A.T. said:
I don't think so. We often talk about macroscopic vs. microscopic kinetic energy, but this just shifts the definition to what "macroscopic" and "microscopic" is.
I played with a simple one dimensional model.
Scenario 1:
Particles on a line have a uniform velocity distribution from ##-V## to ##+V##.
The Shannon entropy is ##-\frac 1{2V}\int_{-V}^V\frac 1{2V}log_2(\frac 1{2V}).dv=1+log_2(V)##.
The energy is ##V^3/3##.
Scenario 2:
We now add the same velocity ##U## to every particle.
The energy increases but the entropy remains the same.
Scenario 3:
Instead of adding ##U## to every particle, we add ##U<V## to half and ##-U## to the rest. This leaves the mass centre of the system static.
The velocity distribution has the original density in ##(-V+U, V-U)## and half that density in ##(-V-U, -V+U)## and ##(V-U, V+U)##.
The Shannon entropy is now
##-\frac{4U}{4V}log_2\frac 1{4V}-\frac{2V-2U}{2V}log_2\frac 1{2V}##
##=1+log_2V+\frac UV##
an increase of U/V.
Note that this makes sense at both U=0 and U=V since at the latter the scenario is equivalent to doubling V in scenario 1.
The energy is ##\frac 13(V+U)^3##.
Scenario 4:
As scenario 1 but increasing V to V+U. This produces the same energy as scenario 3 but the entropy increase is instead ##log_2(U+V)-log_2(V)=log_2(1+U/V)##.
Over the range 0<U<V, ##log_2(1+U/V)>U/V##, so the entropy is lower in scenario 3 than in scenario 4. That difference represents available work.