How to approach unstructured email triage

Hello all! I am trying to figure out how to create a solution for this email triage use case, but I am hitting a wall…

For context, I am a designer by trade, actually working in AI dev tools, but I am not a developer. I have basic html, css, and JavaScript skills…

I am trying to create a workflow that can scrape emails for payments (via Zelle or Cash App), identify people who haven’t made their payment, send late notices to those people, and send a summary email to the manager of who paid and how much, and who didn’t

Getting the emails has worked fine. I pull in emails from the last month, then filter out all the ones that don’t match specific conditions. I am not sure exactly what the best pattern for a solution would be, but I am trying to append all the payments into a google sheet so I can then compare it to another sheet with the list of months payees to see who hasn’t paid.

At one point I had created an agent that could do this reliably, but I was using samples from my own email. Once I started using the actual emails from the clients accounts, things stopped working. Before I was using “snippet” from the email, but with the expected email, sometimes the payment amount gets cut off in the snippet. I started using the full email text, but there is HTML in there.

What are some methods I might try to find a path forward? THANK YOU in advance! I am new, but I am really loving all that I can do with n8n.

Hi,

What kind of info does the payment provider issue upon payment. What can we match to?

From what I understand each app support email for sending payment requests. Wouldn’t it just be possible to retrieve the payment requests for a certain period and check the status? Anything other than paid

If your requests come in from the same email address or even from people from the same company domain you could block.

Another approach would be.
Have one workflow to check payment status and send reminders and another process to label incoming email as paid or unpaid (only for internal reference)
This way the process is less prone to errors involving direct client interaction on random mail content

Reg
J.