Chatbot conversation history: how to review chats and take over

If you don't review your chatbot conversation history, it's easy to miss a loss: the bot accepted a request, the client is waiting for a call, but there's no lead in the CRM — for example, because the balance ran out. The client thinks they're booked, and the manager doesn't even know they exist. Or the bot repeats the same mistake for months and quietly loses clients. Without reviewing the conversations, you'd never find out.

“Dialogs” is a log of every conversation the bot has with real clients across all channels. Here you see what the bot answered, whether a lead was created and, if not, why. A bad reply is fixed right here, and the next client gets the right one. A manager can stop the bot for one client and reply personally. Viewing, filters, manual replies and export are not charged separately.

Watch the two-minute video, and below you will find step-by-step setup.

Why businesses need it

• Online school owner: every morning, they set the “With number” filter for yesterday. The “Shown N of M” counter instantly gives the share of conversations where the bot got a phone number — that's the bot's conversion rate.
• Clinic, salon, car service: a photo from the client appears in the feed together with what the bot recognized in it. The administrator takes a tricky question over, and the client doesn't get two different answers.
• Online store: phone search finds the conversation whatever the number format — with or without the country code, with brackets — even if the number came up in the middle of the dialog.
• Marketer: the “Source” filter shows which link to the bot brought conversations and leads.
• Spam protection: the login of an annoying chatter is copied into the bot's “Login blacklist”. The spammer no longer drains your balance or creates leads.

How to view chatbot conversations

Open “AI chat bot” → the bot's gear → “Dialogs”. The newest conversations are at the top. Each row shows the client, the last message, the number of messages and the activity time. Row buttons: “Open lead #… in CRM”, “Copy login” and “Delete dialog”.

The list doesn't update itself: to see new messages, click “Refresh”. Test conversations from “Training AI” don't appear here.

The “Dialogs” section matters most in the first days after launch — it's the last step of how to train an AI chatbot. Here you see where the bot still makes mistakes and where it already copes on its own.

How to find a conversation with a client

1. 1
   Search

   Enter a name, @login or phone number. For a phone, three digits are enough and the format doesn't matter: the number is searched across the whole conversation.
2. 2
   Leads with a phone

   Click “Target action”, tick “With number” and click “Apply”. Add the “Activity date” to choose a period.
3. 3
   Channel and ads

   “Platform” selects Telegram, MAX, “Chat on site” or MAX direct. The “Source” list in the target action window selects the tag from the link to the bot.
4. 4
   Long dialogs

   In the “Messages” field, set “from” to, say, 8. Rows without a lead button show where the bot failed to close the client.
5. 5
   Cold base

   The “Write-first base” filter shows who replied to the first-message campaign — a MAX messenger broadcast to your own contact base. It appears once at least one base is uploaded.

The “Shown N of M” counter shows how many dialogs passed the filters. The “Reset filters” button clears everything at once.

Why the chatbot didn't create a lead: the dialog card

Clicking a row opens the message feed. Under the client's message you see “🎯” badges — the target actions that turn the bot into a chatbot with CRM integration. A green badge means the condition was met, but not yet that the lead was created. A grey “did not fire” badge means the AI decided the client hasn't given the required data yet. If a lead really was created, there's a “Lead #N” link next to the badge.

A green badge without a link — open it. The window shows the reason, for example “Action NOT executed · No account balance” or “Skipped by frequency”. For AI actions, the same window has a “Training by launch” tab: explain when the action should fire.

Under the bot's reply there's “Fix reply”. The rule is written into the instruction and saved right away — the next client already gets the correct reply.

How to reply to a client instead of the bot

1. 1
   “Stop bot”

   Click the button in the card header. The bot stops replying to this client and won't send anything on its own. It handles all the other dialogs as usual.
2. 2
   Reply

   Type in the “Write a message…” field and click “Send”. The message goes out on behalf of the bot in the same channel. You can attach files up to 50 MB: a price list, a contract, photos.
3. 3
   “Start bot”

   Click it when you're done, and the bot will reply to this client again.

How to export chatbot conversations and share a dialog

“Download all” packs all dialogs of all the account's bots into one ZIP: a folder per bot, a file per dialog and a summary table. You can copy the address of an open card and send it to a colleague — the link opens that dialog right away.

Read also: How to train an AI chatbot without prompts when it answers wrong, Chatbot CRM integration: creates leads, sends files and follows up, Customer profile in CRM: full history, client replies and tasks