Datasets:
slug string | title string | category string | score int64 | risk_level string | salary_low_usd int64 | salary_high_usd int64 | salary_trend string | tasks_at_risk_count int64 | tasks_growing_count int64 | defense_skill_1 string | defense_skill_2 string | defense_skill_3 string | insight string | score_q1_2025 float64 | score_q3_2025 float64 | score_q1_2026 float64 | score_q2_2026 float64 | score_q3_2026 int64 | soc_code string | tier int64 | sub_task_automation int64 | sub_tool_maturity int64 | sub_adoption int64 | sub_agentic_exposure int64 | agentic_risk string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
it-support-specialist | IT Support Specialist | Technology | 66 | High | 45,000 | 75,000 | declining | 5 | 3 | Administering AI Helpdesk Agents | Cybersecurity Fundamentals | Cloud Platform & Identity Management | AI helpdesk agents now resolve most tier-1 tickets -- resets, installs, routing -- without a human touch, and that volume was this role's core. The specialists still hired are the ones running and tuning those agents, or moving into security first-response and cloud identity work. | 50 | 55 | 60 | 65 | 66 | 15-1232 | 1 | 70 | 80 | 68 | 62 | High |
systems-administrator | Systems Administrator | Technology | 51 | High | 65,000 | 115,000 | declining | 5 | 3 | Infrastructure as Code (Terraform, Pulumi) | Cloud Architecture (AWS/Azure/GCP) | Security & Compliance Automation | Cloud automation and AI-ops tooling now run provisioning, patching, and backup verification without a human in the loop. The clearest path forward is cloud architecture, security hardening, and automation design rather than day-to-day server administration. | 42 | 46 | 50 | 55 | 51 | 15-1244 | 1 | 56 | 72 | 56 | 54 | High |
business-intelligence-analyst | Business Intelligence Analyst | Technology | 49 | Moderate | 70,000 | 115,000 | mixed | 4 | 3 | Directing AI BI Copilots (Power BI, Looker) | Metric Definition & Stakeholder Alignment | Anomaly Root-Cause Investigation | AI BI copilots now build and refresh most recurring dashboards without an analyst touching the query. The role's future is in deciding what should be measured in the first place and chasing down why a metric moved, not in maintaining the pipeline that reports it. | null | null | null | null | 49 | 15-1211 | 2 | 52 | 76 | 64 | 56 | High |
network-support-specialist | Network Support Specialist | Technology | 49 | Moderate | 55,000 | 90,000 | declining | 4 | 3 | Administering AI-Driven NOC Triage Tools | Network Security Fundamentals | Multi-Site Outage Diagnosis | AI triage tools now resolve most known-fix network tickets before a specialist sees them. The path forward is running that triage layer and handling the multi-site, security-flavored incidents it can't close on its own. | null | null | null | null | 49 | 15-1231 | 2 | 56 | 68 | 52 | 54 | High |
qa-test-engineer | QA/Test Engineer | Technology | 49 | High | 75,000 | 130,000 | mixed | 5 | 3 | Agentic Test Generation & Self-Healing Suites | AI/ML Model & Output Validation | Security & Resilience Testing | Agentic test tools now write, run, and self-heal most regression suites, which is why routine QA is now a substantially agent-owned workflow. Engineers who validate AI system outputs and own security/performance test strategy are the ones still growing in value. | 40 | 45 | 50 | 55 | 49 | 15-1253 | 1 | 56 | 66 | 52 | 52 | High |
data-analyst | Data Analyst | Technology | 47 | Moderate | 65,000 | 105,000 | mixed | 4 | 3 | Directing AI BI Copilots (Power BI, Tableau, Looker) | Business Problem Framing & Insight Storytelling | SQL + AI-Assisted Data Modeling | AI BI copilots now write the queries and refresh the dashboards that used to be most of this job. Analysts who frame the business question and tell the story behind the numbers are pulling away from analysts who just build reports. | null | null | null | null | 47 | 15-2051 | 1 | 50 | 74 | 62 | 54 | High |
database-administrator | Database Administrator | Technology | 43 | Moderate | 80,000 | 140,000 | stable | 4 | 3 | Cloud Database Architecture (AWS/Azure/GCP) | Vector Database & RAG Data Layer Management | Data Security & Compliance Automation | Autonomous cloud databases now handle tuning, patching, and backup verification without a human in the loop, which is why routine DBA work is the most automated slice of the Technology category. The growth path is data architecture for AI workloads, vector data layers, and security -- not day-to-day administration. | 38 | 42 | 46 | 50 | 43 | 15-1242 | 1 | 48 | 72 | 55 | 46 | Moderate |
web-developer | Web Developer | Technology | 40 | Moderate | 70,000 | 140,000 | stable | 4 | 3 | AI-Powered Development Tools (v0, Cursor, Copilot) | Full-Stack Architecture | AI/Agent Integration in Web Apps | AI page-builders now ship standard sites and components with little human editing. Developers who own complex application logic, accessibility, and AI feature integration are the ones still commanding a premium. | 30 | 35 | 40 | 45 | 40 | 15-1254 | 1 | 44 | 76 | 54 | 44 | Moderate |
devops-engineer | DevOps Engineer | Technology | 35 | Moderate | 115,000 | 185,000 | rising | 4 | 3 | AIOps & Agentic Pipeline Supervision | Cloud Security & Zero Trust Architecture | Platform Engineering Leadership | AI now writes and maintains most CI/CD configuration and IaC boilerplate directly. Engineers who move into platform strategy, incident command, and security architecture are the ones whose value keeps climbing as the routine plumbing disappears. | 22 | 25 | 28 | 30 | 35 | 15-1252 | 1 | 38 | 68 | 54 | 50 | Moderate |
data-engineer | Data Engineer | Technology | 34 | Low | 115,000 | 190,000 | rising | 4 | 3 | Agentic Pipeline Monitoring & Self-Healing ETL | Data Architecture for AI/RAG Workloads | Data Governance & Lineage Tooling | AI now writes and self-heals most routine pipeline code, from ETL scripts to data-quality checks. The lasting work is architecting data platforms for AI training and RAG workloads, where governance and lineage decisions still need a person accountable for them. | null | null | null | null | 34 | 15-1252 | 1 | 38 | 64 | 50 | 46 | Moderate |
mobile-developer | Mobile Developer | Technology | 33 | Low | 95,000 | 165,000 | stable | 4 | 3 | AI-Assisted Native Development (Swift, Kotlin Copilots) | Cross-Platform Architecture (Flutter, React Native, KMP) | Mobile Performance & Battery Optimization | AI now scaffolds standard screens and common features for both major mobile platforms directly from a spec. Developers who own cross-platform architecture and can debug real device fragmentation issues are the ones clients still need. | null | null | null | null | 33 | 15-1252 | 2 | 36 | 76 | 54 | 40 | Moderate |
site-reliability-engineer | Site Reliability Engineer | Technology | 31 | Low | 130,000 | 200,000 | rising | 4 | 3 | Agentic Incident Triage & Runbook Automation | AIOps & Observability Platforms | Reliability Architecture & Capacity Planning | Agents now run first-pass triage and execute known runbooks before a human is paged. The SREs still in demand are the ones who own reliability architecture and can run incident command when the runbook doesn't cover it. | null | null | null | null | 31 | 15-1252 | 2 | 34 | 68 | 52 | 48 | Moderate |
