Jobs Most at Risk From AI [Mar 2026]

🔴 The 30 Jobs Most at Risk From AI

These are the roles our scoring framework flags as most vulnerable. Every one scores below 33 on the JobZone Score — meaning AI can already perform the majority of their core tasks. The work is digital, the patterns are repeatable, and the regulatory barriers are minimal. If your role is on this list, the displacement risk is real and the timeline is years, not decades.

44.3M US workers — 26% of the mapped workforce — work in RED zone roles. That is not a theoretical number. It maps to real BLS employment data for each role. Below are the 30 roles with the lowest JobZone Scores, ranked from most vulnerable.

These roles represent the front line of AI displacement. They share a common DNA: the work is entirely digital, follows predictable patterns, faces no regulatory barriers to AI performing it, and requires no physical presence. Data entry, basic bookkeeping, routine customer service, and content moderation lead the list.

Being in the RED zone does not mean the role disappears overnight. It means the core tasks can already be performed by AI. The timeline for actual displacement depends on employer adoption speed, cost comparisons, and organisational inertia. Some RED zone roles will persist for years because “good enough” AI output still requires human quality control. But the direction is unambiguous.

The roles at the bottom of this list represent a distinctive cluster. They are not low-skill in the traditional sense — many require training, software proficiency, and domain knowledge. What they lack is the structural protection that keeps AI from performing their core workflows. A bookkeeper needs accuracy, not physical presence. A data entry specialist needs speed, not a licence. A customer service representative needs patience, not unpredictable judgement. AI delivers all three.

Across the 3649 roles in our database, the average score is 45.1 out of 100. The RED zone threshold sits at 33. Roles below that line have scored poorly across all five assessment dimensions: resistance to AI, evidence of AI capability, barriers to automation, protective principles, and AI growth correlation. A score below 33 means the role has essentially no structural defence remaining.

What the Institutions Say

Major research institutions have published their own displacement estimates. The numbers vary widely because they measure different things — full replacement, task automation, or occupational exposure. Our RED zone data sits alongside these forecasts as the role-by-role, ground-truth layer.

🎯 What Makes These Jobs Vulnerable?

The most at-risk roles don’t fail on one dimension. They fail on all five. Our scoring framework evaluates resistance, evidence, barriers, protective principles, and AI growth correlation. RED zone roles score poorly across the board — and they share a distinctive profile.

The pattern is consistent: roles that fail on all four traits end up in the RED zone. Roles that hold one or two protections — a licensing requirement, a physical component — land in YELLOW. Roles with multiple structural barriers land in GREEN.

Our five-dimension scoring framework captures these traits precisely. Resistance measures how well current AI can perform the role’s core tasks. Evidence tracks real-world examples of AI already doing the work. Barriers identify licensing, physical, and regulatory protections. Protective principles cover ethical and trust requirements. AI growth correlation measures whether improving AI models will further erode the role. RED zone roles score low on every dimension.

The Five Dimensions of Vulnerability

Our AIJRI scoring framework evaluates every role across five dimensions. RED zone roles typically score low on all five — there is no single dimension carrying them. Here is how each dimension maps to vulnerability:

The compound effect matters. A role that scores poorly on one dimension but strongly on others stays in the YELLOW or GREEN zone. Bookkeepers, for example, have low resistance (AI handles the maths) but no barriers, no physical component, and strong AI growth correlation. Every dimension reinforces the vulnerability. That is the RED zone pattern.

⚠️ RED Imminent — Already Being Replaced

70 roles score below 20 on the JobZone Score, placing them in the RED Imminent category. Below 20 on the JobZone Score, roles enter RED Imminent territory. These aren’t jobs that might be automated in five years. The tools exist today. AI chatbots handle reception. Automated systems process invoices. LLMs draft the emails that once filled an office clerk’s day.

RED Imminent is the most acute category in our framework. These roles don’t just face theoretical risk — the AI tools to perform their core functions are commercially available today. Chatbots already handle front-desk queries. Invoice processing is automated end-to-end. Template-based content generation has moved from experimental to production-grade.

The gap between “AI can do this” and “companies have replaced this” is closing. For RED Imminent roles, the main factors keeping humans in the loop are organisational inertia, IT migration costs, and customer preference for human contact. None of those barriers are permanent.

