The Atomic UX Research is a methodology born around 2016 to organize findings into four linked levels: experiments, facts, insights, and recommendations. With this way, the evidences stay traceable, but it historically failed due to maintenance burden, difficulty spotting patterns at scale, and poor accessibility outside research teams.

I wrote a Medium article which describes how I build a Claude skill to fix this.

The skill runs three workflows:

• Creating a new experiment.

• Cross-referencing new facts against existing data to flag overlaps/insights.

• Pushing approved data into the repository.

I estimate this skill can save ~4 hours of manual work and reflexion per study.
It's more a human-machine conversation than a job replacing. The automation skill handles only the plumbing part, which let me the time to manage the reflexion without dedicating my day for that.