ux-designer | UX Designer | Technology | 30 | Moderate | 85,000 | 150,000 | stable | 4 | 3 | AI Interaction & Conversational Design | User Research Methods | Directing AI Design Tools (Figma AI, Galileo) | AI can generate mockups and component variations fast, but understanding real user needs and designing for AI-human interaction remain judgment calls a model can't make on its own. | 25 | 28 | 32 | 35 | 30 | 15-1255 | 1 | 36 | 62 | 46 | 34 | Low |
ai-engineer | AI Engineer | Technology | 29 | Low | 140,000 | 220,000 | rising | 4 | 3 | Multi-Agent Orchestration & Tool Use | RAG & Retrieval System Design | AI Safety, Guardrails & Eval Design | The tooling this role builds with is itself good enough to automate a meaningful share of its own scaffolding -- prompt templates, RAG wiring, eval harnesses. Engineers who own guardrail design and multi-agent orchestration for genuinely novel workflows are the ones the market is still short of. | null | null | null | null | 29 | 15-1252 | 1 | 32 | 58 | 50 | 46 | Moderate |
platform-engineer | Platform Engineer | Technology | 29 | Low | 120,000 | 195,000 | rising | 4 | 3 | Internal Developer Platform (IDP) Design | Agentic Golden-Path Template Maintenance | Platform Cost & Reliability Architecture | AI now maintains a lot of the golden-path templates and CI/CD scaffolding platform teams used to hand-tend. The differentiator is driving adoption across teams and making the standardization-versus-autonomy tradeoff correctly, which no template can decide. | null | null | null | null | 29 | 15-1252 | 2 | 32 | 66 | 46 | 42 | Moderate |
data-scientist | Data Scientist | Technology | 28 | Low | 120,000 | 190,000 | rising | 4 | 3 | Directing Agentic Data Pipelines & AutoML | LLM Fine-Tuning & Evaluation | AI Strategy & Business Translation | Agentic notebooks can now run an entire clean-model-report loop on structured data unattended, which pushes routine modeling toward a commodity task. Data scientists who frame the right problem, design the experiment, and translate results into a business decision remain the ones firms pay for. | 14 | 16 | 18 | 20 | 28 | 15-2051 | 1 | 32 | 58 | 48 | 40 | Moderate |
machine-learning-engineer | Machine Learning Engineer | Technology | 28 | Low | 130,000 | 210,000 | rising | 4 | 3 | Agentic MLOps Pipeline Supervision | Model Evaluation & Production Risk Assessment | LLM & Foundation Model Fine-Tuning | Training pipelines and hyperparameter sweeps are now largely agent-run, but production risk sign-off and novel failure-mode debugging still require an engineer who understands the model, not just the framework. The premium is shifting toward engineers who can evaluate and govern what agents ship. | null | null | null | null | 28 | 15-1252 | 1 | 31 | 64 | 52 | 44 | Moderate |
security-engineer | Security Engineer | Technology | 28 | Low | 115,000 | 185,000 | rising | 4 | 3 | Security Automation & Policy-as-Code | AI/LLM Application Security & Threat Modeling | Zero-Day Response & Exploit Analysis | Automated remediation now closes most known vulnerability classes without an engineer touching a ticket. The differentiator is security architecture for genuinely new systems -- including the AI applications the rest of the org is shipping. | null | null | null | null | 28 | 15-1212 | 2 | 32 | 60 | 50 | 42 | Moderate |
software-engineer | Software Engineer | Technology | 28 | Low | 110,000 | 200,000 | rising | 4 | 3 | Directing Agentic Coding Tools (Claude Code, Cursor, Copilot Workspace) | System Design & Architecture Ownership | Reviewing & Verifying Agent-Generated Code | AI coding agents now handle scaffolding, tests, and routine bug fixes end-to-end, but architecture, security tradeoffs, and cross-team judgment still gate the work that matters. The role is shifting from writing every line to directing and verifying what agents produce. | 18 | 20 | 23 | 25 | 28 | 15-1252 | 1 | 28 | 82 | 62 | 42 | Moderate |
cybersecurity-analyst | Cybersecurity Analyst | Technology | 26 | Low | 90,000 | 165,000 | rising | 4 | 3 | AI Security & Adversarial ML Defense | Agentic SOC Tool Supervision & Tuning | Cloud Security Architecture | AI SOC copilots now clear most tier-1 alert volume, but they also expand the attack surface faster than they close it. Analysts who supervise agentic triage and specialize in AI-native threats -- prompt injection, model poisoning -- are pulling ahead of generalists. | 15 | 17 | 18 | 20 | 26 | 15-1212 | 1 | 30 | 58 | 48 | 38 | Moderate |
game-developer | Game Developer | Technology | 26 | Low | 75,000 | 140,000 | mixed | 4 | 3 | AI-Assisted Engine Scripting (Unity, Unreal Copilots) | Procedural Content & Asset Generation Pipelines | Game Feel & Player-Experience Design | AI now writes standard gameplay systems code and can generate a fair share of routine assets, but the creative judgment behind what makes a game actually feel good to play is still squarely a human skill. | null | null | null | null | 26 | 15-1252 | 2 | 30 | 58 | 40 | 38 | Moderate |
network-engineer | Network Engineer | Technology | 24 | Low | 80,000 | 135,000 | stable | 4 | 3 | Network Automation (Ansible, Terraform for Networking) | SD-WAN & Hybrid-Cloud Connectivity | Network Security & Zero Trust | AI can now generate standard router and switch configurations from a template, but multi-vendor network design and diagnosing an unfamiliar topology under outage pressure still need an engineer who has seen the failure mode before. | null | null | null | null | 24 | 15-1241 | 2 | 28 | 54 | 42 | 36 | Moderate |
blockchain-developer | Blockchain Developer | Technology | 23 | Low | 100,000 | 170,000 | mixed | 4 | 3 | Smart Contract Security Auditing | AI-Assisted Contract Testing & Formal Verification | Cross-Chain Protocol Architecture | AI can now scaffold a standard smart contract quickly, which makes security auditing -- where a missed edge case costs real money on-chain -- the part of this role that is actually growing in value. | null | null | null | null | 23 | 15-1252 | 2 | 30 | 48 | 28 | 32 | Low |
cloud-architect | Cloud Architect | Technology | 23 | Low | 140,000 | 215,000 | rising | 4 | 3 | Multi-Cloud Architecture (AWS/Azure/GCP) | AI-Assisted IaC & Landing Zone Design | Cloud Cost & Resilience Governance | AI can draft a reference architecture and the IaC to match it, but choosing between vendors, owning the cost/resilience tradeoff, and getting a regulated workload through security review still take an accountable architect. | null | null | null | null | 23 | 15-1241 | 1 | 26 | 56 | 48 | 36 | Moderate |
database-architect | Database Architect | Technology | 22 | Low | 115,000 | 175,000 | stable | 4 | 3 | Enterprise Data Architecture Strategy | AI-Assisted Schema Design & Migration Tooling | Data Governance for AI/Analytics Workloads | AI can draft a reasonable schema for a well-understood problem, but the decisions that actually matter -- how data flows across systems, who governs it, and how it serves both transactional and AI workloads -- still need an accountable architect. | null | null | null | null | 22 | 15-1243 | 2 | 26 | 52 | 40 | 34 | Low |
solutions-architect | Solutions Architect | Technology | 20 | Low | 125,000 | 195,000 | stable | 4 | 3 | AI-Assisted Solution Diagramming & Documentation | Enterprise Requirements Discovery & Negotiation | Multi-Vendor Architecture Tradeoff Analysis | AI now handles the first draft of the diagram and the proposal document. What still requires a person is sitting across from a client, understanding what they actually need, and owning the tradeoff decision when the textbook answer doesn't fit their constraints. | null | null | null | null | 20 | 15-1211 | 2 | 24 | 50 | 42 | 30 | Low |
