use Airtable AI to analyze pictures and documents

[Kai_Muscher]

Dear community,

I am thinking about to turn on Airtable AI, but first I'd like to ask whether the Airtable AI has the power to do what I want.

A brief introduction of my base and what I want to achieve:

We provide logistics management for cruise vessels when they are in dry dock. We manage the complete supply chain and on site coordination. Building a detailed inventory of all items for the ship and keep track is one of our tasks. 

Now, we work with barcode tracking which means we have to assign a barcode to a record (primary field = barcode). This can be done by scanning the label (regret to say that, but the Airtable mobile app sucks!) , or manual input (risk of typos), or third party scanning integration. We also take photos of every single item (attachment field).

And here is what I want (primary goal):

• AI should analyze the photo and search for a barcode label, then update the primary field accordingly.

So much more what I want, for e.g.:

• AI should analyze the packing list (attachment field) for cargo description and update cargo description (text field) of the record.
• AI could analyze a photo of a label attached to a pallet and retrieve the description. I did that successfully with ChatGPT. ChatGPT was able to recover text from very poor quality photos.

...

Can someone tell me if Airtable AI has the capability to achieve what I want? If not, maybe there is an alternative way to look for.

Thanks a million!

Kai

[Sachin_191]

Hey Kai 

Welcome to the community

Hey there,

we can integrate AI to automate barcode and cargo description updates in Airtable. By leveraging AI, we can analyze photos to detect barcodes and automatically update the primary field, minimizing manual input errors and enhancing efficiency. Furthermore, the AI can parse packing list attachments to populate cargo descriptions and extract label information from photos, ensuring accurate and streamlined data management.

[magicalpotion]

@Kai_Muscher As you already tried it using ChatGPT, you can "easily" integrate ChatGPT using Make (ex-Integromat) with their Analyze Image (Vision) module. Flow is as follows:

1. Setup an Airtable webhook that is triggered from an Airtable button (you can also watch new records)
2. Fetch the Airtable record
3. Call the ChatGPT Analyze Image (Vision) module from Make using your prompt, the attachment URL and picking the model you wish. For the demo I used "Extract barcode as number"
4. Update the code field in your table

You can engineer your prompt so that ChatGPT only returns the number so that you don't need any further parsing. Also you will need some error handling when ChatGPT doesn't manage to analyze the image. But this is the simplest design to kickoff your flow.

NB: you need to fund your OpenAI account and generate an API Key to trigger the module.

Just in case you didn't saw it, Airtable offers native barcode reader: https://support.airtable.com/docs/using-the-barcode-field-in-airtable.

For further assistance DM me, I'm an automation dev-consultant (Airtable/Make/OpenAI).

[Rutger]

Adding the ability to use attachments in the Airtable AI fields would allow us to summarize documents. Having it natively in Airtable would be much easier than going through Make etc.

[Rutger]

Cool, I saw this feature has now been implemented. Although not yet for Excel documents. Is anything on the roadmap for Excel?