Creator messaging and a reply assistant
A first message tailored to each creator goes out through your agency's Backstage and replies are collected on one screen; the AI assistant replies and sends the invite link, and you can take over any conversation at any moment.
Finding a suitable creator is half the job. The other half is reaching out with the first message and getting back to replies on time. As the team grows, this becomes both the most time consuming work and the easiest to get wrong. In TT Panel each creator gets a first message tailored to their profile, prepared with the agency's template and rules, and it goes out through your agency's own Backstage. When a creator replies, the conversation lands in the panel; if the agency wants, the AI assistant answers with the agency's own information and, at the right moment, sends the application and invite link, and you decide which conversations to take over and when.
After finding creators, the most time consuming work is running first contact and getting back to replies on time. Each creator gets a first message tailored to their profile, and it goes out through your agency's own Backstage; the AI assistant handles incoming replies and, at the right moment, sends the application and invite link; every conversation stays in front of your team.
- The first message is prepared for each creator from their profile, with the agency's template and rules, and goes out through your agency's Backstage. Right before sending, eligibility is checked from your Backstage and seven checks run; if one fails, the message does not go. Each sending account sends a creator one first message; if the creator does not reply, no follow up goes out; if a creator accepts the message request but does not write, the agency can turn on a single reminder if it wants.
- If the agency switches the reply assistant on, the AI assistant answers creators' questions: what the agency does, earnings, gifts, requirements, stream hours, applying. At the right moment it sends the application and invite link with your agency's card. Replies are based on your agency's approved knowledge base and pass fixed rules that no setting can switch off.
- Every conversation has a mode: assistant, human or muted. The moment an operator writes in a conversation, it moves to a human and any pending assistant reply is cancelled. A message the system is unsure about is marked "needs a human".
- Messages with harassment, insults, spam or scam attempts get no reply, and the conversation is muted. A creator who asks you not to write again goes on your block list and gets no further messages.
From first message to reply
- Step 1:
Seven checks
Right before the first message, eligibility is checked from your Backstage and seven checks run. If even one fails, the message does not go.
- Step 2:
First message
The message is prepared for the creator from their profile and goes out through your agency's Backstage. Each sending account sends a creator one first message.
- Step 3:
The reply lands in the panel
When the creator replies, the conversation is matched to that creator's record. You see the earlier messages on the same screen.
- Step 4:
The assistant or you
If the agency turned it on, the AI assistant replies based on the agency's knowledge base and sends the invite link at the right moment; no reply is used before it passes the rules. The moment you write from the panel, you take over.
Read the steps in detail
How the first message works
The first message is prepared for each creator from their profile; your agency's template sets the frame and the rules of the text. The message goes out through your agency's own Backstage; if you want, it is also sent from the 3 extra TikTok accounts you connect.
Each sending account sends a creator one first message. If the creator does not reply to the first message, no follow up goes out. If a creator accepts the message request but does not write, the agency can turn on a single reminder if it wants. After your Backstage has written, one of your extra accounts can send the same creator its own first message; the panel never writes again on its own to a creator who does not answer.
Right before the first message, the creator's eligibility is checked through your Backstage and seven checks run: fresh eligibility evidence, identity, the block list, no earlier conversation, the daily quota, the emergency stop and working hours. If even one fails, the message does not go; working hours are skipped only for a first message an employee sends by hand from the panel, one at a time. The result of the check is written to the evidence record at that moment, so you can later show which check came before the message.
Three conversation modes: assistant, human, muted
Assistant
A reply that passed the rules is used
Every conversation has a mode, and the mode decides who answers it. In assistant mode, replies that passed the rules are used. Model replies also depend on a switch the agency controls; while it is off, no model reply is sent, but messages your team writes in the panel still go out.
Human
Your team writes the answer
In human mode the assistant stays out of that conversation and your team writes. The moment an operator writes in a conversation from the panel, it switches to human mode and any assistant reply waiting in the queue is cancelled, so the team and the model never talk over each other. The system also hands a conversation to the team on its own and marks it "needs a human" in the list: when the message is unclear, the topic is outside the knowledge base, or the reply fails the rules.
Muted
No reply is prepared
In muted mode no reply is prepared. Messages with harassment, insults, spam, sexual content or scam attempts get the conversation muted, and it shows as muted in the list. A creator who asks you not to write again goes on your block list and gets no further messages.
What you get
Find the right conversation
Clear who wrote what
The details
Why messaging falls apart
Writing to creators looks simple: a few lines, then waiting for an answer. But when several people on the team write, two of them can message the same creator, and a creator who never answered can get a second message. Replies do not follow office hours either: creators tend to answer after streaming, late at night or at the weekend. If nobody is watching, the reply waits until morning and the creator cools off.
Read more
Then there is consistency. Two people give two different answers to the same question; one copies the application address from an old note, another explains what the agency offers in their own words. With messages scattered across screens, working out later who promised what gets hard.
