How about asking almost any question about your *structured* data?


Here is how I would want to use this ability:

I will write comprehensive documentation for my company-wide info system that explains all the different DBs, how they are related, and all the different common procedures (e.g. how to create a new Client, a new Project, etc).

I would like to see a “Help” button or field at the top of every page where users can type a natural language query to be processed in the context of the custom Help docs I have written - e.g. the user might ask “how to create a new website project for this client”.

Also fun would be automatically generated context-sensitive help for any page (generated from these same Help documents) - i.e. a list of “things you can do from here”, common processes, etc.

This could be very useful as a floating/popup box that follows the user across different pages, collecting additional context, giving additional hints, etc.



That’s awesome! Would it also be possible to analyze combined data (if the data is present and correct of course)?

We have data such as

  • LTV of a contact
  • First (marketing) channel and most important (marketing) channel
  • Amount of touch points during customer journey

Then it would be awesome to let AI analyze that data (so that you don’t need a report anymore):

  • What (marketing) channel resulted in the contacts with highest LTV
  • How many touch points did the contacts with highest LTV have


Quick idea → on the top of every entity we have an ‘explanation’ entity with ‘how it works’.

If user clicks, the help text opens split screen.



Excited about this!!!

This is definitely the dream. I have been surprisingly impressed with the AI help on website host Siteground, FWIW. They appear to possibly be using Cypress AI:
Totally different system than Fibery of course, but possibly an example of something similar in action, in terms of providing actually effective, contextual help based on a large corpus of proprietary data/knowledge.

Oh and I forgot to mention this interesting related announcement: