I would like to ship some new features just for a specific group of users to better test it in production and then release it progressively to everyone, should i put IFs in my code and assign specific policies to users in the database? Is there a better way to do it?
The normal way to handle this is have two versions of your software. The "main" version is the one most people are on, but you also release an "experimental" version which has the new features.
There are various ways to manage the software, but you should look to use strong version management practices in your source code repository, perhaps using some good branching techniques. You should avoid the two versions from diverging too much.
You can choose to invite certain users to the "experimental" version, or have them opt in but give the necessary caveats that things might not work as well, and if you have any SLAs then you might want to caveat them. If you are hoping users will provide you with feedback then make sure there is a good mechanism for that and that the users are aware of it.
If you have client software then uses will need to get hold of the new version themselves. If your software is purely server side (eg a web application or SAAS platform) then you might look at a routing layer eg in the load balancer which automatically sends users to the normal or experimental version depending on whether they are part of the relevant group.
This is a common scenario in software and you should be able to do some good research. I suggest you start by looking into A/B testing.