ai-research-scientist | AI Research Scientist | Technology | 17 | Low | 150,000 | 260,000 | rising | 4 | 3 | Agentic Experiment Pipelines & Self-Driving Research Loops | Frontier Model Evaluation & Interpretability | Technical Writing for Research Publication | Agentic experiment loops are starting to run their own ablation studies and literature synthesis, but the research direction and theoretical insight that make a result matter still come from a scientist, not the loop. | null | null | null | null | 17 | 15-1221 | 2 | 20 | 54 | 34 | 32 | Low |
embedded-systems-engineer | Embedded Systems Engineer | Technology | 15 | Low | 90,000 | 150,000 | stable | 4 | 3 | Real-Time & Safety-Critical Firmware Design | AI-Assisted Low-Level Coding (embedded C/C++ copilots) | Hardware-in-the-Loop Testing & Debugging | This is one of the least agent-exposed engineering roles today: AI tooling for low-level, hardware-coupled firmware still lags mainstream software copilots badly, and physical hardware-in-the-loop debugging keeps a person firmly in the loop. | null | null | null | null | 15 | 17-2061 | 2 | 20 | 42 | 30 | 22 | Low |
it-manager | IT Manager | Technology | 14 | Low | 105,000 | 165,000 | stable | 4 | 3 | AI-Augmented Reporting & Portfolio Dashboards | AI Tooling Strategy & Vendor Evaluation | Team Leadership Through an AI Transition | AI now drafts the status reports and vendor summaries that used to eat a manager's week, but leading a team and deciding where to invest a shrinking headcount budget is still an entirely human call. | null | null | null | null | 14 | 11-3021 | 2 | 18 | 38 | 44 | 16 | Low |
computer-hardware-engineer | Computer Hardware Engineer | Technology | 13 | Low | 105,000 | 175,000 | stable | 4 | 3 | AI-Assisted EDA & Simulation Tools | Hardware Prototyping & Lab Validation | Power/Thermal/Manufacturing Tradeoff Analysis | AI-assisted simulation and EDA tools can speed up routine schematic work, but hardware design still lives or dies in physical prototyping and lab validation that no model can substitute for. | null | null | null | null | 13 | 17-2061 | 2 | 18 | 40 | 26 | 16 | Low |
tax-preparer | Tax Preparer | Business & Finance | 69 | High | 40,000 | 75,000 | declining | 5 | 3 | Advanced tax planning and strategy | AI tax software proficiency | Business advisory services | AI tax software now files the majority of standard individual returns with minimal human input, and the segment keeps shrinking. Preparers who specialize in multi-state, international, or business tax planning are the ones with a growing practice left. | 58 | 63 | 68 | 72 | 69 | 13-2082 | 1 | 70 | 88 | 72 | 62 | High |
insurance-underwriter | Insurance Underwriter | Business & Finance | 62 | High | 55,000 | 110,000 | mixed | 5 | 3 | AI risk modeling and insurtech platforms | Specialty lines underwriting | Data analytics for risk assessment | AI now underwrites the overwhelming majority of standard personal lines with no human touch. The career path forward is complex commercial risk, emerging categories like cyber and AI liability, and broker relationships that carriers still need a person to run. | 52 | 58 | 63 | 68 | 62 | 13-2053 | 1 | 62 | 88 | 72 | 62 | High |
bank-teller | Bank Teller | Business & Finance | 58 | High | 30,000 | 40,000 | declining | 4 | 3 | Complex product cross-selling and financial guidance | Dispute resolution and exception handling | Community banking and relationship building | Branch teller headcount has been shrinking for over a decade as mobile and ATM/ITM channels absorb routine transactions, and AI-driven self-service accelerates that shift further. | null | null | null | null | 58 | 43-3071 | 1 | 64 | 75 | 60 | 52 | High |
loan-officer | Loan Officer | Business & Finance | 57 | High | 45,000 | 120,000 | mixed | 5 | 3 | Commercial lending expertise | Financial advisory and planning | Fintech and AI lending platforms | Online lenders now originate most standard consumer loans with essentially no loan officer involvement. Officers who build deep client relationships and master complex commercial or construction lending remain the ones borrowers and banks still need. | 50 | 55 | 60 | 65 | 57 | 13-2072 | 1 | 58 | 85 | 68 | 58 | High |
accountant | Accountant | Business & Finance | 56 | High | 55,000 | 105,000 | stable | 5 | 3 | AI accounting platforms (Vic.ai, Botkeeper, QBO AI) | Tax strategy and advisory | Forensic accounting and fraud analytics | Routine bookkeeping, reconciliation, and standard tax prep are now largely agent-run rather than merely assisted. Accountants who move up into advisory, complex multi-entity tax strategy, or forensic work keep pricing power; those who stay in pure transaction processing compete directly with software. | 48 | 53 | 58 | 62 | 56 | 13-2011 | 1 | 55 | 90 | 72 | 62 | High |
bookkeeper | Bookkeeper | Business & Finance | 56 | High | 38,000 | 55,000 | declining | 5 | 3 | AI bookkeeping platform administration (Botkeeper, QBO AI) | Client advisory and cash-flow coaching | Books cleanup and audit-readiness | Bookkeeping is one of the most fully agent-automatable roles in finance today: categorization, reconciliation, and reporting now run with minimal human input. The path forward is overseeing the AI, handling exceptions, and moving into client-facing advisory work the software can't do. | null | null | null | null | 56 | 43-3031 | 1 | 58 | 85 | 65 | 58 | High |
auditor | Auditor | Business & Finance | 53 | High | 60,000 | 115,000 | stable | 5 | 3 | AI and algorithm auditing | Data analytics for audit (ACL, IDEA, continuous audit tools) | IT audit and cybersecurity risk | Full-population continuous auditing has moved from pilot to default at large firms, replacing sample-based testing outright. Auditors who specialize in judgment-heavy risk assessment, AI-system auditing, or fraud analytics are the ones clients still pay a premium for. | 45 | 50 | 55 | 60 | 53 | 13-2011 | 1 | 54 | 85 | 65 | 62 | High |
financial-analyst | Financial Analyst | Business & Finance | 50 | High | 65,000 | 120,000 | stable | 5 | 3 | AI-powered financial modeling (Copilot, AlphaSense) | Strategic finance and executive advisory | Data science for finance | AI now builds the model and pulls the data in minutes; the analysts who keep their seat are the ones translating the output into a judgment call an executive can act on under real uncertainty. | 45 | 50 | 54 | 58 | 50 | 13-2051 | 1 | 50 | 88 | 68 | 55 | High |
budget-analyst | Budget Analyst | Business & Finance | 49 | High | 60,000 | 100,000 | stable | 4 | 3 | AI-assisted FP&A platforms (Planful, Vena, Datarails) | Strategic finance and resource-allocation advisory | Data visualization and executive storytelling | AI now closes the variance-report-to-forecast loop with minimal input, but budget analysts who can advise leadership on trade-offs and broker cross-department negotiation still set the agenda. | 40 | 45 | 50 | 55 | 49 | 13-2031 | 1 | 50 | 85 | 65 | 52 | High |
claims-adjuster | Claims Adjuster | Business & Finance | 48 | Moderate | 48,000 | 85,000 | mixed | 4 | 3 | Complex and catastrophe claims investigation | AI touchless-claims platform oversight | Fraud detection and escalation analysis | Touchless claims already settle most straightforward auto and property losses without a human adjuster. The remaining career path is in complex, contested, and catastrophic claims where judgment and negotiation still decide the outcome. | null | null | null | null | 48 | 13-1031 | 2 | 52 | 70 | 58 | 58 | High |
