AI Agent Use Cases
Views —

AI Agent Use Cases

Added today

10All 241

1–20 of 241

Work & ResearchWalkthrough

Automation That Turns Whiteboard Photos into Meeting-Minutes Documents

When a photo of handwritten meeting notes is sent by email, it extracts attendees, a summary, and action items and creates a Google Doc.

n8n

How it works

  1. The creator built an n8n workflow that starts when a whiteboard or notebook photo is sent to a dedicated inbox.
  2. It extracts the meeting title, date, attendees, summary, and action items from the image, creates a Google Doc in a Drive folder, and appends a full transcription at the bottom.
  3. The creator said the key was specifying in the field descriptions that unclear handwriting must not be guessed and should be returned as an empty value.

Results

The creator said it takes about one minute from receiving the photo to saving the document, and that the reply includes the document link and the number of action items found.

Notes

The workflow and setup instructions are available in the creator's GitHub repository. The post provides no measured handwriting-recognition accuracy.

Source

Reddit Original post ↗

Example prompt · n8n

Build an n8n automation that turns an emailed photo of handwritten meeting notes into a meeting-minutes document.

Inputs: dedicated email address for receiving photos [address], Drive folder for saving documents [folder], fields to extract [meeting title, date, attendees, summary, action items and owners], and email address for receiving the result [address]. First, verify that email triggers and Google Drive and Docs connections work in my n8n environment, and determine which model to use for reading text from images.

Implementation steps: 1) Receive image attachments from a new email. 2) Extract the fields from the image into a structured format, and state in each field description that difficult-to-read handwriting must not be guessed and should be left blank. 3) Create a Google Doc, organize the fields, and append the full transcription at the bottom. 4) Reply to the sender with the document link, the number of action items, and the fields left blank. If there is no image or it cannot be read, do not create a document; report only that fact.

Deliverables: a workflow file and setup guide. First, test it with one clear photo and one intentionally blurry photo, and verify that it does not invent names or dates in the blurry image.