Suppose there is a primitive human settlement on the edge of a dense forest. The forest is very difficult to traverse, especially with equipment. However, the villagers need the fruits of the forest, so they create a pathway through the forest. Eventually, the pathway becomes a road that reaches the other side of the forest. One day, by chance, a villager uses the food trail to take some digging tools to the other side of the forest, where he digs for minerals. There, thanks to his tools, he finds an immensely valuable mineral. Soon, the forest road is used as much for mineral transport as for food. Finally, the village learns to plant nutritious crops in the flatlands near the village, and stops collecting low nutrition foods in the forest.
Now, look at the final state. The villagers are using the road only to transport minerals and tools for digging. Yet, if the road had not existed, they could never have transported the tools to make the initial discovery. It's just too improbable that the village folk would create a road to nowhere at massive expense, transport their equipment along this trail, dig, and find minerals. Since the villagers don't feed in the forest, and have forgotten that they did, this system is apparently irreducibly complex (IC). Not knowing this history, it appears that the villagers would have had to know about the minerals before they built the road and dug for them in the right spot.
By co-opting components of a system that serve alternate functions (nutrition, not just mining), one significantly reduces the improbability involved versus having all the components appear at once. You get the illusion of foresight.
And yet, intelligent design advocates like Behe and Dembski continue to deny that IC systems can evolve through incremental steps. This is willful blindness (or worse). Their claims have been thoroughly refuted, but they continue to make them.