It's not hard to tell people to do that. It is hard to get people to really comply and avoid changing behavior just by enrolling them in a study. It's hard to select 100 workplaces that are similar enough that they can be compared to eachother and also representative of the myriad complex types of social areas that people congregate in - opening a window might work in one type of office but not at all in another situation. And also, you need a control. You can tell a group of people to not change their behavior in a global pandemic to see what the impact of each intervention was. But getting people to comply? You even have problems in clinical trials where groups of patients find ways to get together and discuss side effects they've been having to try and figure out who has the placebo and who has the real drug killing the trial. Running statistically rigorous trials on that scale to get good information is just really difficult because of the study subjects.