← Back to library

Handle inquiries from Telegram while you still approve the reply

Toma describes an n8n support desk that gathers form and email inquiries into ClickUp, notifies him on Telegram, and uses AI to finish the busywork (outbound mail or ticket state) only after he answers in Telegram. The point is help with context, not full autopilot.

  • OutcomeCapture each inquiry with an ID, open a task, ping you on Telegram, and let AI draft the next email or status change only after you reply in chat.
  • SetupAn afternoon+
  • LevelIntermediate
  • CreditToma (@Toma_web3)
View original
Works with
n8nChatGPTClaude

Published Oct 23, 2026Updated Oct 23, 2026AI dry run on sample input: passed Oct 23, 2026Source verified Oct 23, 2026

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?

  1. Capture onceCreate a task and ID when an inquiry arrives; store the full thread.
  2. Ping a personSend yourself a Telegram note with the ID and a short gist.
  3. Reply in chatAnswer in Telegram in plain language.
  4. 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)

Prompt4 variables
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?
The poster wanted to answer from one place on the phone. Any chat you actually check works if the workflow waits on your message.
Can it close tickets on its own?
Only when your reply clearly says to close. Ambiguous replies should ask you again.

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.

More in Customer Success

All 8 Customer Success playbooks →

Related playbooks

Customer SuccessRoutine✓ AI dry run

Send travelers a pre-trip checklist and a welcome-back note

Have a daily workflow read the trip sheet and message each traveler a checklist before departure and a check-in after they return, without duplicates.

Credit: Tongshishak Gojim | AI Automation Specialist (@tongshishak)

Under an hour
Customer SuccessPrompt✓ AI dry run

Send your buyer a phone-friendly list of contract deadlines

Give the ratification date and the key contingency deadlines and get a friendly email to your buyer that lists each date in order, says what happens at each step and what the buyer must do.

Credit: David Stroh

Minutes
Customer SuccessWorkflow✓ AI dry run

Analyze a few thousand app reviews inside a code editor

Turn a CSV of customer reviews into ranked complaints, a severity-by-frequency grid and a journey map in about half a minute, then prototype from the findings.

Credit: Eric Xiao

Minutes