What it is
An automation builder: it connects services to each other (CRM, spreadsheets, messengers, warehouse) without writing code from scratch for every link.
How we use it
We deploy it self-hosted — on your server, not in someone else's cloud — and build chains like "website request → Telegram notification → spreadsheet row" on it.
Where n8n helps a business
- Leads arrive in a Telegram bot and are then copied into the CRM and a spreadsheet by hand.
- A manager learns about a deal status change only when they open the CRM themselves.
- Payments land in one system while the books are kept in another, and reconciliation eats time.
- The number of systems grows, and connecting each pair with a separate program gets expensive.
Options
- n8n on your own server. No limit on executions, data does not leave for someone else's cloud, and the service will not be cut off by regional restrictions. This is our default.
- Cloud builders (Zapier, Make, cloud n8n). Quick to start, but the price grows with the number of steps and access depends on the vendor.
- A dedicated Python service. When there is no ready-made node or you need custom logic and frequent polling. That is how Wildberries slot monitoring is built: no n8n there, just our own bot.
How we use it
The lead automation for a small business has four groups of workflows:
- Lead capture. The first node is subscribed to the Telegram bot's messages. The message is split into fields and created as a record in Bitrix24.
- Sync. The same lead and its status are appended as a row to Google Sheets, which stays as a fallback record if the CRM is unavailable.
- Notifications. When a deal status changes, Bitrix24 calls n8n itself, and the workflow finds the responsible manager and messages them in Telegram.
- Payment webhooks. Payment events arrive at the same hub and are recorded without manual reconciliation.
The systems are not connected to each other directly; each talks only to n8n, so replacing one does not break the others. Workflows and prices for such connections are under Automation.
Common problems
- Keys do not travel with the workflow. A workflow exports to JSON without tokens: on a new server it is imported, the Telegram and Google credentials are assigned again, and it is activated. Good for security, but the step is easy to forget.
- "Did the lead arrive?" The answer is in the execution log, where every run is visible without reading code. If nobody looks at the log, a failure is noticed from a customer complaint.
- No ready-made node. Then we write a node or a separate Python service and plug it into the chain.
- The server is the only ceiling. There are no plan limits, but capacity is limited by your machine, and it needs maintenance.
When you do not need it
If there is one simple connection, such as a new-lead alert in Telegram, it is cheaper to build it into the bot. And when the job is frequent polling with filters, like delivery slots, a small dedicated service is more reliable than a workflow.