I think I am in one of the camps that says no to entry level data scientists, so keep in mind I'm presuming this person has some data experience already and my focus is on the side of analytics groups (which probably has the majority of data scientists though this is anecdotal information).
They should understand the basics\* of data self support (e.g. how to write SQL queries, table joins, ability to get data from different sources and bring it together, and ability to get automation set up for same), understand the basics\* of algorithms (e.g. what types are available for kinds of problems and be able to get most of them set up and running even if it's not tuned yet, this includes programming and output necessary), understand how to interpret results of statistical tests (e.g. what a chi-sq looks like and why you'd use it), and the ability to implement those things conceptually to solve a business problem. They probably also need experience with solving some kinds of business problems with data.
I'd also like them to be able to do some basic requirements gathering but that's not for everyone and in some places you might have other folks to help with that part.
\*Basics are important here - you can have data or ML engineers help with high scale automation, but you should be able to get what you need for initial analysis or pilot-scale stuff yourself