A developer is tasked with optimizing a search algorithm for a large dataset of 1 million sorted records. The current implementation uses linear search which takes too long. The team suggests implementing binary search instead. What would be the time complexity improvement?