Data science is the work of answering questions with data: gathering it, cleaning it, exploring it, sometimes building a model that predicts something, and telling the result so someone can decide. It mixes statistics, programming (Python and SQL, for example) and understanding the business. Machine learning is one of its tools, not the whole job.
A question from the app. Pick an answer.
Management asks why sales dropped in March and wants to see it in a chart. Who?
Gets the data ready to query. Explaining what happened in March and telling it with charts is another job.
Works with models running in production. Explaining a drop just takes querying and comparing data, no training.
Queries the data (often with SQL), compares months and tells the story with charts management understands. Answers business questions.
A store wants to know which customers are about to stop buying. First the question gets sharper: what counts as "stopping"? Then the sales data is gathered, cleaned (wrong dates, duplicate customers) and explored with charts. Only then, if it helps, a machine learning model is tried to flag customers at risk. And at the end you explain what you found and how sure you are.
Cleaning and understanding the data usually takes longer than the model. Errors in the data can carry over into the model. And if it does well on the training examples but poorly on similar new cases, that is overfitting.
They overlap, but they are not the same. AI aims to have a machine do tasks that look intelligent. Machine learning is one way to get there by learning from examples. Data science uses machine learning when it helps, but many questions are answered with a good query, a well calculated average and a clear chart.
Data science jobs show up under English titles, and they overlap: a Data Analyst tends to focus on analysis and reporting; a Data Scientist, on experiments and models; a Data Engineer, on preparing and moving the data. What each one does is in AI and data roles. Entry level ads usually ask for SQL, Python, some statistics and being able to explain a result to someone who is not technical.