A world model opens up possibilities for the AI. Reactive techniques can operate with memory and significantly benefit from it. That said, there are limitations with reactive techniques (for instance, memory consumption) that make them unsuitable for large problems—hence the need for architectures and decomposition.
Search algorithms shine where reactive techniques fail (rarely in game AI), trading off additional computation for memory. In particular, search techniques are appropriate for establishing plans in the future, considering the consequences of each of the possible actions.
In terms of increasing the capabilities of the NPC, planning techniques based on search algorithms have an important role—although it should not be overestimated. Pathfinding aside, deliberative AI still hasn't found a place in the games industry. (Only a handful of games use deliberative AI.) The web site contains some good starting points for investigating techniques.