You mean misunderstanding the data, coming to the wrong conclusions, etc? Data science always has an issue with bullshit KPIs, shallow depth of statistics, and mostly mangling stuff keeping the manager happy. Still it's much better than not having any data analysis.

Whether it benefits from being in a single datalake idk. We really don't know how the operations are being done, we're mostly just reacting to news reports and outside guessing.

I'm assuming it will be basically how Palantir works in government health care and intelligence agencies where they aggregate multiple data sources from a bunch of old and new databases and have complex analytical tools on top.