Many AI systems make costly mistakes, due to a number of problems which are inherent to the algorithmic methods they are based on. Even though these types of AI failures cause clear damage, they are not always easy to detect or avoid. Deepchecks is an early-stage start-up, tackling this huge problem with a unique algorithmic approach. We focus on warning organizations before their deployed AI systems make bad calls, and offer a variety of solutions for dealing with different types of AI failures.