business-analyst | Business Analyst | Business & Finance | 43 | Moderate | 70,000 | 120,000 | stable | 4 | 3 | AI process automation and agent orchestration | Product management fundamentals | AI-enhanced business intelligence | AI now drafts requirements and pulls reports directly from meetings and data, but reading organizational politics and redesigning workflows around AI remain distinctly human work. | 35 | 40 | 44 | 48 | 43 | 13-1111 | 1 | 46 | 78 | 58 | 45 | Moderate |
credit-analyst | Credit Analyst | Business & Finance | 41 | Moderate | 50,000 | 95,000 | stable | 3 | 3 | Complex commercial credit and covenant analysis | AI-driven credit scoring platform oversight | Portfolio risk trend analysis | Algorithmic scoring already handles most consumer and small-business credit decisions. Analysts who move into complex commercial deals and portfolio-level risk strategy keep the work software can't yet do. | null | null | null | null | 41 | 13-2041 | 2 | 44 | 75 | 58 | 45 | Moderate |
compliance-officer | Compliance Officer | Business & Finance | 40 | Moderate | 70,000 | 135,000 | rising | 4 | 3 | AI governance and ethics frameworks | RegTech and compliance automation | Risk management certification | AI regulation is exploding worldwide, and monitoring/checklist work is automating fast, but every one of those new rules needs a human who can interpret it and build the internal case for compliance. AI-governance specialists are the fastest-growing niche in the field. | 38 | 40 | 42 | 45 | 40 | 13-1041 | 1 | 45 | 70 | 58 | 42 | Moderate |
fraud-investigator | Fraud Investigator | Business & Finance | 33 | Low | 55,000 | 100,000 | rising | 3 | 3 | AI/AML transaction-monitoring platform expertise | Investigative interviewing and case-building | Emerging fraud typology analysis (deepfake, synthetic ID) | AI now catches the overwhelming majority of anomalies automatically, which is actually growing the investigator role: someone still has to confirm intent, interview the subject, and build a case that holds up, and fraud volume keeps climbing. | null | null | null | null | 33 | 13-2099 | 2 | 36 | 68 | 50 | 48 | Moderate |
financial-advisor | Financial Advisor | Business & Finance | 29 | Low | 55,000 | 140,000 | stable | 3 | 3 | Comprehensive financial and estate planning | AI portfolio and robo-advisory tools | Behavioral coaching and client trust-building | Robo-advisors have commoditized basic portfolio management, but the advisors who keep and grow books of business are the ones clients call during a market crash or a life event, not the ones who just rebalance a portfolio. | null | null | null | null | 29 | 13-2052 | 1 | 32 | 75 | 50 | 35 | Low |
real-estate-appraiser | Real Estate Appraiser | Business & Finance | 29 | Low | 45,000 | 90,000 | declining | 3 | 3 | Complex and unique-property valuation judgment | AVM oversight and exception review | Litigation and dispute appraisal expertise | Appraisal-waiver programs and automated valuation models are eroding standard-appraisal volume steadily, but licensed sign-off on complex or disputed properties still requires a certified human on site. | null | null | null | null | 29 | 13-2023 | 2 | 34 | 65 | 45 | 30 | Low |
management-consultant | Management Consultant | Business & Finance | 28 | Low | 90,000 | 200,000 | rising | 4 | 3 | AI strategy consulting | AI-powered research and analysis | Change management for AI adoption | AI now handles research and deck-building faster than any analyst team, but clients still pay for judgment, relationships, and someone accountable for the outcome. Consultants who specialize in AI transformation itself are in peak demand. | 25 | 28 | 32 | 35 | 28 | 13-1111 | 1 | 30 | 70 | 52 | 40 | Moderate |
credit-counselor | Credit Counselor | Business & Finance | 25 | Low | 40,000 | 65,000 | stable | 3 | 3 | Empathetic financial counseling and coaching | Creditor negotiation | Debt-relief and bankruptcy case guidance | Debt-plan math is easy to automate, but people in financial distress come to a credit counselor for a person who will negotiate on their behalf and help them change behavior, not a calculator. | null | null | null | null | 25 | 13-2071 | 2 | 32 | 52 | 38 | 28 | Low |
real-estate-agent | Real Estate Agent | Business & Finance | 25 | Low | 35,000 | 95,000 | mixed | 4 | 3 | AI-powered listing and comp-generation tools | Negotiation and local market expertise | Client relationship management (CRM/AI follow-up) | AI now writes the listing and pulls the comps, but buying and selling a home is still an emotional, high-stakes, in-person negotiation. Agents who lean into local expertise and relationship management outcompete discount and AI-only listing models. | null | null | null | null | 25 | 41-9022 | 1 | 28 | 68 | 48 | 30 | Low |
financial-risk-specialist | Financial Risk Specialist | Business & Finance | 24 | Low | 70,000 | 135,000 | rising | 3 | 3 | AI/ML model risk validation and governance | Quantitative stress-testing platforms | Regulatory capital strategy | Regulators now treat AI/ML model risk as a top priority, which is turning financial risk specialists into the certified check on every automated pricing and credit model a bank deploys. | null | null | null | null | 24 | 13-2054 | 2 | 28 | 58 | 45 | 32 | Low |
financial-examiner | Financial Examiner | Business & Finance | 22 | Low | 60,000 | 115,000 | stable | 3 | 3 | Regulatory enforcement judgment and interpretation | AI-driven exam data analytics | Emerging risk examination (crypto, AI lending) | Examination data review is getting faster with AI, but the statutory finding and enforcement decision stay with a certified examiner, and new risk categories like crypto and AI-driven lending are expanding the job, not shrinking it. | null | null | null | null | 22 | 13-2061 | 2 | 28 | 48 | 35 | 25 | Low |
actuary | Actuary | Business & Finance | 19 | Low | 75,000 | 150,000 | stable | 3 | 3 | AI/ML model validation and governance | Predictive modeling for novel risk (climate, cyber) | Actuarial credentialing (FSA/ASA/CAS) | Actuarial work is professionally licensed and carries personal liability for the certified opinion, which keeps AI in an assistive role even as modeling tools improve quickly. Actuaries who also validate AI/ML risk models are becoming the bridge between data science and regulatory-grade certification. | null | null | null | null | 19 | 15-2011 | 1 | 24 | 55 | 38 | 22 | Low |
purchasing-manager | Purchasing Manager | Business & Finance | 19 | Low | 75,000 | 130,000 | stable | 3 | 3 | Supplier negotiation and sourcing strategy | AI spend-analytics platforms | Supply chain risk management | AI now surfaces spend patterns and reorder points automatically, but negotiating supplier terms and managing a sourcing team under real supply-chain risk stays a human judgment call. | null | null | null | null | 19 | 11-3061 | 2 | 24 | 50 | 42 | 25 | Low |
economist | Economist | Business & Finance | 18 | Low | 65,000 | 150,000 | stable | 3 | 3 | Original research design and causal inference | AI-assisted data analysis and modeling | Policy communication and stakeholder briefing | AI speeds up the literature review and the first regression, but the value of an economist is still in the research question and the judgment call about what the numbers actually mean for policy. | null | null | null | null | 18 | 19-3011 | 2 | 24 | 45 | 32 | 22 | Low |