The answer is not speed without rules. A tool that messages creators without checking eligibility, writes to the same person again and again, keeps no records and ignores the platform's daily limits does not speed anything up; it puts your account at risk. TT Panel keeps speed inside the rules.
When the creator replies
A conversation starts when a creator writes back. The panel matches the message to that creator's record, so the earlier messages and who the creator is sit on the same screen. Under each message you see whether the creator, your team or the assistant wrote it.
Read more
The reply assistant is something the agency switches on or off. When it is on, each incoming message is classified first: is the creator asking something, chatting, saying they are not interested, or is the meaning unclear? At this step the model cannot write a sentence; it only picks one label from a fixed, closed list.
A label does not trigger anything on its own; which label leads to which outcome is written in a fixed table in the system. A question or small talk gets a reply prepared. For some kinds of question no new text is generated, and an approved answer from the knowledge base is used word for word. A message that is unclear or suspicious is left to the team.
When a reply is written, the model relies on the agency's knowledge base: what the agency does, how earnings work, gifts, requirements, stream hours, how to apply. If an application or invite link is needed, the model does not type the address, it only signals that a link is needed; the system takes the agency's card and link from the agency's record and adds them.
No assistant reply is used without passing the rules
Text prepared by the model is never used as it is. It first passes through a fixed rule layer. There is no AI inside that layer: its rules reach the same verdict on the same text every time. A message cannot talk it round, and no setting switches it off.
Read more
The rules look at content and form. The text must not contain a link or an address, drift into another script, or carry invisible characters or long dashes. No fragment of the system's own instructions may leak into it. Overlong replies, repeated sentences and template fields left unfilled are turned back too. A reply that fails is not trimmed or patched; it is not sent.
When a reply is rejected, an approved answer for that situation is used if one exists; otherwise the conversation stays with your team. Approved answers added to the knowledge base pass the same rules when they are saved.
The assistant has no authority
The model only proposes text. It cannot change records, switch settings, run tools, send a message itself or make up a link. The only things in front of it are the conversation and the agency's knowledge base.
Read more
Which message gets a reply, whether a text goes out and who a conversation is handed to are decided by the system's rules. The text of a queued reply is fixed at that moment and is not rewritten at send time.

The Messages screen
A conversation in the panel: the first message, the creator's reply and the assistant's answer. Taken from the panel as it is.
Who it helps
Agencies that deal with many new creators.
Planning first contact can leave the team's hands, and the team spends its time on creators who replied.
Agencies that do not want night and weekend messages left waiting.
With the reply assistant on, common questions do not wait for morning, and your team looks at the conversations that need a decision.
And agencies that would rather not hand conversations to an assistant.
You can leave the layer off or move single conversations to human mode; the panel then works as a shared screen where replies collect and your team answers. The agency owner sees who was written to and which conversation is waiting for the team on one screen.
Questions about this feature
Which views does the Messages screen have?
All, Replied, Wrote to us first, Silent signal, Unread, Needs a human and Muted, plus a search box for creators. Switching view does not delete any conversation; it only filters the list.
How do I see which creators replied?
The Replied view collects the conversations where the other side wrote at least one message after our first message. The message can be text, an image, a sticker or a voice message.
How do I avoid missing a new reply?
The Unread view shows the conversations whose last message is from the other side and that you have not opened since.
Can I see the earlier conversation?
Yes. When you open a conversation, the earlier messages and the creator's record are on the same screen. Under each message it says whether the creator, the team or the assistant prepared it.
Can I write to a creator from the Messages screen?
Yes, into an existing conversation; your messages go out in order. This screen does not start a new conversation. The moment you write from the panel, the conversation switches to human mode.
Plan and price
A first message tailored to each creator and the AI reply assistant are included in the plan and run on the server reserved for your agency. TT Panel Core is 299 USD a month and includes every feature: first messages go out through your agency's Backstage and, if you want, from 3 extra TikTok accounts, and up to 4 people use the panel.
Prices are in USD, and tax is added at the payment step.
Related features
- Panel and teamTikTok LIVE agency team panel: personal sign in for every employee, creator files and conversations in one place, every first message logged by account.How the team panel works
- Evidence recordThe eligibility check you run before writing to a creator is recorded the moment it happens and cannot be edited later. In a dispute, you export that record.How the evidence record works
- Candidate discoveryFind TikTok LIVE creators in one candidate list, ranked by score and league history. No duplicates, and eligibility is checked before anyone writes.How candidate discovery works
Guides that go deeper
- Reaching creators by TikTok DMCovers the platform's limits and the rules to follow when you message a creator.
- Who messaged whom in the teamDescribes the routine that stops the same creator getting two messages when several people write.
- The risk of messaging a tied creatorExplains why the check has to come before the first message.
- How agencies find creatorsWalks through the process from discovery to the eligibility check and the first message.
Contact us with your questions or for detailed information about the system
Write to us on WhatsApp or by email; we answer your questions and explain the system in your own language.