$\newcommand{\g}{\mathfrak g}$
Let $G$ be a reductive group and $U_q(\g)$ the associated quantum group. One can argue that the classical limit of $U_q(\g)$ is $G$ or $\g$, with some Poisson structure, but here's a picture I like better about the category of modules over those:

The basic observation is that a $G$-module is the same as a $G$-equivariant quasi-coherent sheaf on a point, i.e. a quasi-coherent sheaf on the quotient stack $BG=pt/G$. This stack has a 2-shifted symplectic structure, and one can define a Poisson sigma model which to a topological surface $S$ attaches 
$$QC(Map(S,BG)).$$
So this thing is a categorified 2-dimensional topological field theory, or equivalently the 2-dimensional part of a would be 4-dimensional field theory. Note that $Map(S,BG)$ is identified with the moduli stack of $G$-local systems on $S$ so this thing really is the classical limit of Chern-Simons theory. This is in accordance with the old idea that Chern-Simons/Reshetikhin-Turaev really is secretly a 4, not 3, dimensional TFT. 

Hence it is expected that Khovanov homology and its generalizations should come from a partially defined 5-dimensional TFT, which mathematically should be computed by a still conjectural braided monoidal 2-category structure on categorified quantum groups. Witten has a lot of ideas about the physical side of this story, which I honestly don't understand.

So I'm wondering if at least the classical part of this theory has a nice description similar to he one above, i.e. a nice geometrical construction from $G$ of a 2-dimensional TFT valued in (dg) 2-categories.

**Edit**: To be a bit more precise, I guess what i'm really looking for is some natural way of associating some geometric object $Z(M)$ to a topological 3-manifold $M$ in such a way that for a surface $\Sigma$, $$Z(S^1\times \Sigma)=Map(\Sigma,BG).$$
Then $QC(Z(\mathbb{R}\times \Sigma))$ will be a monoidal category, i.e. a 2 category, and $Z(\mathbb{R^3})$ should then automatically be a braided (in fact symmetric, probably) monoidal 2-category categorifying $G$-mod.

I feel like such a $Z$ should exist and be defined purely in terms of $G$ from some physical considerations, i.e. without assuming a priori the existence of a categorifcation of $G$ but rather with the aim of producing such a categorification from "first principles".