25 Jobs Most at Risk from AI in 2026 (Ranked by Exposure Score)

The Highest-Paid Professionals Are the Most Exposed

Here's what almost everyone gets wrong about AI job displacement: they think it starts at the bottom.

It doesn't.

Jobs paying under $35K average a 3.4 AI exposure score. Jobs paying over $100K average 6.7. The more you earn, the more you're in the crosshairs. That's not intuitive. But it makes sense when you trace what high-income work actually is: pattern recognition, synthesis, structured communication. Exactly what large language models do best.

The global average across 500+ occupations is 5.3 out of 10. Most jobs are in the middle, partially restructured, not eliminated. But that middle is moving fast. And the jobs at the top of the risk curve are moving fastest.

This list is built from our displacement scoring methodology, applied to Bureau of Labor Statistics occupation data, cross-referenced with AI adoption rates and 10-year employment outlook projections. A high score alone isn't a death sentence. But a high score paired with a declining outlook? That's the danger zone.

How to Read the Score

The AI Displacement Score runs 0-10. Here's what the ranges actually mean:

• 9-10: Disruption happening now. Core tasks are being automated today.
• 7-8: Restructuring within 2-3 years. The role survives but shrinks.
• 5-6: Pressure arriving in 5+ years. Time to adapt, not panic.
• 0-4: Relatively insulated. Physical, social, or perceptual tasks dominate.

Only 3% of occupations score 9-10. Most of the damage is in that 7-8 band. It's not mass extinction. It's mass restructuring. Which, for the people in those roles, is its own kind of disruption.

The 25 Jobs Most at Risk from AI in 2026

The 10s and 9s: Disruption Now

1. Medical Transcriptionist (Score: 10)
The benchmark case. Every task in this job, listening, transcribing, formatting clinical notes, is now handled by AI tools like Nuance DAX and Suki. Job outlook: -8%. This is what full displacement looks like.

2. Data Entry Clerk (Score: 10)
Structured input tasks are the lowest-hanging fruit for automation. High volume, low ambiguity, no social component. Already gone in most modern organizations that have adopted ERP or RPA tooling.

3. Telemarketer (Score: 9.8)
Conversational AI now handles outbound calls with near-human cadence and objection handling. The humans who remain are closing, not dialing. The dialers are mostly gone.

4. Paralegal (Score: 9.5)
Document review, case research, contract summarization. These are LLM-native tasks. Top law firms already use AI for first-pass discovery. The paralegal role is being absorbed into the attorney's workflow, not eliminated cleanly, but hollowed.

5. Insurance Underwriter (Score: 9.5)
Risk classification at scale is a prediction problem. AI handles it faster, cheaper, and with fewer errors on structured data. Employment in this category is already declining. Job outlook: -4%.

6. Loan Officer (Score: 9.2)
Same pattern. Creditworthiness assessment is a structured decision tree. AI doesn't tire at 4pm on a Friday. The human loan officer is increasingly a relationship layer on top of an automated recommendation engine.

7. Tax Preparer (Score: 9.1)
Not the CPAs doing strategic tax planning. The preparers handling W-2 filings and standard returns. TurboTax was an early harbinger. Current AI tools make that look slow.

8. Bookkeeper (Score: 9.0)
Accounts receivable, accounts payable, bank reconciliation. QuickBooks, Xero, and their AI layers handle this with minimal human input. The bookkeeper is increasingly a QA function on an automated system.

The 8s: Restructuring Within 2-3 Years

9. Radiologist (Score: 7.8)
This one deserves its own section. Radiologists earn $400K+. They also analyze images for patterns, which is exactly what convolutional neural networks are trained to do. FDA-approved AI tools already match or exceed human accuracy on specific scan types. The job isn't gone. But fewer radiologists will read more scans. That's compression, not elimination. And compression means fewer positions.

10. Financial Analyst (Score: 7.6)
Junior analysts building models in Excel are the first to go. Senior analysts interpreting those models still matter. But the entry-level pipeline, the training ground that historically made senior analysts, is evaporating.

11. Legal Secretary (Score: 7.5)
Document drafting, scheduling, correspondence formatting. AI handles the drafts. Humans handle the judgment calls. The ratio is shifting fast.

12. Market Research Analyst (Score: 7.4)
Survey design, data synthesis, trend reporting. LLMs ingest and summarize competitive landscapes faster than any human team. The insight layer still requires judgment. The synthesis layer doesn't.

13. Software Developer (Score: 8.5)
Here's the paradox that breaks most people's mental model.

Software developers score 8-9 on AI exposure. Job outlook is +25%. High score. Booming demand. The same tool that threatens the role is also expanding what the role can build.

The developers who thrive are the ones using AI as a force multiplier. The ones who resist are the ones who get replaced by developers who don't. This is a taste of what's coming for every knowledge worker.