investment-banker | Investment Banker | Business & Finance | 15 | Low | 85,000 | 275,000 | mixed | 3 | 3 | Deal origination and client relationship management | AI-assisted financial modeling and comps | Cross-border and regulatory deal structuring | AI now drafts the pitch book in a fraction of the time, but the job has always really been about winning and closing the mandate, which is still a trust-and-negotiation business. | null | null | null | null | 15 | 41-3031 | 2 | 20 | 48 | 32 | 20 | Low |
telemarketer | Telemarketer | Marketing & Sales | 77 | Very High | 27,000 | 45,000 | declining | 5 | 3 | AI Voice-Agent Campaign Management | Regulated-Sales Compliance (TCPA, disclosures) | Complex Objection Handling & Escalation | Scripted cold calling is now close to a pure agent job: an AI voice agent dials, pitches, and logs outcomes at a fraction of the cost. The remaining human seats sit in regulated products and campaigns where a compliance disclosure or an angry prospect needs a person on the line, and even those seats are shrinking as ca... | null | null | null | null | 77 | 41-9041 | 1 | 80 | 85 | 60 | 86 | Very High |
content-writer | Content Writer | Marketing & Sales | 71 | High | 42,000 | 85,000 | declining | 5 | 3 | Agentic Drafting Pipeline Ownership | Original Reporting & Source Development | Editorial Strategy & Voice | Generic drafting is now a commodity an agent can run start to finish across a full content calendar. Writers left standing are the ones who bring reporting AI can't originate, subject expertise it can't fake, and editorial judgment over what an agent produces. | 48 | 55 | 62 | 68 | 71 | 27-3043 | 1 | 72 | 90 | 76 | 58 | High |
email-marketing-specialist | Email Marketing Specialist | Marketing & Sales | 63 | High | 50,000 | 90,000 | mixed | 5 | 3 | Agentic Lifecycle Automation (Klaviyo, Braze) | Deliverability & Sender Reputation Management | Revenue Attribution & Lifecycle Strategy | Build, test, and send is now a loop an agent runs unattended inside modern ESPs. Specialists who own deliverability health and revenue attribution across the lifecycle stay indispensable. | 42 | 48 | 55 | 60 | 63 | 13-1161 | 1 | 64 | 88 | 72 | 54 | High |
media-buyer | Media Buyer | Marketing & Sales | 61 | High | 40,000 | 90,000 | declining | 4 | 3 | AI-Managed Programmatic Bidding (DSPs) | Publisher Negotiation & Custom Deals | Brand Safety & Fraud Oversight | Bid pacing and reporting now run almost entirely inside AI-managed DSPs. The remaining work is negotiating the deals platforms can't auto-execute and catching the fraud and brand-safety issues an algorithm misses. | null | null | null | null | 61 | 41-3011 | 2 | 62 | 88 | 72 | 58 | High |
seo-specialist | SEO Specialist | Marketing & Sales | 60 | High | 55,000 | 100,000 | mixed | 5 | 3 | AI Search Optimization (AI Overviews/AIO) | Technical SEO & Site Architecture | Content Strategy for Zero-Click Search | Audits and keyword clustering now run on agents, and the audience is shifting from search results to AI answers. Specialists who master AI-search visibility and can still touch the technical stack remain in demand. | 42 | 48 | 53 | 58 | 60 | 13-1161 | 1 | 62 | 86 | 70 | 54 | High |
copywriter | Copywriter | Marketing & Sales | 58 | High | 45,000 | 90,000 | declining | 5 | 3 | Agentic Copy Testing & Prompt Direction | Brand Voice Architecture | Conversion & Launch Copywriting | Catalog and variant copy is now an agent job, run and A/B-tested at volumes no human team could match. Copywriters who survive own the brand voice an agent works from and the high-stakes launch copy nobody trusts it to write alone. | 45 | 52 | 60 | 65 | 58 | 27-3043 | 1 | 60 | 88 | 70 | 50 | Moderate |
growth-marketer | Growth Marketer | Marketing & Sales | 55 | High | 55,000 | 110,000 | mixed | 4 | 3 | Agentic Growth Experimentation Platforms | Growth-Loop & Funnel Design | Statistical Interpretation of Test Results | The build-launch-measure experiment loop is now something an agent runs continuously. Growth marketers who pick the right bets and read ambiguous results correctly are the ones still worth the seat. | null | null | null | null | 55 | 13-1161 | 2 | 56 | 85 | 70 | 56 | High |
social-media-manager | Social Media Manager | Marketing & Sales | 53 | High | 50,000 | 90,000 | stable | 5 | 3 | Agentic Content Scheduling & Creation Tools | Community Management & Crisis Response | Influencer Strategy & Social Analytics | Posting, captioning, and basic reporting now run on agents around the clock. Managers who own real-time community trust and crisis response are the ones a brand still can't automate away. | 38 | 42 | 48 | 52 | 53 | 13-1161 | 1 | 54 | 87 | 70 | 52 | High |
market-research-analyst | Market Research Analyst | Marketing & Sales | 49 | High | 55,000 | 100,000 | stable | 4 | 3 | Agentic Analytics & Survey Tooling | Qualitative Research Methods | Data Storytelling for Leadership | Tabulating data and tracking competitors is now agent work. Analysts who ask the sharper research question and translate findings into a strategic story for leadership are the ones that stay in the room. | 40 | 45 | 50 | 55 | 49 | 13-1161 | 1 | 54 | 74 | 58 | 48 | Moderate |
pr-specialist | PR Specialist | Marketing & Sales | 47 | Moderate | 50,000 | 95,000 | stable | 4 | 3 | AI Media Monitoring & Drafting Tools | Crisis Communication Strategy | Journalist & Influencer Relationship Building | Routine releases and coverage tracking now run themselves. What still needs a person under pressure is a crisis response, a real relationship with a reporter, and a narrative built to survive scrutiny. | 35 | 38 | 42 | 45 | 47 | 27-3031 | 1 | 50 | 78 | 62 | 48 | Moderate |
advertising-manager | Advertising Manager | Marketing & Sales | 44 | Moderate | 70,000 | 130,000 | stable | 4 | 3 | Agentic Ad Platform Management (Advantage+, PMax) | Privacy-First Measurement Strategy | Creative Direction & Campaign Architecture | Bidding, targeting, and creative rotation now run themselves inside agentic ad platforms. Advertising managers who still own the budget strategy, agency relationships, and the creative narrative across channels are the ones platforms can't route around. | 32 | 35 | 38 | 42 | 44 | 11-2011 | 1 | 46 | 82 | 64 | 46 | Moderate |
marketing-manager | Marketing Manager | Marketing & Sales | 41 | Moderate | 75,000 | 140,000 | stable | 4 | 3 | Agentic Marketing Automation Oversight | Marketing Analytics & Attribution | Cross-Channel Brand Strategy | Execution and reporting now run largely on autopilot. Marketing managers who own positioning, customer insight, and orchestration across an ever-multiplying set of channels are becoming more valuable, not less. | 30 | 33 | 37 | 40 | 41 | 11-2021 | 1 | 42 | 84 | 66 | 48 | Moderate |
sales-representative | Sales Representative | Marketing & Sales | 40 | Moderate | 50,000 | 130,000 | stable | 4 | 3 | Agentic Sales Tools (Gong, Outreach, Clari) | Consultative Selling Methodology | Sales Analytics & Forecasting | Prospecting and pipeline admin now run largely on agents. Reps who win are the ones clients trust to understand a nuanced problem and negotiate a complex deal, which no agent closes alone yet. | 30 | 33 | 35 | 38 | 40 | 41-3091 | 1 | 42 | 78 | 62 | 48 | Moderate |
