your agents read every message in their own language and reply in the customer’s, without leaving the conversation.

  • Translate their message into your language, right under the original.
  • Translate your reply into theirs, and check it against your own words before you use it.
  • Their language is remembered from what they have written, so no one has to pick it.
  • Names, places, dates and numbers stay exactly as written.

In the languages your AI model handles, including Swahili, Hausa, Yoruba, Amharic, Twi, French, Arabic and Portuguese.

Customer context

The whole customer, beside the chat.

Agents shouldn’t have to ask what you already know. The customer profile brings in the person’s records from your own data stores, and their history from every channel.

  • Records from your data stores, matched on the phone number or email the conversation carries.
  • One person across channels: see when the same customer has also written on another channel.
  • Past conversations, and the cases raised on them.
  • Internal notes your team keeps private, and a note from the flow when it hands over.
Beside the conversation, the customer’s profile: her WhatsApp identity, and her account and repaired ticket from the utility’s own records, matched on her phone number.
The Inbox queue: conversations from WhatsApp, web chat, Telegram, Messenger, Instagram and Slack in one list, each with its state, its channel and how long it has been open.

One inbox

Every channel in one queue.

WhatsApp, web chat, Telegram, Messenger, Instagram and Slack land in the same list, with the oldest unanswered first when you want it.

  • Assigned to me, unassigned, waiting: filters kept in the link, so a view can be shared.
  • Search every message, not just names.
  • Canned replies and tags your whole team shares.
  • USSD sessions appear alongside, so a follow-up can go out by SMS.
  • On a phone too: the whole Inbox works at phone width.

Running the team

Built for the people who answer.

The right person, automatically

Round robin or least busy, agent groups and a chat limit per agent. A declined chat passes to the next free agent.

A queue, not silence

When everyone is busy, the customer is told they are waiting, and the chat goes to the next agent who frees up.

Bot and person, one conversation

Take over from the bot, or pass the chat back to it. The bot stays quiet while a person has the chat, and the customer is told who joined.

WhatsApp’s 24-hour window, handled

Outside the window an approved template goes out instead, and your reply is delivered as soon as the customer writes back.

Alerts that find your team

An email when an assignment is waiting, optional WhatsApp alerts, notifications in the browser, and quiet hours.

Close the loop

A satisfaction survey when a chat is resolved, and conversation analytics: volumes, response and wrap-up times, satisfaction.

How it works

From the flow to the right person.

  1. Step 1

    Hand off

    Add a Live Chat step anywhere in a flow, or take over a conversation the bot is handling.

  2. Step 2

    Route

    It goes to the next agent by round robin or availability, within a group, or waits in the queue.

  3. Step 3

    Answer

    The agent sees the history, the customer’s records and a reply drafted from your documents.

  4. Step 4

    Resolve

    Close with a satisfaction survey, pass to a colleague, or hand the customer back to the bot.

Channels

Where the conversations come from.

Chat channels
WhatsApp, web chat, Telegram, Messenger, Instagram and Slack, two-way, in one inbox.
USSD
Sessions appear in the queue for an agent to claim and record an outcome. The reply goes by SMS.
Voice
Coming with Voice and IVR: calls handed to a person ring your team’s browser softphone.
SMS
Outbound only: confirmations, and replies to USSD sessions.

Tell us how your team works.

We’ll show you how Buni.ai would fit it, and what it would take.