We experimented with various AI-powered features at incident.io, learning that investing in team expertise and tooling upfront is crucial for successful implementation. We found it helpful to discover what's technically possible through short experiments and iterated quickly to build conviction in the team about where AI-powered features would be most compelling. Investing in tools and developer experience up front allowed us to experiment faster and test more ideas, resulting in a significant reduction in development time. It's equally important to invest in tooling as it is to invest in the team. We also learned that having principled product principles, such as keeping a human in the loop and subtly automating existing workflows, can help narrow down what to build and how to build it. Additionally, using large language models during development can be incredibly powerful, but it's essential to stay up-to-date with foundation model changes and focus on the big picture. Launching AI-enabled features in phases, being close to customers, and not waiting for perfection are also critical for success. Ultimately, building AI features requires a combination of technical expertise, product principles, and a willingness to learn and adapt.