customer-success-manager | Customer Success Manager | Marketing & Sales | 39 | Moderate | 65,000 | 120,000 | mixed | 4 | 3 | AI Health-Scoring & Playbook Platforms (Gainsight) | Enterprise Escalation & Renewal Negotiation | Account Expansion Strategy | Low-touch onboarding and renewal reminders now run on automated playbooks, which is exactly why headcount is under pressure at SaaS companies in 2026. CSMs who survive own the enterprise accounts where an escalation or expansion conversation still needs a trusted human relationship. | null | null | null | null | 39 | 11-9199 | 1 | 42 | 72 | 58 | 48 | Moderate |
account-executive | Account Executive | Marketing & Sales | 37 | Moderate | 70,000 | 160,000 | stable | 4 | 3 | AI SDR & Deal-Intelligence Tools (Gong, Clari) | Enterprise Negotiation & Closing | Multi-Stakeholder Solution Selling | Prospecting and pipeline admin now run largely on AI SDR agents. What still requires a human is reading the room across a multi-stakeholder buying committee and closing a deal a prospect has to trust. | null | null | null | null | 37 | 41-3091 | 2 | 38 | 80 | 68 | 46 | Moderate |
insurance-sales-agent | Insurance Sales Agent | Marketing & Sales | 34 | Low | 40,000 | 85,000 | mixed | 4 | 3 | AI Quote-Engine Oversight & Exception Handling | Complex-Risk Advisory Selling | AI-Assisted Client Outreach & Retention | Personal-lines insurance is already sold by software; agents surviving the shift are advisors on complex risk, not quote processors. | null | null | null | null | 34 | 41-3021 | 2 | 38 | 68 | 48 | 45 | Moderate |
wholesale-sales-representative | Wholesale Sales Representative | Marketing & Sales | 31 | Low | 40,000 | 95,000 | stable | 3 | 3 | B2B AI Ordering & Reorder Platforms | Buyer Relationship & Account Retention | Volume Pricing & Shelf-Space Negotiation | Reordering and catalog lookups are moving to self-service B2B portals. Reps who keep the relationship that wins shelf space and negotiates volume pricing are harder to route around. | null | null | null | null | 31 | 41-4012 | 2 | 36 | 68 | 46 | 34 | Low |
brand-manager | Brand Manager | Marketing & Sales | 28 | Low | 70,000 | 130,000 | stable | 3 | 3 | AI Brand Analytics & Tracking Tools | Brand Positioning & Portfolio Strategy | Creative Direction & Packaging Strategy | AI now handles the competitive tracking and documentation that used to eat a brand manager's week. The job increasingly is positioning strategy and creative judgment, which is exactly what AI can't originate. | null | null | null | null | 28 | 11-2021 | 2 | 30 | 75 | 58 | 36 | Moderate |
retail-salesperson | Retail Salesperson | Marketing & Sales | 19 | Low | 25,000 | 48,000 | stable | 4 | 3 | AI Clienteling & POS Recommendation Tools | Personalized Styling & Consultative Selling | In-Store Experience & Loyalty Building | Self-checkout and inventory tools automate the paperwork, but the floor interaction that actually closes a sale is still an in-person, physical job AI doesn't reach yet. | null | null | null | null | 19 | 41-2031 | 1 | 24 | 50 | 42 | 18 | Low |
product-demonstrator | Product Demonstrator | Marketing & Sales | 11 | Low | 24,000 | 42,000 | stable | 3 | 3 | AI-Assisted Pitch Script Drafting | Live Demonstration & Audience Engagement | Event Brand Experience Delivery | This is one of the most physically-grounded, in-person roles in marketing, and it stays that way: AI can write the script, but it can't stand at the sample table and make the pitch. | null | null | null | null | 11 | 41-9011 | 2 | 16 | 40 | 22 | 8 | Low |
data-entry-clerk | Data Entry Clerk | Administrative & Operations | 92 | Very High | 28,000 | 42,000 | declining | 5 | 3 | AI Document Intelligence & OCR Pipeline Configuration | Data Quality Auditing & Exception Handling | RPA & Workflow Automation Basics | Document AI and agentic extraction pipelines now handle the full capture-validate-file loop for structured data, leaving almost no defensible core to this role as historically specified. The realistic path is moving into the exception queue and the pipeline-configuration work that keeps the automation running. | 78 | 82 | 85 | 88 | 92 | 43-9021 | 1 | 90 | 92 | 80 | 85 | Very High |
administrative-assistant | Administrative Assistant | Administrative & Operations | 80 | Very High | 35,000 | 55,000 | declining | 5 | 3 | Agentic AI Workflow Orchestration (Copilot, Gemini agents) | Cross-Functional Project Coordination | AI Output Review & Executive Judgment | Agentic assistants now run the full scheduling-to-inbox-to-expense loop with minimal supervision, which is why routine admin work keeps climbing toward Very High risk. The defensible ground is judgment-heavy coordination — reading a room, managing competing executive priorities, catching what an agent gets subtly wrong... | 62 | 68 | 74 | 78 | 80 | 43-6014 | 1 | 82 | 88 | 75 | 65 | High |
call-center-agent | Call Center Agent | Administrative & Operations | 73 | High | 28,000 | 40,000 | declining | 4 | 3 | Complex Call De-escalation & Retention | AI Voice-Agent Supervision & Escalation Handling | CRM & Contact-Center Platform Fluency | AI voice agents now run the bulk of routine call center volume end-to-end, which is why this role carries the steepest agentic exposure in the category. Agents who move into escalation handling, retention calls, and AI-voice-agent supervision are the ones still needed on the floor. | null | null | null | null | 73 | 43-4051 | 1 | 74 | 85 | 76 | 66 | High |
payroll-specialist | Payroll Specialist | Administrative & Operations | 73 | High | 45,000 | 70,000 | mixed | 4 | 3 | HRIS & Payroll System Administration | Multi-State & International Tax Compliance | Compensation Analytics | Standard payroll runs are now close to fully automated by platforms with built-in compliance engines, pushing this role toward High risk. The remaining ground is genuinely hard compliance — multi-state, cross-border, system-migration work — where a mistake is expensive enough that employers still want a specialist. | 58 | 63 | 68 | 72 | 73 | 43-3051 | 1 | 74 | 90 | 75 | 58 | High |
receptionist | Receptionist | Administrative & Operations | 72 | High | 28,000 | 42,000 | declining | 4 | 3 | In-Person Customer Experience & Hospitality | Facility Security & Emergency Response | AI Front-Desk Tool Administration | AI voice attendants and check-in kiosks now handle most routine front-desk traffic, which is why this role sits at High risk. The remaining value is in-person: reading a nervous visitor, handling a security issue, making a VIP feel looked after — things a kiosk can't do. | 60 | 65 | 70 | 75 | 72 | 43-4171 | 1 | 76 | 82 | 70 | 58 | High |
customer-service-representative | Customer Service Representative | Administrative & Operations | 67 | High | 30,000 | 45,000 | declining | 4 | 3 | De-escalation & Complex Complaint Resolution | AI Chatbot Handoff & Escalation Management | Multi-Product Technical Troubleshooting | AI chat and voice agents now close a majority of tier-1 tickets without a human touch, which is why routine customer service sits at High risk. The remaining seats go to reps who handle the escalations an agent bounces back — angry customers, ambiguous multi-product problems, and judgment calls the bot won't make. | null | null | null | null | 67 | 43-4051 | 1 | 68 | 85 | 78 | 62 | High |