How RED Imminent Differs From RED

The distinction between RED (<33) and RED Imminent (<20) is not just numerical — it reflects a qualitative difference in how close the role is to full automation. RED zone roles face significant AI pressure on their core tasks. RED Imminent roles face near-total AI coverage: the tools exist, they work at production quality, and they cost a fraction of human labour.

A RED zone role scoring 28 might have AI covering 60-70% of tasks, with meaningful human components remaining. A RED Imminent role scoring 12 has AI covering 85-95% of tasks, with only edge cases and quality review remaining. The practical difference: RED zone workers see their jobs changing. RED Imminent workers see their jobs disappearing.

For career planning, the implication is clear: if your role is in RED Imminent, the transition timeline is not 2-5 years. It is 6-18 months. Companies that have not already automated these roles are typically held back by IT migration costs, not by AI capability gaps. Once those migrations complete, the human position is eliminated or fundamentally restructured.

⚡ YELLOW Urgent — Next in Line

287 roles sit in the YELLOW Urgent category. YELLOW Urgent roles score between 33 and 45. They’re not in the RED zone yet, but the gap is narrow. Most have one defensive trait keeping them above the line — a requirement for some human judgement, a partial physical component, or a regulatory overhead that slows full automation. Remove that single defence, and they cross over.

YELLOW Urgent is the transition zone. These roles are not safe — they are one AI breakthrough away from crossing into RED. The typical profile: mostly digital work with a thin layer of human judgement, partial regulatory requirements, or occasional physical presence that keeps the overall score above 33.

When combined with the 200 RED zone roles, 487 positions in our database face significant AI displacement risk — 13% of all assessed roles. The distinction between RED and YELLOW Urgent is one of timing, not direction.

Why YELLOW Urgent Matters for Risk Assessment

YELLOW Urgent roles are important because they represent the next wave of displacement. Today’s RED zone was last year’s YELLOW Urgent — as AI capabilities improve, roles that were marginally protected lose their one remaining defence. A role scoring 38 today might score 30 after the next generation of AI models launches.

The typical YELLOW Urgent profile: mostly digital work with occasional physical touchpoints, moderate complexity that requires some human judgement, or partial regulatory requirements that slow but do not prevent automation. These are roles where AI handles 60-70% of the work today and the remaining 30-40% is eroding with each capability jump.

For workers in YELLOW Urgent roles, the strategic calculus differs from RED zone workers. RED zone workers need to pivot entirely. YELLOW Urgent workers have time to strengthen their one remaining defence — pursuing a certification, adding a physical component to their work, or moving into client-facing responsibilities that require human trust. The goal is to add enough structural protection to move the score above 45 and into stable YELLOW territory.

🏭 Which Industries Face the Most Risk?

AI displacement is not evenly distributed across industries. Some sectors have high concentrations of digital, repeatable work that AI handles well. Others are protected by physical presence, licensing, or human trust. The domain-level average scores reveal which industries carry the highest concentration of vulnerable roles.

Average JobZone Scores vary across career domains. Lower averages mean a higher concentration of vulnerable roles. Individual roles within any domain can vary widely — a senior specialist and a junior analyst sit in very different zones despite sharing an industry.

The sector-level pattern maps directly to the vulnerability traits: domains where work is primarily digital and administrative score lowest. Domains where work requires physical presence, licensing, or human trust score highest. This is not a coincidence — it is the defining pattern of AI displacement risk.

Most Exposed Domains by Average Score

The domains with the lowest average scores contain the highest concentration of at-risk roles. Within each domain, the spread from highest to lowest individual score can be dramatic — a finance director and a junior bookkeeper share a domain but not a zone. The tables below show the lowest-scoring roles in the most exposed sectors.

Creative and marketing roles appear prominently in the at-risk sectors because their core output — text, images, video concepts, campaign copy — is exactly what generative AI produces. Junior creative roles are the most exposed. Senior strategic and client-facing roles retain more protection because they involve relationship management, brand judgement, and cross-functional coordination that AI cannot yet replicate.

