The 10 Roles Most Exposed to AI Displacement in 2025–2026
Not all knowledge work is equally exposed to AI. Based on current AI capability benchmarks, labor market data, and Runway assessment data, these ten roles face the highest structural exposure — measured by the proportion of their task mix that falls within current or near-term AI automation capability.
This is not a prediction of job losses. It is an assessment of structural exposure — how much of the work can AI already do, or will be able to do within 24 months.
1. Data Entry & Administrative Specialist
Exposure score: 89/100
The most exposed role category. AI can already handle document processing, form filling, data extraction, email triage, and scheduling at production quality. The remaining human-required tasks (exception handling, relationship management) represent a small fraction of the total role.
What to do: Transition toward process design, workflow automation oversight, or exception management. The administrative role is not disappearing — it is transforming into an automation management role.
2. Junior Content Writer / Copywriter
Exposure score: 82/100
First-draft content creation is now an AI commodity. Blog posts, marketing emails, social media copy, product descriptions — AI produces these at acceptable quality faster and cheaper than a junior writer. The remaining value is in strategy, brand voice judgment, and editing.
What to do: Move up the value chain. Content strategy, brand positioning, editorial judgment, and audience insight are all defensible. The skill to build is not better writing — it is better thinking about what to write.
3. Financial Analyst (Reporting-Focused)
Exposure score: 76/100
Financial analysts who primarily build reports, consolidate data, and create standardised models are highly exposed. AI can already generate financial summaries, create charts from data, and identify patterns in financial statements.
What to do: Specialise in judgment-intensive analysis — scenarios with genuine uncertainty, stakeholder advisory, and strategic interpretation that requires understanding context machines cannot access.
4. Customer Support Representative
Exposure score: 74/100
AI chatbots and automated resolution systems now handle 60–70% of Tier 1 customer inquiries at major companies. The remaining human work is complex escalation, empathy-requiring situations, and edge cases.
What to do: Move toward escalation specialisation, customer experience design, or AI training and quality assurance. The generalist support role is compressing fast.
5. Paralegal / Legal Research Assistant
Exposure score: 71/100
Document review, legal research, contract analysis, and case summarisation are all within current AI capability. Law firms are already reducing paralegal headcount by 20–30% through AI adoption.
What to do: Specialise in areas requiring licensed judgment, client-facing work, or regulatory expertise that changes faster than training data can capture.
6. Junior Software Developer
Exposure score: 68/100
AI coding assistants now generate 30–50% of production code at companies that adopt them. Junior developers who primarily write boilerplate code, fix simple bugs, and implement well-defined features are the most exposed segment of the software engineering workforce.
What to do: Focus on system design, architecture decisions, code review judgment, and understanding business context. The skill that protects software engineers is not coding speed — it is knowing what to build and why.
7. Marketing Coordinator
Exposure score: 65/100
Campaign execution, social media scheduling, email template creation, basic analytics, and report generation — the core tasks of a marketing coordinator — are all within AI capability. The strategic and creative elements remain human-dependent.
What to do: Build expertise in marketing strategy, audience insight, brand positioning, or marketing technology orchestration. Move from execution to direction.
8. Translation & Localisation Specialist
Exposure score: 63/100
AI translation quality has crossed the threshold for most business content. Technical documentation, marketing copy, and business communications can now be translated at production quality. Literary translation and culturally sensitive adaptation remain human-dependent.
What to do: Specialise in cultural adaptation, transcreation, or quality assurance for AI-generated translations. The pure translation role is compressing, but the quality layer remains valuable.
9. Bookkeeper / Accounting Assistant
Exposure score: 61/100
Transaction categorisation, reconciliation, invoice processing, and basic compliance reporting are increasingly automated. The remaining human work centres on judgment calls, exception handling, and advisory.
What to do: Move toward advisory accounting, financial planning, or compliance specialisation. The transactional core of bookkeeping is being automated across the industry.
10. HR Administrator
Exposure score: 58/100
Benefits administration, onboarding paperwork, policy document management, and basic employee inquiry handling are all within AI capability. The relationship-dependent and judgment-intensive elements of HR remain defensible.
What to do: Transition toward workforce intelligence, AI ethics governance, employee experience design, or change management. The administrative core is compressing.
What these numbers mean
An exposure score of 70 does not mean 70% of people in that role will lose their jobs. It means that 70% of the typical task mix for that role falls within current or near-term AI automation capability.
The practical impact is role compression: fewer people needed to produce the same output. The professionals who survive compression are the ones who do the work AI cannot — judgment, relationships, creativity, and adaptation.
If your role is on this list, the right response is not panic. It is action. The displacement timeline gives you a window. What you do with that window determines whether AI is a threat or an opportunity.
Exposure scores are based on Runway's scoring methodology, which analyses task vulnerability against current AI capability benchmarks. Individual scores vary significantly based on work environment, seniority, and specific task distribution. Take your own assessment for personalised results.