MicrosoftNightmareRambo77 - Topics is not intended for small data sets (so it can be a challenge in a non-prod tenant).
It works best when it can work its way across / discover at least 20,000 items. I won't pretend to know the ins and outs of "why", but it relates to the large scale compute behind the scenes and machine learning which does not perform well on small datasets (if you look into ML you will often see reference to complex models relying on large sample sets etc).
It will in any case take several days (anywhere up to 2 weeks we are seeing) to start providing good Topics results, and it will continue to discover topics over time. It will also improve over time as connections to topics and the topics themselves get curated.
If you are looking for a winning business case, then start with the business problems / user scenarios that relate to discovering (and making sense of) data that otherwise might be "in the dark". If you have suitable user scenarios where the data/people are in M365 today, and where you have high enough data volumes, then Topics may be suitable help you build the connections between the data and the people.
