Due to its novel, significantly different integration method ApartCI requires a mental shift, so you may want to:

Create a project configuration. You can start with a simple one which you can gradually refine. Peek at existing public configurations for inspiration.

Simulate your configuration. It will give you an idea of how a real project using that configuration will behave.

Create a test branch and update your configuration for it. Create a real ApartCI project from your configuration and start developing/testing your Agent. We'll soon have some examples and skeleton projects you can use.
It is not mandatory to automate 100% of the agent functionality to be able to get your project running, manual operation is perfectly fine as long as the overall performance expectations are adjusted accordingly.
Take advantage of the free quotas during this time, until you get the entire end-to-end flow functional.

Pick one of your busiest, most difficult or unstable project/integration branches and pull a test branch shadowing it. Create and activate an ApartCI project for this test branch. For every changeset committed to the integration branch also submit a corresponding changeset to the ApartCI project.
The side-by-side comparison of the speed and stability of the two branches will outline ApartCI's advantages. You may even use the ApartCI decisions to speedup recovery from the breakages from the integration branch.
You need to keep in mind that the test branch will not contain the rejected changesets, so the two branches will gradually diverge. This will make porting the changesets from the integration branch into the test branch increasingly difficult, so you'll have to make efforts to cancel the differences now and then. Or simply re-set the experiment using a fresh test branch.

When you decide to switch the actual integration branch to ApartCI you can just clone your test project configuration and update it for the actual integration branch.