I’d focus on understanding data at a deeper level rather than just learning how to query it.

AI can generate SQL and build reports, but it can still produce technically valid outputs based on incorrect assumptions. Understanding data grain, join cardinality, NULL handling, metric definitions, and how aggregation affects results is what helps an analyst catch those errors.

I’d also prioritise statistical reasoning and experimentation. Knowing whether a change is meaningful, whether a correlation is misleading, or whether a metric has been affected by selection bias is much harder to replace with simply generating a query.

Another important skill is understanding data pipelines and lineage. When two dashboards show different numbers, an analyst should be able to trace the discrepancy back to transformations, source systems, or inconsistent business logic.

AI literacy matters, too, but not just prompt engineering. It means validating generated queries, testing edge cases, and understanding the limitations of the tools being used.

SQL is becoming easier to produce. Knowing whether the result is correct, meaningful, and useful for a business decision is where an analyst can differentiate themselves.

I think junior data analysts should focus less on memorising syntax and more on understanding what happens behind the output.

AI can write SQL queries and generate reports, but knowing whether the query answers the right question is a different skill. Analysts need to understand joins, data grain, metric definitions, edge cases, and how small assumptions can change the results.

I’d prioritise three things: analytical thinking, data validation, and business context. Learn to investigate unexpected changes, identify misleading patterns, and connect findings to actual business decisions. SQL and reporting still matter, but they should be learned as tools for solving problems, not as the end goal.

The ability to use AI effectively is becoming part of the job, too. That means knowing how to give it the right context, review its output, and catch mistakes rather than accepting the first answer.

Ultimately, AI can reduce the effort required to produce an analysis, but it doesn’t remove the responsibility to make sure that analysis is correct and useful.

I’d focus on learning how to think through a data problem rather than just how to produce the output.

AI can generate SQL queries and build reports, but someone still needs to know whether the query answers the right question, whether the joins and aggregations are correct, and whether the results make business sense. Without that foundation, it’s easy to accept an answer that looks convincing but is wrong.

For someone starting out, I’d prioritize three things:

1. SQL and data fundamentals: Understand joins, aggregations, window functions, data quality, and how tables relate to one another.

2. Business problem-solving: Learn to translate vague business questions into measurable metrics and identify what the numbers actually mean.

3. Validation and communication: Check AI-generated outputs, investigate unexpected results, and explain findings clearly to people who need to make decisions.

Python and visualization are valuable too, but I wouldn’t try to learn every tool at once. I’d rather see a junior analyst solve one real business problem end to end than build ten dashboards without understanding the decisions behind them.

The advantage will go to analysts who can use AI to work faster while still exercising sound judgment over the results.