Build an AI Vendor Follow-Up Agent in n8n

Chasing late purchase orders is a daily job in operations, and it usually lives in someone's head and inbox. In this free lesson you build an AI vendor follow-up agent in n8n for Ridgeline Supply Co. You give it a list of late or pending orders, it writes a polite, firm message for each one, and it logs every follow-up with a due date in Google Sheets.

What you will build: an AI agent that chases late orders

You build a small n8n workflow with four parts: a chat trigger so you can talk to the agent, an AI Agent with a system prompt that knows Ridgeline Supply Co., an OpenAI chat model with simple memory, and a Google Sheets tool so it can write rows into a spreadsheet.

Then you test it. You paste three open orders: Tallpine Pallet Co. on PO 4471 is five days late, Corrugo Box Supply on PO 4480 has no ship date, and Northwind Wrap shipped PO 4492 twelve rolls short. The agent writes a short message for each. Every message names the PO, states the problem in one sentence, asks one clear question, and gives a reply-by date. When you say "save those follow-ups to the sheet", it writes each one as a row in a spreadsheet called Vendor Follow-Ups.

Who this is for

This lesson is for operations managers, purchasing coordinators and warehouse leads who spend part of every week asking vendors "where is my order?". You know the pain: a short shipment nobody logged, a ship date you were promised but never wrote down, a follow-up that went out late because the person who sends them was on the dock. No coding and no AI background are needed.

Step by step: what you do in the lesson

• Get n8n: start at n8n.io and click Get started free.
• API key: add a $10 credit balance in the OpenAI dashboard, create a secret key named "n8n agent", and copy it.
• New workflow: create a workflow in n8n.
• Chat trigger: add "On chat message" as the first step.
• AI Agent: add the AI Agent node, then add a System Message and paste the Ridgeline Supply Co. system prompt.
• Chat model: add the OpenAI Chat Model, create a credential with your API key, and pick gpt-5-mini.
• Memory: add Simple Memory so the agent remembers the conversation.
• Sheets tool: add the Google Sheets Tool.
• Create the spreadsheet: name it Vendor Follow-Ups and add five headers: Vendor, Order, Issue, Follow-up Message, Due Date.
• Configure Sheets: sign in with Google, choose Append Row, pick the Vendor Follow-Ups document and Sheet1, and let the model fill each of the five columns.
• Test the agent: open the chat, paste your open orders, then ask it to save the follow-ups to the sheet.

What the system prompt tells the agent

The system prompt is the standing instruction the agent reads before every request. It says who Ridgeline Supply Co. is: a small warehouse and distribution business that buys pallets, boxes, stretch wrap and janitorial supplies. It says who the messages are for: vendor sales reps and customer service teams.

It sets the tone as polite, plain, firm and brief, like an operations manager who values the relationship but needs an answer. It lists what each message must include, and it sets the reply-by rule: one business day if the order is late, two if it is pending. It also sets two limits. The agent may not threaten to cancel an order or invent penalties, and each message stays under 80 words. The five return fields match the five sheet columns exactly, which is why the Sheets tool knows where everything goes.

What you need

For the lesson itself, nothing but a browser. To build the agent for real you need an n8n account, an OpenAI API account with a small credit balance, and a Google account for Sheets. Keep the API key private: the lesson reminds you never to paste it into a chat, a shared file or a screen recording. Have a list of two or three real open orders ready so your first test is useful.

Why it works

A chatbot only talks. An agent can also act, and the Google Sheets tool is what lets it act. Memory lets you say "save those" and have it know which messages you meant. Because the model fills each column from your headers, you do not map fields by hand. The result is a running log of every vendor follow-up, with the issue and the date you expect an answer, in a sheet your whole team can see.

How to use this in your operation

Change the system prompt to describe your business, your usual vendors and your own reply-by rules. Each morning, paste the open POs from your purchasing report and let the agent draft the follow-ups. Review the messages, send them from your normal email, and use the Due Date column to see who owes you an answer today. Sort by vendor before a quarterly review and you have a clear record of who ships late and how often. The agent writes and logs; a person still decides what gets sent.

Frequently asked questions

Does the agent email vendors by itself?

Not in this lesson. It writes the follow-up messages and logs them in Google Sheets. You review and send them. The next lesson in this section shows how to add a Gmail tool so an agent can send email.

Is this lesson free?

Yes, Stepthrough is completely free. To build the agent for real you need your own n8n account and an OpenAI API account. OpenAI charges per request; the lesson adds a $10 credit balance, and short messages like these cost a fraction of a cent each. n8n has its own plans, so check its current pricing.

Do I need to know how to code?

No. Every node is added by clicking, and the system prompt is pasted in for you. The lesson is 69 short steps and about 12 minutes.

Can I change the spreadsheet columns?

Yes. Keep the column headers and the fields listed in the system prompt the same, and the agent will fill whatever columns you set up.

How does the interactive lesson work?

Each step is shown to you first on the left (Watch). Then you do it yourself in a practice copy on the right (Your turn), and every click is checked. If you get stuck, press Do it for me. It is a practice copy, so no real accounts are touched.

More in AI Automation for Operations