1 · Chat Trigger
The workflow starts with n8n's native chat interface. A user types something like 'Office supplies $50 today' and the message kicks off the flow.
Case Study · Automation with Explanation
AI Bank Expense Tracker
A chat-driven n8n automation that turns plain-English messages into clean, categorized rows in a Google Sheet — no forms, no manual entry.
The Workflow
Built in n8n: a chat trigger feeds an AI Agent (OpenAI) wired to a memory store and a Google Sheets tool. The agent decides when to log a row.
2 · AI Agent (OpenAI)
An AI Agent node powered by an OpenAI chat model parses the message: it extracts the expense name, amount, category, and date — and asks follow-up questions when fields are missing.
3 · Simple Memory
A memory node keeps short-term context across turns so the agent can hold a multi-message conversation (e.g., 'add it as Food/Supplies instead') without losing state.
4 · Google Sheets Tool
Once the agent has a clean record, it calls the 'Add-expense' tool which appends a new row to the AI Bank Expense Tracker sheet — Expense, Amount, Category, Date, Status, and payment type.
Connecting the Tools
Two credentials power the automation: a Google Sheets OAuth2 connection so n8n can append rows on the user's behalf, and an OpenAI API key for the agent's reasoning model.
The Output
Every chat turn that resolves into a transaction lands as a row in the tracker — ready for pivoting, charting, or month-end review.
The Stack
- n8n (workflow orchestration)
- OpenAI chat model (extraction + reasoning)
- Simple Memory (multi-turn context)
- Google Sheets OAuth2 (append rows)
- Chat trigger (no UI to build)
The Takeaway
A 4-node automation replaces an entire mini-app: no front-end, no database, no custom backend. The AI Agent handles the messy human input, memory keeps the conversation coherent, and Google Sheets becomes the system of record. This pattern — chat in, structured data out — scales to invoices, CRM updates, support triage, and beyond.