Framing the right question
84Turning a vague business or scientific problem into a specific, testable question requires organizational and domain context an automated tool has no access to on its own.
Data science sits at an unusual point in the AI story: many of the tools reshaping the profession — automated machine learning, natural-language-to-SQL translation, AI-assisted charting — were themselves largely built by clinical data scientists and machine-learning researchers, now aimed back at parts of their own daily work.
That exposure is real and uneven. The routine, well-specified end of the job — pulling data, running a standard model, producing a first-draft chart — is already comparably fast for AI tools to do. The end that survives is less about writing code and more about deciding what question is worth asking and whether an answer can be trusted.
Moderate
Share of the work a machine could do
A meaningful share of routine data-science work — SQL querying, exploratory charting, fitting a standard model with AutoML, writing boilerplate pipeline code — is already automatable or AI-assisted today, and that share is growing quickly. What resists automation is framing an ambiguous business problem as a testable question, judging whether a result is real or a statistical artifact, and being accountable when a model's real-world decision turns out to be wrong.
Scored from the tasks, not the job title. Lower is safer.
Jobs AI cannot take →What machines cannot take
Judging whether a pattern is real
80Telling a genuine effect apart from noise, a data-leakage artifact or a lucky train-test split is exactly the skeptical judgment automated pipelines are worst at applying to their own output.
Accountability for a real-world decision
88Someone has to answer for a model that denies a loan, flags a patient or misprices a product — a responsibility that cannot transfer to the tool that produced the number.
Persuading skeptical stakeholders
68Getting a counterintuitive or unwelcome finding actually acted on depends on trust, credibility and negotiation with people, not just a correct analysis.
Designing a genuinely new experiment
74Choosing the right natural experiment, control group or causal-inference strategy for a new question is closer to creative judgment than to pattern-matching over past analyses.
What they already take
Routine SQL queries and data pulls
75Natural-language-to-SQL tools now handle a large share of straightforward data-retrieval requests that once required a clinical data scientist's direct involvement.
First-draft exploratory charts and summaries
65Automated exploratory-analysis tools can generate distribution plots, correlation summaries and basic charts from a raw dataset in seconds.
Fitting and tuning standard models
68AutoML platforms can select, train and tune conventional models like gradient boosting or logistic regression largely unattended, for well-defined, well-labeled problems.
Boilerplate pipeline and ETL code
60Wiring together a standard data-transformation or loading script is now often faster to generate with an AI coding assistant and review than to write from scratch.
How the work is changing
From building models to directing and checking them
A growing share of the job is specifying what an automated pipeline should try and verifying its output, rather than writing every query and model by hand.
The generalist title splits into specialties
Analytics engineering, machine learning engineering and product analytics have increasingly become separate job titles, each absorbing a slice of what a single 'clinical data scientist' was expected to cover a decade ago.
Causal inference and experimentation gain status
As pattern-finding in existing data becomes easier to automate, the harder skill of designing a trustworthy experiment or causal analysis is becoming more valued, not less.
The 'citizen clinical data scientist' spreads basic analysis
No-code and AutoML tools let non-specialists run their own basic analyses, pushing dedicated clinical data scientists toward the harder, more ambiguous problems those tools cannot handle.
New jobs branching off
Analytics engineer
Builds and maintains the data-transformation pipelines — often using tools like dbt — that feed dashboards and models, sitting between data engineering and analysis.
Machine learning engineer
Takes a model out of a research notebook and makes it run reliably, monitored and at scale in a production system.
Data product manager
Owns the roadmap for a data or AI-driven feature, translating business goals into what a data team should actually build.
Responsible AI / model-risk analyst
Audits models and the data behind them for bias, fairness and regulatory compliance before they are allowed to ship.
Outlook
Demand for people who can turn data into a trustworthy decision is very likely to stay strong through the next decade, but the entry-level version of the job is already changing: fewer roles will exist purely to write routine queries and first-draft charts by hand.
The clearest long-run advantage sits with clinical data scientists who can frame a genuinely useful question, spot when an AI-generated result is quietly wrong, and take responsibility for a recommendation in front of skeptical stakeholders — the same judgment the field has needed since Fisher's crop trials, now applied to a workflow that increasingly drafts its own first pass.
Similar professions
Closest neighbours on the six-score profile — not the same field only.
🔐
Cybersecurity Specialist
Protects systems, data and people by finding, preventing and responding to digital attacks.
AI-resistant 63 🌬️Renewable Energy Engineer
Designs, builds and improves wind, solar, storage and grid systems that turn renewable resources into dependable electricity.
AI-resistant 72 🦾Robotics Engineer
Designs the machines that sense, decide and act in the physical world, where the hard problem was never intelligence but the world itself.
AI-resistant 65 🌿Sustainability Manager
Leads the strategy, measurement and reporting that helps organizations reduce environmental and social harm while meeting business obligations.
AI-resistant 66 🔌Semiconductor Engineer
Designs and fabricates the transistors inside every computer, phone and weapon, using machines precise enough that only a few factories on Earth can run them.
AI-resistant 60 📦MLOps Engineer
Builds the systems that train, deploy, monitor and govern machine-learning models in production.
AI-resistant 54Continue exploring
Keep exploring
More in Science & Research
⚛️
Physicist
Derives and tests the mathematical laws governing matter, energy, space and time, from a lone chalkboard to a 3,000-author particle-collider paper.
AI-resistant 65 🧬Biologist
The scientist who studies life itself, from Linnaeus naming species by hand to editing genomes with CRISPR, still testing every idea against a living organism.
AI-resistant 64 🧪Chemist
The scientist who makes and measures matter itself — from Tapputi's Babylonian perfume still to today's robot laboratories, still the one who decides what the spectrum means.
AI-resistant 70 📊Data Scientist
Finds patterns and builds predictive models from data — a 2008 job title built on three centuries of counting, testing and visualizing evidence.
AI-resistant 38 🔭Astronomer
The scientist who measures the universe, from Babylonian clay tablets to space telescopes, still deciding which flicker in the data is a discovery.
AI-resistant 70 🧮Mathematician
From Babylonian scribes to Fields medalists and AI-assisted proof: the profession that turns hard questions into permanent certainty, one theorem at a time.
AI-resistant 70 🧿Quantum Engineer
Builds, measures and controls devices that exploit quantum states for computing, sensing, communication and materials research.
AI-resistant 78