Solution · bots and AI features

A Telegram inbox for your team: messages and voice notes become Notion tasks at once

The team lives in Telegram while tasks live in Notion, and everything agreed in chat has to be copied by hand. The solution is a bot you write or speak to: AI works out whether it is a task, a note or a meeting, parses the date and puts the record in the right Notion database or calendar.

What it looks like

Mock-ups with demo data, no real data

Team inbox

🎙 Voice 0:14 → transcribed

✅ Task: prepare the contract

👤 Maria · 📅 Thursday, 12:00

· Database: Team tasks

11:20

Voice note → task

At a glance

In production: As my own product — the hey, nooka bot
Fits: Teams that talk in Telegram but work in Notion
Needs: Notion and Telegram; Google Calendar for events

Estimate for your company

≈7 h
a month of copying from Telegram to Notion by hand
Assumption: 10 people × 2 records a day × ~1 minute of manual entry × 22 working days (our assumption)

This is a calculation, not a promise: plug in your own volumes and the figure changes.

Before

Something is agreed in chat, and the task exists only in the conversation.
Moving a thought into Notion from a phone takes too long, so nobody does it.
Nobody transcribes voice notes, yet half of the assignments are in them.

What was built

Knows what it is

AI sorts each message into one of six types: task, note, event, schedule question, “done”, reschedule.

Voice to text

Voice messages are transcribed and handled exactly like text.

Exact dates

Dates are parsed by a rules-based service, not the language model, so “Friday at 3” is always Friday.

Covered separately →

Into the right database

Each member connects Notion via OAuth and picks the database with buttons; events go to Google Calendar.

Stack

n8n
Telegram Bot API
Notion API + OAuth
Google Calendar
Supabase
AI models via API
Speech recognition

Where it runs

hey, nooka: a Telegram bot that files your messages and voice notes into Notion

Read the case →

Questions

Is this an off-the-shelf bot or built for us?

The architecture is proven in my own product, the hey, nooka bot. For a team it is set up around your databases, task types and rules.

What if the AI gets the type wrong?

The record can be fixed right in Notion, and frequent mistakes become classifier rules.