How AI is changing computer & math work.
TaskExposed currently tracks 15 occupations in the computer & math family, representing approximately 4.3M workers. The group has an average AI exposure score of 55% and an average resilience score of 67.
The most exposed roles usually contain repeatable, text-heavy, data-heavy, or process-driven tasks. The most resilient roles usually depend on judgment, physical presence, trust, real-time decision-making, or cross-functional human coordination.
Use this page as a career map: compare risk levels, explore lower-exposure adjacent paths, and open individual profession reports for task-level detail.
Most AI-exposed computer & math careers
Roles with the highest task-level exposure scores.
Most resilient computer & math careers
Roles with the strongest human resilience scores.
High, moderate, and low exposure roles.
Explore every computer & math profession.
Web developers face high task-level exposure — much of the implementation work that used to take days is now generated in minutes. Value is shifting toward architecture, UX judgment, and client communication.
Data analysts face high exposure in query writing, report generation, and data cleaning — but strong human advantage remains in hypothesis formation, business translation, and stakeholder storytelling.
QA engineers face high exposure in test generation and automation, where AI now writes test suites from specs. But exploratory testing, test strategy, and the judgment to prioritise what breaks badly still require humans.
Software engineers face heavy task-level exposure to language models, but maintain strong human-critical work in systems design, debugging ambiguous environments, and cross-team negotiation.
Data engineers face growing exposure in pipeline generation and schema work, but the architectural thinking, data contract ownership, and cross-system integration judgment remain strongly human.
Data scientists face high AI assistance in modelling, coding, and analysis, but the durable value is deciding what questions matter, validating messy real-world data, and translating uncertainty into decisions.
DevOps engineers see strong AI assistance in configuration, scripting, and documentation, while real-time incident response, platform strategy, and on-call judgment under pressure stay firmly human.
ML engineers face a paradox: AI accelerates their tooling and code generation, but the research intuition, model evaluation, and production reliability work that defines the role requires deep human judgment.
Platform engineers automate other engineers' toil — and AI now automates theirs: pipelines and IaC generate quickly, while platform strategy and organizational adoption stay human.
Developer advocates see tutorials, sample code, and docs generate on demand, while community trust, live talks, and authentic developer empathy remain the human differentiators.
Cybersecurity analysts benefit from AI in threat detection and log analysis, but adversarial reasoning, incident response under pressure, and attacker psychology remain distinctly human capabilities.
Solutions architects see diagrams, proposals, and reference configs generate quickly, while trade-off judgment across messy enterprise constraints and client trust stay human.
SREs are more resilient than most engineers: runbooks and boilerplate automate, but incident command under pressure, systems intuition, and accountability for uptime stay human.
Security engineers stay resilient because the adversary adapts: AI automates scanning and reports, but threat modeling, incident response, and out-thinking attackers remain human work.
Engineering managers stay resilient as AI absorbs status reports and planning admin: coaching people, making judgment calls, and owning delivery remain management's human core.