order-clerk | Order Clerk | Administrative & Operations | 67 | High | 33,000 | 46,000 | declining | 4 | 2 | ERP & Order-Management System Administration | Order Discrepancy & Customer Dispute Resolution | Custom Order Configuration Support | AI order-management systems now take standard orders from intake to confirmation without a human, which is why this role sits at High risk. Clerks handling discrepancies and custom configurations hold the more defensible ground. | null | null | null | null | 67 | 43-4151 | 2 | 72 | 78 | 68 | 58 | High |
executive-assistant | Executive Assistant | Administrative & Operations | 61 | High | 55,000 | 95,000 | stable | 5 | 3 | Executive Judgment & Confidential Stakeholder Management | AI-Assisted Briefing & Decision-Prep Tools | Chief-of-Staff Style Operations | AI now assembles the calendar, drafts the emails, and preps the briefing deck; what it can't do is read a principal's unstated priorities or manage the politics around them. EAs who move toward chief-of-staff-style trusted advising hold their ground better than those who stay purely logistical. | 48 | 53 | 58 | 62 | 61 | 43-6011 | 1 | 66 | 82 | 62 | 52 | High |
inventory-clerk | Inventory Clerk | Administrative & Operations | 59 | High | 30,000 | 42,000 | declining | 4 | 2 | Warehouse Management System (WMS) Administration | Inventory Discrepancy & Shrinkage Investigation | AI Inventory Vision System Oversight | Barcode, RFID, and computer-vision counting have automated most routine stock tracking, pushing this role toward High risk. | null | null | null | null | 59 | 43-5071 | 2 | 68 | 70 | 55 | 45 | Moderate |
records-clerk | Records Clerk | Administrative & Operations | 59 | High | 32,000 | 45,000 | declining | 4 | 2 | Records & Information Governance | AI Document Classification & Retrieval Systems | Compliance Audits & Legal Hold Management | AI document-management systems now file, tag, and retrieve most records automatically, pushing this role to High risk. Clerks who move into records compliance and legal-hold work hold a more defensible position than pure filing and retrieval. | null | null | null | null | 59 | 43-4071 | 2 | 70 | 68 | 50 | 42 | Moderate |
billing-specialist | Billing Specialist | Administrative & Operations | 58 | High | 38,000 | 55,000 | mixed | 4 | 2 | AI Billing & AR Automation Platforms | Complex Dispute Resolution & Collections | Billing System Integration & Migration | AI billing platforms now generate invoices and chase routine payments automatically, which pushes this role to High risk. The durable work is disputed charges and negotiated arrangements where a system alone can't close the case. | null | null | null | null | 58 | 43-3021 | 2 | 62 | 78 | 65 | 52 | High |
office-clerk | Office Clerk (General) | Administrative & Operations | 56 | High | 35,000 | 50,000 | declining | 5 | 3 | Workflow Automation Oversight (approvals, exceptions) | AI Office-Suite Power Use (Copilot, Gemini for Workspace) | Records Compliance & Data Quality | One of America's largest occupations sits squarely in the path of document AI and office agents; clerks who become the person who runs and audits the automation keep the job. | null | null | null | null | 56 | 43-9061 | 2 | 62 | 74 | 52 | 60 | High |
logistics-coordinator | Logistics Coordinator | Administrative & Operations | 53 | High | 42,000 | 70,000 | stable | 4 | 3 | AI Logistics & TMS Platforms (Project44, FourKites) | Disruption Response & Contingency Planning | Carrier & Vendor Relationship Management | Tracking, routing, and standard rate comparison are now largely automated by TMS copilots, but real supply chains still break in ways only a person can untangle. Coordinators who specialize in disruption response and carrier relationships are more durable than those doing pure status-update work. | 42 | 46 | 50 | 55 | 53 | 13-1081 | 1 | 58 | 75 | 60 | 48 | Moderate |
office-manager | Office Manager | Administrative & Operations | 53 | High | 45,000 | 75,000 | stable | 5 | 3 | Operations Management & Process Design | Workplace AI Tool Administration | People Operations & Culture Leadership | Ordering supplies and tracking a budget are now largely automatable, but running an office still means reading a room and fixing whatever nobody planned for. Office managers who take on real cross-functional operations leadership are outgrowing the role's clerical reputation. | 42 | 46 | 50 | 55 | 53 | 43-1011 | 1 | 58 | 78 | 62 | 48 | Moderate |
procurement-specialist | Procurement Specialist | Administrative & Operations | 52 | High | 55,000 | 90,000 | stable | 4 | 3 | AI Procurement & Spend-Analysis Platforms | Strategic Sourcing & Contract Negotiation | Supply Chain Risk Management | AI now handles the transactional half of procurement — POs, catalog pricing, invoice matching — leaving negotiation and supplier risk as the durable core. Specialists who move up into strategic sourcing hold a clearly defensible High-but-not-Very-High position. | 45 | 48 | 53 | 58 | 52 | 13-1023 | 1 | 58 | 72 | 58 | 45 | Moderate |
mail-clerk | Mail Clerk | Administrative & Operations | 50 | Moderate | 30,000 | 40,000 | declining | 4 | 2 | Mailroom Digitization & Scan-to-Distribute Systems | Secure & Sensitive Document Handling | Facilities & Logistics Support Skills | Scan-and-distribute systems now handle a growing share of digital mail routing, but someone still has to physically sort packages and hand-deliver sensitive documents. That keeps this role at Moderate risk rather than pushing it higher. | null | null | null | null | 50 | 43-9051 | 2 | 66 | 55 | 38 | 35 | Low |
supply-chain-analyst | Supply Chain Analyst | Administrative & Operations | 47 | Moderate | 60,000 | 100,000 | stable | 3 | 3 | AI Supply Chain Forecasting & Optimization Tools | Supply Chain Risk & Resilience Design | Applied Data Science for Operations | Forecasting and inventory math are now largely AI-generated starting points, but building a network that survives a real disruption still takes an analyst who understands the business, not just the model. That keeps this role Moderate rather than sliding toward High. | 38 | 42 | 46 | 50 | 47 | 13-1081 | 1 | 52 | 74 | 58 | 48 | Moderate |
facilities-manager | Facilities Manager | Administrative & Operations | 36 | Moderate | 60,000 | 100,000 | stable | 4 | 3 | Smart Building IoT & Predictive Maintenance Platforms | Sustainability & Energy Management | Vendor & Contractor Relationship Management | AI now triages work orders and flags maintenance needs before a person notices, but someone still has to stand in the mechanical room, manage the contractor, and own the safety call. Facilities managers who add sustainability and smart-building program leadership to core operations are growing in scope, not shrinking. | 28 | 30 | 32 | 35 | 36 | 11-3013 | 1 | 42 | 65 | 50 | 32 | Low |
operations-manager | Operations Manager | Administrative & Operations | 26 | Low | 70,000 | 145,000 | stable | 3 | 3 | AI-Augmented Operations Dashboards & Analytics | People Leadership & Performance Management | Cross-Functional Strategy & Process Design | AI now compiles the reports and drafts the schedules an operations manager used to build by hand, but leading a team through an ambiguous, high-stakes day still needs a person in the room. That is why this role stays firmly Low risk even as the reporting work underneath it automates. | null | null | null | null | 26 | 11-1021 | 1 | 30 | 62 | 48 | 30 | Low |
