Will AI Steal My Job? · Role analysis

Data Scientist

O*NET 15-2051.00 ESCO: Data scientists

Changing

Data scientists extract insights from large, complex datasets using statistical analysis, machine learning, and data visualisation. They design experiments, build predictive models, and communicate findings to technical and non-technical audiences — translating data into decisions that drive product, operations, and strategy.

Section 01

Task Map

Task	AI impact	Why
Write data cleaning and preprocessing code	🔴 High exposure	AI coding tools generate data cleaning pipelines and EDA code rapidly. This was previously one of the most time-consuming parts of data science work and is now significantly accelerated.
Select and apply machine learning models	🟡 Changing	AutoML tools can run model selection automatically, but the judgment about which approach fits the problem, the data quality, and the business context still requires expertise.
Design experiments and A/B tests	🟡 Changing	AI can assist with experimental design, but defining the right question to test — and ensuring the experiment actually tests what it needs to — requires product and statistical thinking.
Build data pipelines and ETL processes	🔴 High exposure	AI tools accelerate pipeline development significantly. Standard ETL patterns are well-understood by code generation tools. Data engineering work is heavily AI-assisted.
Evaluate model performance and bias	🟡 Changing	AI can calculate metrics automatically, but understanding whether a model's errors matter in the real world — and whether it perpetuates unfair bias — requires domain knowledge and ethical judgment.
Communicate insights to non-technical stakeholders	🟢 Safe	Translating statistical findings into business decisions, presenting uncertainty honestly, and persuading sceptical stakeholders requires communication and political intelligence that AI tools cannot provide.
Frame and define the right business problem	🟢 Safe	Deciding what question is actually worth answering — and whether data can answer it — requires business acumen, curiosity, and stakeholder engagement that no automated tool can replicate.
Research and apply novel ML methods	🟡 Changing	AI can summarise research papers, but adapting novel methods to a specific problem, understanding their assumptions, and diagnosing when they fail requires deep technical expertise.

Section 02

What Stays Human

Problem framing and hypothesis generation The most valuable data scientists don't wait for a problem to be handed to them — they look at a business and identify what's most worth understanding. This intellectual curiosity and business judgment is irreducibly human.
Ethical oversight of AI and ML systems As organisations deploy ML models that affect real people's lives — credit decisions, hiring, healthcare — the data scientist who can evaluate fairness, identify bias, and advocate for responsible use is performing essential oversight work with significant social stakes.
Stakeholder influence and scientific communication Getting an organisation to actually act on data findings — changing strategy, allocating budget, overcoming resistance from people who prefer their intuitions — requires persuasion, credibility, and political navigation that no algorithm can manage.

Section 03

What to Do Next

Develop deep expertise in AI/ML systems design and evaluation. Data scientists who understand how to build, evaluate, and maintain LLM-based systems, RAG pipelines, and modern ML infrastructure are at the frontier of demand. This is the specialism with the strongest job market and fastest-growing salary premium.
Build domain expertise alongside technical skills. A data scientist who deeply understands healthcare, finance, climate science, or another specific domain is significantly more valuable than a generalist. The combination of strong ML skills and deep domain knowledge is extremely hard to replicate with AI tools or generalist data scientists.
Invest in communication and influence skills. The data scientist who can run a board presentation, navigate stakeholder politics, and champion data-informed decisions at a senior level is providing leadership value that technical skills alone don't deliver. Practise explaining your work to non-technical audiences and building internal data culture.

Sources: O*NET Online (onetonline.org) · ESCO (esco.ec.europa.eu) · All task data cross-referenced against O*NET occupation profiles. This analysis uses task-level exposure, not occupation-level prediction.