What does data science actually cover?
The path from raw records to a defensible answer. That means acquiring and cleaning data, exploring it, building statistical or machine learning models, and communicating what the result does and does not support. The toolbox spans simple aggregation, clustering to find structure, forecasting, and experimentation, where A/B testing is the workhorse for causal questions. Most of the discipline is making data trustworthy enough to answer anything at all; the model at the end is the visible tenth.
Data science vs AI: what is the difference?
Direction of travel. Data science begins with what happened, your records, and works toward understanding and prediction that inform a human decision. Artificial intelligence begins with a capability target, a system that writes, sees, decides, and works toward behavior. The fields share mathematics and infrastructure, and in practice data science builds and validates the datasets AI learns from. One produces answers for people to act on; the other produces systems that act, which is also why their failure modes differ: a bad analysis misleads a decision, a bad AI system misbehaves at scale.
How did LLMs change the data scientist’s job?
They automated the middle and raised the stakes at both ends. Models now draft the queries, the charts and the boilerplate analysis, so the routine reporting layer thins out. What grew is the ends: upstream, owning data quality and structured data foundations that both analytics and AI systems depend on; downstream, validating what automated analysis claims, because a fluent wrong answer travels fast through an organization. The scarce skill is no longer producing an analysis but standing behind one, and teams staff accordingly: fewer dashboard factories, more data owners embedded where decisions are made.