A high entropy isolated system is shuffling through its more probable micro-states. Local entropy dip produce several structures. These structures, akin to life, feed on free energy to preserve themselves. They then increase in complexity as greater complexity confers a higher preservation rate. Greater complexity imposes more structure on the system, restricting the number of accessible micro-states.
Has entropy of this isolated system decreased?
No. Note that the system is isolated: the overall entropy is still going up. The life is exploiting local disequilibria or free energy to lower their own entropy while increasing the total entropy. If they cannot do this, then they cannot protect themselves from ending up in likely sets of microstates. In a high entropy systems this is hard to do.
The fact that they came about randomly doesn't matter when we look at entropy on the macroscale. You might be thinking about how rare random events cause unlikely patterns to appear in a thermalized closed system: wait long enough, and an ergodic system will end up in super-unlikely low entropy states. This kind of Poincaré recurrence happens in finite systems since the second law strictly speaking only applies to infinite systems. However, the time to wait is typically doubly exponential ($\sim\exp(\exp(N))$) so in practice this is not happening.
Life is actually a good illustration of the second law of thermodynamics, stating that the entropy of the closed systems can never decrease.
For example, a living cell can be viewed as a heat engine, which absorbs energy from the environment, uses this energy to perform work (i.e. building structures, such as proteins, nucleic acids, etc.), and then rejects the unusable heat to the environment. The seeming decrease of entropy (increase of information) in the cell itself is compensated by an even greater increase of overall entropy due to the sugars, supplying the energy, being split into more elementary molecules.
