ChatBot AI Data Training

[Matt-B]

Hey everyone,

I'm excited to try out the HubSpot AI chatbot but I noticed it only accepts training data through URLs. What's the best way to add data from Word documents, PDFs, or Excel files for the chatbot?

Do I need to put that data on a webpage, or is there an easier method, like hosting the file on HubSpot or Google Drive?

Thanks!

[karstenkoehler · Accepted solution]

Hi @Matt-B,

For now, the AI chatbot is limited to Knowledge Base articles and publicly available URLs. As part of that, publicly hosted pdf seem to be supported – I just added some in the 'Import Sources' section and they didn't throw an error –, are however stuck in review for what seems like forever.

If you have files, you can try uploading them in the file manager and then referencing them in the AI chatbot: https://knowledge.hubspot.com/files/upload-files-to-use-in-your-hubspot-content

For other file types (Word, Excel), the import tool is currently returning an error. Keep in mind that this feature is still in beta and under active development. I'm sure HubSpot will increase file type support over time.

If you want to make sure your feedback to include more file types is heard, I'd recommend posting in the HubSpot Ideas section of the community: https://community.hubspot.com/t5/HubSpot-Ideas/idb-p/HubSpot_Ideas

These requests submitted to the HubSpot Ideas section of the community are reviewed by the HubSpot product team, based on their popularity and the assumed demand.

Have a great day!

[SLiu168]

I tried upload Pdf files in "Files" in Hubspot. It seems that the documents never get to be synced. 

[karstenkoehler]

@SLiu168 you're correct, PDFs seem to be stuck in the review stage – something that HubSpot is likely going to fix. Keep in mind that this is still a beta feature.

[SLiu168]

My manager doesn't like Pdf files to be the knowledge base of AI since Pdf is not easy to be updated by itself and prefers to use URL of web pages. However, there are some web pages Hubspot AI cannot recognize, like this page. https://www.norsat.com/products/3000-c-band-single-band-pll-lnb?variant=41050456817847. After it is synced, the specification  of the product  is not extracted from the page. Did you have the similar experience? Thanks