Can ChatGPT decipher workflow hand-written on a whiteboard?

For today’s workflow analysis conversation with ChatGPT, I hand-wrote a basic loan workflow on a whiteboard using various colored markers of questionable markability. After taking a picture of the completed workflow, I entered my analysis request in ChatGPT and uploaded the whiteboard image. The purpose of the conversation was to determine ChatGPT’s ability to decipher workflow data written with marker on a whiteboard.

ChatGPT PLUS attempted to use Optical Character Recognition (OCR) to analyze the whiteboard information. (I didn’t even know it had OCR capability.) The OCR failed. ChatGPT switched to “manual” analysis and provided a response to my request.

Despite the fact the handwritten text was difficult to interpret, ChatGPT accurately identified most of the processes, as you’ll see in the screenshots below. Key takeaway: if you intend to submit a whiteboard workflow image to AI for analysis, ensure each word is clear. Eliminating the hurdle of text deciphering will pave the way for meaningful analysis on the first or second pass.

Also, some very interesting questions came out of this episode.

Why did the OCR fail?
What is the difference between AI using OCR to analyze the flowchart and using '“manual” analysis?
And what, precisely, occurs when ChatGPT “manually” analyzes?

I asked ChatGPT and will share the answers in my next blog entry.

Enjoy the read!