warehouse-manager | Warehouse Manager | Administrative & Operations | 23 | Low | 65,000 | 110,000 | stable | 3 | 3 | Warehouse Safety Leadership & Compliance | AI-Augmented WMS Analytics & Forecasting | Frontline Team Management | AI now handles the scheduling math and throughput reporting, but running a physical floor safely still takes a person who can read the room and react to a forklift breakdown. That keeps this role solidly Low risk. | null | null | null | null | 23 | 11-3071 | 2 | 28 | 58 | 42 | 28 | Low |
medical-transcriptionist | Medical Transcriptionist | Healthcare | 75 | Very High | 30,000 | 50,000 | declining | 5 | 3 | Clinical Documentation Improvement (CDI) | Health Informatics | Medical Coding Certification | Ambient AI scribes now draft the majority of clinical notes directly from the patient encounter, collapsing the separate transcription step almost entirely. The remaining career path runs through documentation-quality oversight, not typing. | 70 | 74 | 78 | 82 | 75 | 31-9094 | 1 | 75 | 92 | 75 | 60 | High |
medical-coder | Medical Coder | Healthcare | 57 | High | 40,000 | 70,000 | mixed | 5 | 3 | Agentic Coding Pipeline Auditing | Clinical Documentation Improvement | Healthcare Compliance & Auditing | AI coding engines now assign codes for routine single-diagnosis visits with production-grade accuracy, and agentic pipelines are starting to run the full code-verify-submit loop with a human auditing exceptions. Coders who specialize in complex multi-comorbidity cases and own that audit layer stay essential. | 48 | 53 | 60 | 65 | 57 | 29-2072 | 1 | 62 | 80 | 55 | 55 | High |
pharmacy-technician | Pharmacy Technician | Healthcare | 42 | Moderate | 33,000 | 45,000 | mixed | 4 | 3 | Pharmacy Automation System Oversight | Sterile Compounding Certification | Insurance & Prior-Authorization Specialization | Robotic dispensing and automated insurance adjudication now run routine fills close to end to end, which is why this is the most exposed role in the healthcare category. Technicians who move into compounding or automation-system oversight have the clearer path forward. | null | null | null | null | 42 | 29-2052 | 2 | 50 | 68 | 48 | 38 | Moderate |
pharmacist | Pharmacist | Healthcare | 35 | Moderate | 120,000 | 160,000 | mixed | 4 | 3 | Clinical Pharmacy Specialization | Pharmacogenomics | AI in Drug Development & Safety | Robotic dispensing and AI interaction-checking now run routine fills close to end to end, with a pharmacist reviewing exceptions. Pharmacists moving into clinical consultation and medication-therapy management are seeing their scope grow, not shrink. | 30 | 33 | 37 | 40 | 35 | 29-1051 | 1 | 40 | 70 | 55 | 35 | Low |
radiologist | Radiologist | Healthcare | 34 | Moderate | 300,000 | 500,000 | stable | 4 | 3 | AI Imaging Tool Proficiency | Interventional Radiology | AI Validation & Clinical Informatics | AI now triages and pre-screens most normal-study volume and drafts routine reports, compressing read time on straightforward cases. The complex, ambiguous and interventional work that defines the specialty still sits with the radiologist. | 38 | 40 | 42 | 45 | 34 | 29-1224 | 1 | 36 | 80 | 58 | 38 | Moderate |
End of preview. Expand in Data Studio
AI Career Threat Index
An open dataset scoring 300 US professions on AI displacement risk. Every role carries four published sub-scores — task automation potential, AI tool maturity, employer adoption, and agentic exposure (exposure to autonomous multi-step agents, distinct from copilot-style tools) — with a one-sentence written rationale per factor, combined by an open formula into a single 0–100 score (higher = more exposed). MIT licensed, reviewed quarterly, versioned on GitHub.
Most AI-and-jobs discussion is anecdotal. This dataset gives one numeric answer per role and
publishes the factor judgments behind it so they can be argued with — for journalists,
researchers, career tooling, and workforce analysis. To our knowledge the agenticExposure
factor is the first published per-occupation scoring of exposure to autonomous agents
specifically.
Files
ai-career-threat-index.csv— flat table, one row per role (loads as the default config)ai-career-threat-index.json— full structured records: sub-score rationales, task lists, defense skills, industry modifiers, quarterly historysoc-crosswalk.csv— every BLS SOC 2018 detailed occupation (867 codes) mapped to its nearest scored role with a match-quality label
Columns (CSV, one row per role)
| Column | Type | Description |
|---|---|---|
slug |
string | URL-safe id (e.g. software-engineer) |
title |
string | Role display name |
category |
string | One of 15 role families |
score |
int 0–100 | AI displacement risk; higher = more exposed |
risk_level |
enum | Low (0–35), Moderate (36–50), High (51–75), Very High (76–100) |
salary_low_usd, salary_high_usd |
int | USD range anchored to BLS OES (~p25–p75) |
salary_trend |
enum | rising / stable / mixed / declining |
tasks_at_risk_count, tasks_growing_count |
int | Task counts behind the score |
defense_skill_1..3 |
string | Top skills to build for the role |
insight |
string | Headline finding |
score_q1_2025 … score_q3_2026 |
int | Quarterly history (new roles start Q3 2026 — never backfilled) |
soc_code |
string | BLS SOC 2018 detailed occupation code |
tier |
int 1–2 | 1 = head role with full editorial depth |
sub_task_automation, sub_tool_maturity, sub_adoption, sub_agentic_exposure |
int 0–100 | The four published factors |
agentic_risk |
enum | Band on agentic exposure alone |
How the score is built (methodology v2, Q3 2026+)
score = taskAutomation × (0.45 + 0.30·toolMaturity/100 + 0.25·adoption/100) + 0.10·agenticExposure
Task automation potential sets the ceiling (you can't be displaced from work AI can't do); tool maturity and employer adoption determine how much of that ceiling is realized today; agentic exposure adds forward pressure from autonomous agents. Full rubric with scoring anchors and calibration examples: pipeline/RUBRIC.md.
Quarters through Q2 2026 were published under methodology v1 (50/30/20 composite, no agentic factor) and are retained as published. Quarterly moves over ±8 points carry a stated restatement reason (four in Q3 2026).
Sources: O*NET task lists, public AI capability evidence, employer surveys (BCG, Gallup, WEF, McKinsey), and BLS Occupational Employment Statistics.
Load it
from datasets import load_dataset
ds = load_dataset("Jott2121/ai-career-threat-index")
df = ds["train"].to_pandas()
# Roles most exposed to autonomous agents specifically
print(df.nlargest(10, "sub_agentic_exposure")[["title", "sub_agentic_exposure", "score"]])
# Most AI-resilient roles
print(df.nsmallest(5, "score")[["title", "score"]])
Interactive explorer: https://jott2121.github.io/ai-career-threat-index/
Honest limitations
The scores are structured editorial estimates made against a published rubric — not empirical measurements of realized job loss. The formula and every sub-score are published precisely so they can be challenged; treat the dataset as a structured, versioned prior for analysis and reporting, not ground truth. Scores within 5 points of a band boundary may retain their prior band across quarterly reviews to avoid noise.
Citation
@dataset{ai_career_threat_index,
title = {AI Career Threat Index},
author = {Otterson, Jeff},
year = {2026},
url = {https://github.com/Jott2121/ai-career-threat-index},
doi = {10.5281/zenodo.21227026},
note = {Open dataset of AI displacement risk for 300 professions with published
sub-scores including agentic exposure, MIT licensed}
}
Canonical repo: https://github.com/Jott2121/ai-career-threat-index — License: MIT.
- Downloads last month
- 41