TL;DR: You already know how to make a shot look cinematic. The place that control slips is a phone shot, a video call or a selfie, where your film instinct takes over and the result reads as a film still, not a phone. Take it back by keeping the cinematic base and directing the phone-camera grammar on top: name the device, put the body in the lens, add the phone tells. Tell Agent One on invideo to blend both.
Where the control slips
A pure cinematic look is gorgeous, which is exactly why a video call or a selfie comes out wrong. Shallow depth of field, film grain, dramatic lighting, a tripod-stable color-graded frame. But a phone is not shot like a movie, so the audience reads it as a film still and never believes there is a phone in his hand. The polish is the tell, and you handed that detail to the model instead of directing it.
Take it back: direct the phone grammar
Keep the cinematic base, then layer the phone on top. Three moves carry it. First, name the device: a front phone camera, a video-call feed, so the model knows what is recording. Second, put the body in the lens: his arm extended holding the phone, his eyes looking straight into it. Third, add the phone tells: the wide selfie lens, the on-screen call UI, the screen glow on his face.
Let Agent One blend both
Tell Agent One on invideo to keep the cinematic look but render it as a phone POV, and it blends both into one believable shot. You direct both layers. The film grade gives it weight, the phone grammar makes it real.