Skip to content

Train your AI assistant in 30 minutes

The AI assistant is the persona that replies inside your Telegram workflow. It is configured during onboarding and can be refined later from Settings -> AI Chatter.

This guide covers the creator-facing setup decisions. It does not document internal prompt logic, safety rules, or automation thresholds. The goal is simple: give the assistant enough public brand context to sound consistent on day one.

Set the public identity fans should recognize:

  • Display name.
  • Basic persona profile.
  • Brand tone.
  • Details that are safe for the assistant to reference.
  • Details that should stay off-limits.

Use the identity your audience already knows. Do not invent a new character unless you are intentionally launching a new persona.

Tune how messages feel:

  • Short, medium, or more adaptive replies.
  • Softer or more direct tone.
  • Emoji level.
  • Sales pace.
  • Manual-review preference for delicate moments.

Start close to how you already reply. If you are unsure, choose the middle option and adjust after reading real conversations in Inbox.

Select the languages the assistant can use confidently. tease.bot supports fan chat across the product’s supported language set, and the assistant remembers each fan’s language on their fan record.

If you only want English replies, keep the setup narrow. If your audience is multilingual, add the languages you genuinely want the assistant to handle.

Boundaries are the things the assistant should not reveal, promise, or discuss.

Good boundaries are concrete:

  • Personal details that should stay private.
  • Topics that should escalate to a human.
  • Requests you never want accepted.
  • Tone limits for sensitive situations.

The assistant can handle a lot of repetition, but it should not be the final decision-maker for situations that need human judgement.

Set pricing limits for paid media and tips. These are guardrails for the assistant and the funnel builder.

You can set broad minimums and, where needed, stricter floors for specific content categories. Keep the first version conservative enough that you would be comfortable honoring it in a live conversation.

If voice is enabled on your plan, configure it before fans hear it.

Use the voice setup to choose the voice style, then test a short message. If it does not fit the persona, keep voice off until it does. A silent assistant is better than a voice that breaks the brand.

Choose how fans reach the assistant:

  • Channel connection through @myteasebot.
  • Telegram Business Mode for private DMs.
  • Both, if your audience uses both surfaces.

For the step-by-step, see Connect Telegram to @myteasebot and Enable Telegram Business Mode.

Send messages from a test Telegram account and read them in Inbox.

Look for:

  • Does the first reply feel on-brand?
  • Does the assistant keep the right length?
  • Does it avoid topics you marked as off-limits?
  • Does manual takeover work cleanly?
  • Does the funnel offer feel natural when it appears?

Change one setting at a time in Settings -> AI Chatter. Then test again. Small changes give you clearer signal than rewriting the whole persona every time.

After launch, use real conversations to refine the setup:

  • Add missing boundaries when fans find edge cases.
  • Nudge tone if replies feel too formal or too intense.
  • Add languages only when you are ready to support those fans.
  • Review voice again before using it at scale.

The assistant picks up saved changes on future replies.