Commonly found in human decision-making, cognitive or heuristic biases are mental shortcuts that may help individuals make more efficient decisions but often result in errors. We propose that Human-AI decision systems should be designed utilizing knowledge of how to mitigate cognitive biases and thus improve human decision making. Our research examines whether algorithmic advice can mitigate human biases and how the timing of such advice influences the extent of bias mitigation. We focus on conservatism bias, a bias which causes individuals to underreact to new information, in financial decision making. The experiment compared three conditions, varying the timing of algorithmic advice. The results indicate that timing of algorithmic advice matters – individuals who received algorithmic advice prior to making a decision exhibited less conservatism bias than those who received algorithmic advice after making an initial decision and those who received none at all.