The use of AI coding tools has become increasingly necessary in software development, but most companies and developers work with large, legacy codebases that can be millions or billions of lines long. The cost of mistakes in these environments is huge, with estimates suggesting over $2 trillion per year. To harness the benefits of AI coding tools responsibly, developers need to establish clear best practices, including ensuring no obvious bugs or vulnerabilities, maintaining performant and robust code, and keeping it readable for humans and large language models. Organizations should adopt seven key "habits" that promote accountability, simplicity, and security in their use of AI coding tools, such as the Golden Rule, over-documenting project context, keeping code simple and avoiding stray code, analyzing everything, writing mandatory unit tests, conducting rigorous code reviews, and prioritizing prevention over cure.