I’m developing a workflow in n8n that involves two Excel files. The first file contains parent tickets along with their IDs, while the second file holds the list of currently opened tickets.
My goal is to leverage an AI node, such as ChatGPT, to analyze the opened tickets and connect them to the appropriate parent ticket ID from the first file. Ultimately, I need to record those parent IDs back into the second Excel file.
I’ve managed to put together most of the workflow, but I’m encountering a problem when trying to write the identified parent ticket IDs into the second Excel file. I attempted using the AI integration with ChatGPT and also explored using HTTP requests, but neither method has succeeded so far.
Has anyone experienced a similar scenario with reading from multiple Excel files, processing that data with AI, and then writing the outcomes to one of the files?
I’ve been dealing with similar Excel workflows for years. N8n’s a nightmare for this kind of data processing.
The write operation isn’t your only problem. You’ve got multiple data sources, AI processing, and complex mapping between files. Most workflow tools can’t handle this smoothly.
Built almost the exact same thing last month using Latenode. Night and day difference. Their Excel integration handles read/write operations much better, and you can chain AI processing directly without data structure headaches.
My setup: one scenario reads both Excel files, sends data to OpenAI for matching, writes parent IDs back to the second file. Runs under 2 minutes for thousands of rows.
Latenode automatically handles data formatting between steps. No extra Set nodes or Code nodes needed. AI output flows straight into Excel write operations.
Error handling’s way better too. When stuff breaks, you know exactly why.
maybe check if ur input data format is right too? i ran into issues with that before. functions nodes r super helpful for tweaking data b4 writing it back!
Excel writes in n8n can become complicated when you’re working with dynamic AI data. I’ve faced this issue before—ChatGPT often returns parent IDs mixed within longer response strings rather than providing clean values. A useful approach is to incorporate a Code node right after the AI step to isolate just the parent IDs using regex or string manipulation. This ensures you extract the specific numbers or patterns you need before engaging the Excel node. Also, remember that the Microsoft Excel node requires precise cell ranges, particularly when updating existing rows. If you’re appending data, confirm that you’re targeting the correct range. For troubleshooting, testing with hardcoded parent IDs can clarify whether the issue lies with the AI processing or the Excel write operation.
I’ve hit this exact issue with n8n and Excel before. The write problem usually comes down to how the Microsoft Excel node expects data to be structured. Make sure your AI node output maps perfectly to your second Excel file’s columns - the structure has to match exactly.
Here’s what fixed it for me: throw a Set node right after your AI processing. Use it to structure the data properly before it hits the Excel write operation.
Also check if your Excel file is open somewhere else or locked by another process. That’ll kill the write silently and you won’t know why it failed.
If you’re going the HTTP route instead, you might be running into auth problems with Microsoft Graph API.
same issue here - excel writes failing with no error messages. the problem was chatgpt’s response format. it doesn’t always return clean ids. i fixed it by adding a function node with javascript to extract just the numbers before the excel node. also check your column mappings in the write operation - that trips people up too.
Your Excel write failures are probably from data type mismatches. ChatGPT might be sending parent IDs as strings, but Excel wants numbers - and n8n’s Excel node won’t give you a clear error about it. I hit this exact issue building a ticket classification workflow. Add a validation step before writing to Excel that converts data types explicitly. Also, use absolute cell references instead of relative ones when updating rows. What worked for me: read the Excel structure first, then match the exact column headers and data types when writing. The Excel node is super picky about formatting, so treat it that way and your workflow will be way more reliable.