I'm a senior principal engineer. I use AI at work, but some tasks you should do from scratch — here's how I decide.

Brandon Pedersen, a senior principal engineer at GoTo, emphasizes the importance of not blindly trusting AI-generated work. He advises asking AI for its sources, following the links, and assessing its reasoning. Pedersen, who has been at GoTo since 2013, works as the technical lead of an AI enablement team responsible for maintaining AI platforms, creating tools to assist employees, and providing training and consulting support. He shares five tips for deciding when to use AI:

1. Do not trust AI blindly. Pedersen has always reviewed AI's output for mistakes, which have decreased but still appear in larger tasks. He trusts AI to handle larger tasks after three years of experience but remains vigilant.

2. Be precise with your prompts. Assume AI output can be improved and approach it with a critical mindset. Attention to detail when evaluating AI outputs is crucial.

3. Do mission-critical work yourself. Pedersen was tasked with building an internal AI governance policy and chose to write it himself instead of relying on AI to generate the document. AI-generated policies may appear good on the surface but fail to account for the specifics of a business's day-to-day operations.

4. Know when to delegate - use the two-second rule. Avoid automating tasks that take less than two seconds to complete. Avoid using AI for personal communication or making judgment calls about people, as it can lead to legal risks and not fully understanding the context.

5. Validate and verify factual claims. When AI provides an answer, ask it for its sources, follow the links, and verify the logic. Question the output suspiciously, especially for unfamiliar subjects, to prevent missing errors or misleading information.