Speed is the easy part.
Building a working application in days is a solved problem. What determines whether it is still useful in two years is everything around it: who owns it, who may change it, where the data lives, how it is tested, what happens when the person who built it leaves, and whether it is one of five applications doing the same job in different departments.
Unmanaged, it becomes shadow IT with a licence.
The failure mode is not a bad application. It is forty applications nobody catalogued, with overlapping data, inconsistent access control and no clear owner, built by people who have since moved on. That is harder to unwind than the spreadsheets it replaced, because it now has users who depend on it.
We put an operating model around it before scale.
Decide which classes of application belong on the platform and which do not. Agree environments, naming, access model, data boundaries and a review gate proportionate to risk. Establish who supports what. Then build, with the first few applications setting the pattern for everyone who follows.
What you are left with.
Applications delivered in weeks rather than quarters, a catalogue that says what exists and who owns it, a review step that is light enough to be used, and a clear line between what business teams may build themselves and what needs engineering.