The debate over which jobs AI will take has been running since ChatGPT launched in late 2022. Most articles cite the same handful of academic studies from 2023. This one draws on DisplaceIndex's own occupation scoring database, built on two-model consensus scoring (Claude Sonnet and GPT-4o) across 57 US occupations and updated through Q1 2026.
Here is what the data shows.
How We Score AI Job Risk
DisplaceIndex's AI Exposure Score runs from 0 to 100. It measures what percentage of the typical tasks in a role today's AI tools can credibly perform, based on current systems rather than speculative future capability. The full methodology is documented on the DisplaceIndex methodology page.
A score of 84 means roughly 84% of the core tasks in that occupation can be completed by existing AI. Task coverage is not job-loss probability. An occupation can have high task coverage and stable employment if AI handles peripheral tasks, if regulatory constraints prevent automation, or if demand for the service is growing faster than AI reduces headcount.
The 10 Jobs Under Most AI Pressure Right Now
Each occupation below scored 58 or higher on the DisplaceIndex AI Exposure Scale. All scores reflect two-model consensus methodology.
Data Entry Clerk: 92/100
The highest-scoring occupation in the DisplaceIndex database. Structured data input from standardized sources is fully automatable with current technology. Remaining demand is concentrated in unstructured or context-dependent input, or in sectors where digital transformation has been slow such as parts of healthcare and government. Volume is declining year on year.
Insurance Underwriter: 84/100
The core task of reviewing applications against risk criteria and setting premiums is almost entirely rule-based. AI platforms are automating standard personal lines and commercial SMB underwriting, which accounts for the bulk of entry and mid-level underwriting volume. According to the BLS Occupational Outlook Handbook, employment of insurance underwriters is projected to decline 4% through 2033. Complex or unusual risks still require human judgment.
Legal Secretary: 82/100
Document drafting, case management, scheduling, and correspondence are all tasks where AI tools have become standard in mid-size and large law firms. The ABA Legal Technology Survey confirms measurable uptake of AI document tools across US law firms. Partners and litigators face lower exposure because their value is tied to judgment and client trust rather than document execution.
Copywriter: 78/100
Standard commercial copy including product descriptions, ad copy, email sequences, and landing pages is the category most directly affected by large language models. The constraint has shifted from writing speed to strategy, brand voice calibration, and the ability to brief and edit AI output effectively. Copywriters who reposition as content strategists are better placed than those competing on output volume.
Financial Analyst (Junior): 75/100
Data collection, model building, and report generation that occupy junior analyst hours are rapidly automated. Bloomberg, FactSet, and AI-native financial platforms now produce equity research summaries, comparables analysis, and sector overviews in seconds. Senior analysts with client relationships and original investment frameworks face lower risk because their value is embedded in judgment, not execution.
Instructional Designer: 72/100
AI authoring tools have compressed content development from weeks to days. Compliance training, product onboarding, and skills training programs are the highest-volume learning and development categories, and these are the formats most susceptible to AI generation at scale. Custom leadership development and bespoke certification programs remain more defensible.
Real Estate Appraiser: 70/100
Automated Valuation Models have been part of mortgage origination for years, and their accuracy has improved significantly. Fannie Mae's appraisal modernization program now accepts desktop and hybrid appraisals for a large proportion of conforming loans, reducing the need for full in-person appraisals. Unique properties, disputed valuations, and litigation support still require licensed human expertise.
Customer Service Representative: 68/100
AI handles Tier 1 support effectively: password resets, order status queries, straightforward refund requests. The limitation is context. AI struggles with emotionally charged interactions, unusual edge cases, and customers who are determined not to engage with a bot. Companies are bifurcating their support operations, automating the high-volume routine queries while retaining agents for complex and high-value interactions.
Actuary: 62/100
AI is accelerating front-end modeling work including data preparation, scenario generation, and preliminary pricing runs. Senior actuarial judgment on product design, regulatory sign-off, and risk strategy remains genuinely difficult to automate. The profession is being restructured around oversight of AI-generated analysis rather than displaced outright, but the number of actuaries needed to produce a given volume of work is falling.
Corporate Trainer: 58/100
AI e-learning authoring platforms are generating course content in hours that previously required weeks of instructional design work. Compliance training automation is commoditizing the largest category of corporate training delivery. Live facilitation, executive coaching, and bespoke leadership programs remain more defensible because their value is relational, not informational.
What At Risk Actually Means
A high AI Exposure Score signals where AI is capable today, not where jobs will disappear on a fixed schedule. The Goldman Sachs AI employment study estimated 300 million full-time equivalent jobs globally are exposed to automation, but stresses that exposure does not equal displacement. Three factors determine whether technical capability translates to actual job loss.
Economic friction: replacing a human workflow costs money, time, and organizational change management. Many firms are absorbing AI productivity gains through attrition rather than active layoffs. The transition is happening, but it is slower and messier than a capability score implies.
Regulatory constraints: appraisers, underwriters, and actuaries operate in regulated environments where AI output requires sign-off from a licensed professional. Regulation is slowing displacement even where AI is technically capable of performing the task independently.
Demand elasticity: in some roles, AI reduces cost enough that demand expands. If AI-assisted copywriters can produce more content and clients buy more content at lower prices, net employment may stay flat even as output per worker increases dramatically.
The DisplaceIndex homepage tracks real employment data from the Federal Reserve alongside these task coverage scores. The combination gives a fuller picture of where hard displacement is already showing up in the labor market numbers.
Jobs Being Augmented Rather Than Replaced
Not every high-exposure role is heading for displacement. Roles scoring 40 to 60 on DisplaceIndex often show AI augmenting output rather than replacing workers. Physicians using AI diagnostic tools, architects working with generative design software, and engineers using AI-assisted simulation are increasing output without reducing headcount.
The dividing line tends to be whether AI handles the repetitive, standardized component of the work and leaves humans to do the judgment-heavy part, or whether the judgment component itself is being automated. When AI can do the reasoning, not just the execution, the displacement dynamic is more severe.
What to Do if Your Job Is on This List
The occupations most at risk share a characteristic: their value was historically measured in time-to-complete rather than judgment quality. As AI compresses execution time, differentiation moves to the elements AI cannot yet replicate, including accountability, trust, ethical reasoning, and the ability to navigate genuinely ambiguous situations.
Three moves the data supports:
- Reframe toward outcomes, not tasks. Document the decisions you make and the judgment calls you are accountable for, not just the procedures you execute. AI can follow procedures; it cannot yet hold accountability for consequences.
- Acquire AI operation skills. The fastest-growing job titles in 2026 all involve using AI tools as a core part of the role. The workers displaced by automation are increasingly those who chose not to adopt it.
- Identify the AI-resistant fraction of your role. Check the task-level breakdown on DisplaceIndex for your specific occupation to see exactly which tasks score highest for AI coverage and which remain lower.
Scores and task-level breakdowns for all 57 occupations are on the DisplaceIndex occupations page. The data is updated whenever new scoring runs are completed and the full methodology is documented on the methodology page.