Linear regression, logistic regression and k-means clustering, if you can get a project into actual real-money production on one of those, you are already well ahead of 90% of data scientists. And these techniques are decades old!

But also recall, that there are alot of people with general computer backgrounds and a whole lot more real-world work experience who could pick up the basics of data scicnce stuff as well..

Places where these jobs tend to be needed are often very highly placed and very strategic, with the decisions being made based on the data very critical to the overall company health. Some of what is being requested by requiring this level of credentials is real-world work experience and a certain level of maturity/professionalism that is often implied by this as well..

If the level of skill is basic enough that any other regular employee could learn it in short order, why wouldn't they and then have the cushy 'sit around and play with numbers all day' job too.. There needs to be enough 'there there' to placate internal politics as well..

Further, having a more advanced person would also likely need to be taking leadership roles in picking and rolling out solutions with long term costs associated with them, training other employees, etc, which also comes with the professionalism/experience part..