14. Content Writer (Score: 7.3)
Not all content writing. Long-form strategic content with genuine expertise and perspective is still human work. SEO-optimized articles, product descriptions, templated reports? Already automated in most digital marketing stacks.

15. Customer Service Representative (Score: 7.8)
Tier 1 support is largely gone in organizations that have invested in AI. Tier 2 is under pressure. Tier 3 complex, judgment-intensive interactions remain human. The funnel has inverted.

16. Graphic Designer (Score: 7.2)
Midjourney, DALL-E, Adobe Firefly. Production design and asset generation are heavily compressed. Brand strategy and creative direction still require humans. The middle band, production execution, is gone.

17. Accountant (Score: 7.1)
Same story as bookkeepers, one level up. Compliance and audit work is increasingly AI-augmented. Strategic advisory work is not. The profession is bifurcating.

18. Sales Development Representative (Score: 8.0)
This is where the second-order effects get interesting. VP of Sales scores 6. The SDR reporting to them scores 8. AI handles prospecting, sequencing, and first-touch outreach at a scale no human team can match. The SDR role as currently defined is one of the most at-risk jobs in tech sales.

19. Proofreader (Score: 8.5)
Grammar, style consistency, factual flagging. These are solved problems for current AI tooling. This role is effectively absorbed into the writer's workflow.

20. Pharmacy Technician (Score: 7.0)
Dispensing, labeling, inventory management. Automated pharmacy systems are deployed in hospital networks today. The technician role survives in patient-facing contexts. The back-of-pharmacy tasks are already automated.

High Score, Surprising Reasons

21. Translator (Score: 7.5)
Not literary translation. Not diplomatic interpretation. But technical documents, standard business correspondence, product localization? Machine translation quality has crossed the professional threshold for most content types.

22. Claims Adjuster (Score: 7.3)
Standardized claims, property damage assessments, auto accident evaluations. AI tools are deployed by major insurers for first-pass review. Complex litigation remains human. Simple claims processing doesn't.

23. Dispatcher (Score: 7.1)
Route optimization, scheduling, resource allocation. These are solved optimization problems. AI handles them continuously, not in eight-hour shifts.

24. HR Coordinator (Score: 7.0)
Resume screening, onboarding documentation, benefits administration. AI handles the volume. Strategic HR, culture building, conflict resolution, remains human. The coordinator layer is compressing.

25. Research Assistant (Score: 7.6)
Literature reviews, citation management, data compilation. Andrej Karpathy's 342-occupation analysis from March 2026 reinforces this. Research synthesis is now a hybrid workflow, with AI doing the first pass and humans doing the interpretation. The pure assistance role has less to do.

The Comparison That Should Make You Uncomfortable

Radiologists score 7.8. Surgeons score 3.

Same hospital. Same medical school pipeline. Radically different futures.

The radiologist's work, image interpretation, pattern matching at scale, is a prediction task. AI does prediction tasks. The surgeon's work requires physical dexterity, real-time judgment in uncontrolled environments, and tactile feedback no current system can replicate. Same career prestige. Completely different exposure.

This is the lesson. Your job title is almost meaningless. The tasks inside it determine your actual risk. A radiologist who pivots to interventional radiology is buying time. A radiologist who doubles down on pure diagnostic reading is not.

81% of physicians now use AI daily, up from 38% in 2023. The adaptation is already happening at the professional level. The question is whether individuals inside those professions are adapting with it.

What This Actually Means for You

Three things worth knowing before you panic or dismiss this entirely:

First, score plus outlook equals your real signal. A score of 9 with +15% job growth (software development) is different from a score of 9 with -8% outlook (medical transcription). One is transformation. One is elimination.

Second, AI skills command a 56% salary premium today. The people learning to use these tools are not losing ground. They're pulling ahead. Rapidly. The gap between AI-augmented and non-augmented workers is widening every quarter.

Third, 42% of Gen Z is pursuing trades. Plumbers score 1 on AI exposure. HVAC technicians score 0-2. The market is already pricing in AI risk. The surge toward physical skilled work isn't nostalgia. It's rational.

Bottom Line

The jobs AI will replace first are not the lowest-paid or the lowest-skilled. They're the ones most built on pattern recognition, synthesis, and structured communication at scale. That description fits a lot of expensive people.

Your exposure score is not your destiny. But ignoring it is a choice with consequences.

The workers gaining ground in 2026 are not the ones waiting to see what happens. They're the ones who already looked at the data, identified the specific tasks inside their role that are automatable, and started building the skills that aren't.

Survival in this shift has less to do with your job title and more to do with how clearly you see your own task stack. The people who see it clearly can act. The ones who don't will be surprised.

And surprise, in a restructuring economy, is expensive.