Point 1 interests me because I have seen it play out.
I work for a small unit of a large organization now and we have a highly formal process for buying SaaS subscriptions. You fill out some paperwork explaining why you need it, in 1 to 2 weeks you tend to hear it is approved or (best of all) the large organization has a site license already they can put you on.
You have to have some brakes on subscriptions.
I worked at a small co where we had maybe 20-30 different SaaS tools that we stored "documents" in. Every week we had an all hands and somebody would bring up the problem that they can't find documents and that the answer is that we need to have a new place to put documents.
The group seemed impervious to the obvious objection that having more possible places to put documents would make it harder to find documents. So we wound up with a lot of SaaS subscriptions, a lot of talking, but very little listening.
The obvious business response that to problem is to build "one ring to rule them all"; e.g. a small instance of the thing that the n.s.a. uses to read your email, sans the cryptoanalysis bit.
That's an astonishingly hard sell. On one hand that system is an existential threat to all SaaS vendors from Salesforce.com to the very smallest. So you know you'll get resistance.
Even though Google is great at making money from ads, the search is not impressive in 2020. (Look at the second page of results, then the third, and ask "why do they bother showing any of this crap?") Corporations still buy OpenText. Ask people what is a scalable full-text search engine and they say "Solr" (have you really seen how it executes queries?) The only full-text search I like these days is pubmed.
We can do better.