What AI does to data analysis work

AI tools have significantly automated the routine parts of data analysis. Natural language query interfaces (in tools like Microsoft Copilot in Excel, Tableau Pulse, Google Looker Studio AI, and various others) allow non-technical users to ask questions of their data in plain English and receive visualisations and summaries without SQL or Python. Automated insight generation tools identify anomalies, trends, and correlations in datasets and surface them proactively, without waiting for an analyst to ask the right question. Large language models can write SQL queries, explain data discrepancies, and produce first-draft analysis narratives with minimal human input.

The tasks most automated are: building standard dashboards and reports, writing routine SQL queries to extract defined data, producing recurring analysis that follows a known template, and explaining what happened in a data set. These tasks were the bread and butter of junior data analyst roles and they are significantly disrupted.

What data analysts still do that AI cannot

The parts of data analysis that remain distinctively human are: formulating the right questions in the first place (business question design), validating whether AI-generated analysis is correct and meaningful, interpreting results in the full business context, identifying why something is happening rather than just what is happening, designing new analysis approaches for novel business problems, and communicating insights to stakeholders in a way that drives action.

The most AI-resistant data analyst is one who operates less as a query builder and more as a business analyst who uses data: someone who understands the business deeply, knows what questions matter, can identify when an AI-generated insight is misleading because of a data quality issue or a confounding factor, and can translate analysis into decisions that stakeholders actually implement.

How data analysts should adapt

Develop strong business domain knowledge alongside technical skills: the analyst who understands the business as well as the data is irreplaceable in a way that the pure SQL writer is not. Build proficiency with AI data tools rather than competing with them: knowing how to use Copilot, Claude, or similar tools to produce analysis faster and at higher quality is itself a skill that differentiates you. Develop the ability to design and run experiments (A/B tests, causal inference), which requires human judgment about experimental design, ethics, and interpretation. And develop the communication and stakeholder skills that move analysis from a report to a decision.

Get real-time help in your next interview
Live Interview Help listens to your interview and surfaces personalised answers in real time. Free 20-minute trial on Google Meet, Teams, and Zoom.
Install Free on Chrome

Frequently asked questions

Is data science more or less at risk than data analysis?
Data science roles that focus primarily on building predictive models using established techniques (logistic regression, gradient boosting, standard neural network architectures on well-defined problems) are becoming more automatable as AutoML tools improve. Data scientists who do research-level machine learning, develop novel model architectures, work on AI safety and alignment, and lead the definition of what problems are worth solving with ML are substantially less at risk. The distinction is between applying known methods (increasingly automated) and advancing the field (remains human).