It really depends how strong the client brand is and how competent the client staff are that you’re working with.
But the important thing is to take an evidence-based approach.
A good idea is to keep it low fidelity for as long as possible until the UX is worked out and tested. Then high-fi it along the brand guidelines.
Then test test test, again.
The best way of getting a good result for your client is to take an evidence-based approach where the initial product is set up as an experimental baseline and further improvements are tested against that baseline, be that through automated A-B testing or focus groups of representative users.
Focus on getting something small to market as soon as possible and adding/improving features in an evidence-based way.
If the client has feedback you don’t like, test it anyway and charge them. They might be right.
Focus group testing should take place before presenting to clients. Use paper prototypes with paper cutouts or using an app such as POP.
Don’t guess! Test!
As a rule when doing UX you want three possible designs (one of which should be a bit “out there”) and you should test each design with representative users. You learn a lot more from in person research than from A/B testing. A/B testing should always be used to verify changes or features. But you need to focus group at the design stage.