Generate service quotes with GPT-4.1-mini, Telegram approval and Gmail

Who's it for

Service businesses such as trade contractors (roofers, plumbers, electricians, cleaners, facility management) who want to send out customer quotes automatically as PDFs — either triggered by a Tally form embedded on their website or by a Voice AI agent webhook request. The business owner wants to delegate the operational work without giving up control, so every quote passes through an approval instance via Telegram before reaching the customer.

How it works

1. Intake: A Tally form submission triggers the workflow. A parallel Webhook path catches direct API submissions (e.g., from a Voice AI agent) and routes them through a validity check.
2. Normalization: An Edit Fields node maps the raw payload into a clean, consistent schema.
3. AI classification: GPT analyzes the request and assigns one of six categories (single-family quote, multi-family quote, management matter, emergency, on-site appointment, special request).
4. Routing: A Switch node dispatches the request to the matching branch.
5. Approval-gated quoting (for quote-requiring requests): PDFMonkey generates a draft PDF, a Telegram sendAndWait approval request is sent to the business owner, and depending on the response the document is either emailed to the customer via Gmail or discarded.
6. Direct notification paths (management, emergency, on-site, special requests): Each fires a Telegram alert plus a structured Gmail message to the relevant inbox.

How to set up

Estimated time: 30–45 minutes

1. Credentials — connect: Tally (webhook), OpenAI, Telegram Bot, Gmail (OAuth2), PDFMonkey API.
2. Tally form — duplicate your form, copy its ID into the Tally Trigger node.
3. Telegram — create a bot via @BotFather, paste the token, and set the target chat IDs in each Telegram node (one per branch).
4. Gmail recipients — open each Gmail node and set the correct destination address per category (management, emergency, on-site, special).
5. PDFMonkey — create your document template in PDFMonkey, paste the template_id into both PDFMonkey nodes (one for single-family, one for multi-family).
6. AI prompt — open the "Classify Request" node and adapt the classification categories to your business vocabulary if needed.
7. Test — submit a test entry per category through the Tally form and verify each branch fires correctly.

Requirements

• n8n (self-hosted or cloud)
• OpenAI API key (GPT-4o or comparable model)
• Telegram Bot token
• Gmail OAuth2 credentials
• PDFMonkey account with at least one document template
• Tally account with a form (optional: Voice AI agent for the webhook path)

How to customize

• Add more categories: extend the AI prompt and add new Switch routes.
• Swap Telegram for Slack/Discord: replace the Telegram nodes with your channel of choice — the approval flow works identically with Slack's "Send and Wait".
• Change the LLM: replace the OpenAI node with Anthropic, Mistral, or any other supported model node.
• Skip the approval step: bypass PDFMonkey + sendAndWait for low-risk document categories by adding a direct route.
• Trigger via Voice AI agent: point your Voice AI's tool/webhook output at the Webhook node instead of using the Tally form.