The Kerr metric has an explicit $g_{\phi t}$ component. From this component, the timelike killing vector of the spacetime, which is the "direction" of the timelike geometric symmetry, inherits a $\phi$ component from here.
If you attempt to keep an object at a constant $\phi$ coordinate while it evolves forward in time, it will not respect the time evolution symmetry, and this will show up as a net force on the object that makes it want to "orbit" along with the direction of the "frame".
A general metric tensor for a spinning physical object will be more complicated than the Kerr metric, but you will still have this same general picture emerge.