I don't think you need AI but you do need to think about the audience for the printed catalogue.

A few years ago I did the website for a retailer of clothes for the elderly, and they were doing it old school with the catalogue, printed order form and excellent customer service by phone. Their niche was the demographic every other clothing retailer avoided. Unless you have a similar niche, you have to ask about whether a printed catalogue is worthwhile.

AI could potentially help but how do you plan and budget time for that? It could take anything between two minutes and two years to get right. Meanwhile you could do it old-school with artworkers slaving away. Alternatively, you could automate the process to use print stylesheets where you specify the page size and then populate the content with CSS grid layout. The printed catalogue could then be created on demand (and cached) so that it automatically updates itself. This could be a manageable process that you could plan and budget for.

In your product database you could have fields for layout preferences so that you can specify the featured product for each page and what to downgrade in the presentation. I would say this is definitely viable and one reason this is not done is that any company still invested in print catalogues will have an artworker department and nobody in such a department would invest time into automating their job.

No need for AI --- I used to work up automated typesetting systems for a previous employer --- feed in the database as a properly tagged XML file, provide all the graphics in a folder, and a couple of typesetting runs later, one had a fully paginated PDF w/ ToC and Index.

The problem is, no one wants to pay for this since no value is seen in such a paginated view --- even if AI could create such a typesetting routine.

Paginated view is stable. Scrollable is not. This hinders spatial memory immensely, but goes unnoticed.

I mean, what I'm after is a page layout that is designed with compactness and readability in mind. Going from a product database to that requires quite a lot.

An example:

https://archive.org/details/mouserelectronic00unse/page/190/...

A printed catalog also provides discovery that no website has ever matched.

Exactly. So, my hope is that AI can rebuild this lost art that is now deemed too expensive.