Dogfooding, or using your own products internally, helps catch bugs, but it can't replace user research: your team knows too much to represent real users.
Pressure to adopt AI isn't evidence that a tool helps. The PROVE framework tests one tool against one task and produces a provisional decision you can defend.
AI-simulated users can fill in missing data and predict population-level trends. They perform better when they are based on extensive contextual information.
Convenience sampling is quick and cost-effective for UX research, but probability sampling is better when you need representative and generalizable data.