How to use AI automation to extract information from an image in Timetonic.
Automate the analysis and extraction of attachments using Timetonic's AI automation.
This feature makes data integration easier for the management of expense reports or business cards and many other specific use cases.
Before you start, if this is your first time using AI automation in Timetonic, you can watch the setup video right above it.
Let's look at this together in detail.
Here's what we'll be able to do once the image extraction configuration is in place.
I've just inserted an attachment, so my expense report.
I will validate my form and the AI will take care of processing the analysis of my document and therefore categorize, indicate the information, the price, the date and the categorization.
Let's go.
I'm in the table where I want to process data from an attachment.
I go to Automate, then Automate with Timetonic plus LIA and I add an automation.
I name this automation.
I give it a trigger that says when it will receive the attachment by securing an extra condition to say that the field is not empty.
And I'm going to add the automation and I choose ChatGPT's AI.
I arrive in the configuration window of ChatGPT.
I'm not showing it in the tutorial, I previously integrated the API key of my ChatGPT, and you can find, as I said, the information from the configuration tutorial of this window.
Let's go straight to the description of the role we're going to give to Lia.
I tell her you are an OCR expertise system allowing to extract information from expense report.
It therefore has its role.
I will then give her the question or more commonly use a prompt.
And here, I tell her how it's going to analyze the execution of this extraction.
So, in the expense report provided as an attachment, extract as much information as possible in JSON format in the following properties.
Please, read your result well and recompare with the original image to avoid errors.
I give it the information that it will find in my attachment.
Description, name this expense report in two words with the elements you find.
Detail, find information on the expense report and interpret the context in five words.
Example, lunch with potential client to discuss the project, seminar.
I ask her to find the total amount paid.
I ask her to find the date and time and I give it the format to use.
I give it the categories.
With the elements from the image, define a category among these ones: equipment, meals, outings, transports, health and others.
Categorization in multiple choice.
I will then here chose the column of the attachment which is the receipt For the response format, I will select the JSON processing so that the property values are stored in the destination fields that we are about to configure.
I will give again the name of the elements that it should find on the attachment and I will link it to the corresponding field in which it should insert the information.
And I will proceed like this for all the elements that I ask it to extract from the attachment.
You would have previously defined these columns accurately to address them in your automation.
Carrying out a second test account to test its effectiveness and there it is well the case We name in categorization of transport a train ticket, we determine what type of trip, cost, date and time, when you reproduce this tutorial for your specific need, but that the result of the AI does not bring you satisfaction, iterate on the prompt by making modifications until you get the desired result.
You will have understood, this evolution combining AI and the nocode of antenna tonic to extract data from an image that will optimize the management applicable to infinite specific cases.
Booster.
Test it quickly to boost your efficiency.
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
