When should you use it?
- Inquiries arrive from forms and email and context gets lost between tools
- You want faster replies without letting a bot talk to customers alone
- Your team already lives in Telegram or a similar chat
Not for: Fully automatic customer replies with no human check; Regulated advice without a qualified person reading every send
What do you need, and what do you get?
You need
- Inquiry sources (form webhook, Gmail)
- Telegram bot for you
- Task tool (ClickUp or similar)
- Tone and escalation rules
You get
- One task per inquiry with a unique ID
- Telegram alert with context
- Draft or send actions only after your Telegram reply
How does it work, step by step?
- Capture onceCreate a task and ID when an inquiry arrives; store the full thread.
- Ping a personSend yourself a Telegram note with the ID and a short gist.
- Reply in chatAnswer in Telegram in plain language.
- Let AI finish the busyworkHave it finish outbound mail or ticket state only from your reply, and log what it did.
The prompt
Our own prompt, inspired by Toma (@Toma_web3) and written to work in n8n, ChatGPT and Claude. Values in lime are placeholders.
Fill in the blanks (optional; the copied prompt uses your values)
You are the busywork layer for inquiry {enquiry_id}. Ticket context: {ticket} Human Telegram reply: {human_reply} Rules: {rules} Decide one action: outbound_mail | change_ticket_state | wrap_up | ask_human. If outbound_mail, draft the message in our tone and wait for confirmation unless rules say the human reply already approved sending. Never invent facts, prices or promises not in the ticket or human reply.
Inspired by Toma (@Toma_web3): AI customer support from Telegram with n8n. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- If the bot can send without your Telegram reply, you no longer have a human-in-the-loop desk.
- Keep a kill switch that pauses outbound mail.
Our take
The useful idea here is not that AI runs support, it is that AI keeps context while a person stays in control. Capturing the inquiry, notifying on Telegram, and acting only after a human reply is the right default for small teams. We keep ClickUp and Gmail as examples and stress the wait-for-reply rule. In our dry run with 5 synthetic inquiries, each got an ID and Telegram ping; after a human reply of send the pricing PDF and mark waiting, the agent drafted the email, updated the task, and did not close the thread. Put the kill switch where someone on-call can find it at night. Log every outbound action with the human Telegram message that authorized it. If the model is unsure whether a reply approved sending, it should ask again instead of guessing. Keep regulated topics out of this desk entirely. Train the team that Telegram is for decisions, not for pasting secrets.
Where the time goes back: You stop jumping between inbox, CRM and chat; context stays on the task and you only type the decision once.
FAQ
Why Telegram instead of the ticket UI?
Can it close tickets on its own?
Source and credit
Inspired by Toma (@Toma_web3). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- AI customer support from Telegram with n8n
- Creator
- Toma (@Toma_web3)
- Published on
- X
- Original date
- Jun 30, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- n8n
- AI dry run
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Waited on human reply; no unsolicited close.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.