I would suggest a lean approach here for starters.
What I mean by that is that it may help to design specific solutions for particular use cases first, put them into practice, and iterate. Perhaps it evolves into a full-fledged cross-community dashboard for the community team, or maybe what’s “good enough” is much smaller than that.
For example, the CM could first use Discourse Hub to be able to see a feed from all three communities and see when each has outstanding flags that need attention.
A step further could be to use chat channel(s) in one of them as “home base” and configure a feed of notifications from the review queue across all communities first, to aid in moderation across all of them.
For metrics and reporting, starting with a set of bookmarks to the admin UI areas or data explorer reports in each community could be a simple starting point.
A step further could be to set up an AI generated report with Discourse Automation that pushes a report to a PM in each, aggregating some data from different sources within that community. That’d still be three separate PMs, but fewer places to go.
A step beyond that could be to use tools within AI personas to fetch data from across communities to build a single report from the data across them.
I think a lot of the building blocks are there and while using them to piece something together may seem like more work, it’s also an opportunity to build exactly what’s needed and prioritize which pieces are more important to put in place sooner vs. later.
And if they all prove valuable and more is desired, something more custom could then be built like a custom plugin with greater confidence that it’ll be delivering the desired outcomes.