I’ve worked as a SaaS PM for the last 5 years, largely focusing on AI the last two years or so (open to new challenges btw ![:wink: :wink:](https://community.plone.org/images/emoji/twitter/wink.png?v=12)
Some thoughts.
Aspects of content mgmt that come to mind that are relevant for use cases where AI could be used:
- generation of content
- analysis and reporting of content
- content augmentation
- search & retrieval
New content generation is something that mostly happens between the user and the genai system so there is relatively little added value the CMS can provide.
Otherwise, it’s largely about what only the CMS can provide: accessing existing content in it for use in various kinds of RAG pipelines. There’s at least three ways in which you can get content from a CMS:
- via framework / platform APIs or feeds
- accessing the DB or search engine index directly
- via rendered page & metadata
In practice, direct DB / search engine access is rarely the best option or even possible. And while customers frequently expect you to ”just get it from the site, fetch the HTML”, it is rarely a good way for real enterprise solutions either.
For plone to shine as part of enterprise solutions utilizing AI, great APIs and extensive indexing, metadata, search and content feed support are what is needed - and afaik plone has them all. Many things should and could be improved of course and the devil is in the details: Once you actually try to e.g. just sync all content from plone for generating & storing vector embeddings, you’ll find out what needs improvement. Lack of metadata has always been a major problem in CMSes and still is so using genai to add that automatically might potentially help things a lot.
Overall, semantic web tech is having a renaissance thanks to AI. Generating an ontology and a ”knowledge graph” and linking the pages to it is something very interesting that might even make sense to add to the core, although it could be done via an external system as well, if such system is used anyway (to feed a vector db etc).
A lot depends on the site content to be used: To what extent it’s mostly text, or audiovisual, or structured ecommerce product data etc. and how will it be used in the application.
So, real use cases and true customer needs (hidden behind what customers say they want) are key.
What kind of AI related use cases have actually come up in conversations with customers and users of plone that plone should / could help support?