The pattern across all exposed domains is consistent: the more a role’s output is purely digital content, the higher its displacement risk. Roles that involve managing teams, making strategic decisions, or maintaining client relationships retain significantly more protection within the same domain.

💰 Finance & Admin: The Most Exposed Sectors

Finance and administrative support are the two sectors that institutional research consistently flags as most exposed to AI. Goldman Sachs estimates 46% of admin support tasks are automatable. Banking has moved faster than almost any other industry in deploying AI across operations, compliance, and customer service.

The finance sector has moved faster than almost any other industry in deploying AI. Banking compliance, fraud detection, loan processing, and risk assessment are all being automated. For junior finance roles — bookkeeping, data reconciliation, basic analysis — the displacement timeline is compressed because the AI tools are already in production at the institutions that employ them.

Administrative support roles face similar pressure. Goldman Sachs estimates 46% of admin support tasks are automatable by current AI. The roles most affected are those centred on scheduling, document management, data entry, and routine correspondence — all tasks that AI handles efficiently at scale.

Where Finance Is Protected

Not all finance roles face equal risk. The data shows a clear split within the sector. Roles involving client relationships, regulatory judgement, and strategic decision-making retain strong protection. Roles involving routine processing, data reconciliation, and template-based analysis are highly exposed.

The BLS projects a 4% decline in bookkeeper employment through 2033. That modest-sounding figure masks a faster reality: the decline is accelerating as AI accounting tools mature. Meanwhile, compliance officers are projected to grow, and financial managers face steady demand. The sector split maps directly to our scoring framework: process-based roles are being automated; judgement-based roles are not.

📊 Measured Displacement: What the Data Shows

Forecasts are one thing. Measured reality is another. Since ChatGPT’s launch in November 2022, we now have 33+ months of real labour market data to examine. The evidence shows a pattern: AI displacement is real but narrower than predicted, concentrated in digital-first roles, and often anticipatory — companies cutting ahead of AI capability, not in response to it.

The measured data reveals an important nuance: most AI-attributed layoffs so far are anticipatory. Companies are cutting roles in preparation for AI, not in response to proven AI performance. HBR found that 77% of AI layoffs appear anticipatory — preparing for what’s coming rather than reacting to what’s here.

This means the displacement numbers are likely to accelerate. Companies are currently at the “restructure ahead of time” phase. As AI tools mature and demonstrate consistent ROI, the reactive phase — replacing proven-inferior human performance with AI — will follow. For RED zone roles, both phases point in the same direction.

For workers in RED zone roles, the anticipatory pattern creates an uncomfortable reality: you may lose your role before AI can fully perform it, because your employer believes it will be able to soon enough. The 30% of companies that have already replaced workers with AI, and the 59% of hiring managers who admit AI is used as cover for unrelated layoffs, illustrate the messy reality of displacement.

The Displacement Timeline

The Challenger data tracks cumulative AI-attributed layoffs since 2023. The number has grown each year, and January 2026 saw the highest monthly job cuts since January 2009. While AI is not the sole driver, it is increasingly cited as a factor in restructuring decisions across technology, finance, and media companies.

Oxford Economics estimates that AI is currently cited in approximately 4.5% of all job losses. That number sounds modest, but it represents a baseline that is growing quarter by quarter. The trajectory matters more than the current number — and the trajectory is clearly upward.

For RED zone roles specifically, the displacement risk is higher than the economy-wide average. The overall 4.5% figure includes healthcare workers, tradespeople, and other structurally protected roles that drag the average down. Within digital-first sectors like marketing, finance, and administration, the AI-attributed displacement rate is significantly higher.

💻 Freelance Impact: The Early Warning Signal

Freelancers are the canary in the coal mine for AI displacement. They don’t have employment protections, institutional inertia, or organisational friction to slow the transition. When AI can do the work, the gig disappears immediately. Harvard, Upwork, and Ramp all show the same thing: freelance marketplaces are where AI displacement shows up first and hardest.

The freelance data is the clearest real-world evidence of AI displacement. Writing gigs dropped 30% after ChatGPT launched. Software development gigs fell 21%. Graphic design work declined 17%. These aren’t projections — they’re measured changes in platform activity from Harvard researchers tracking millions of gigs.

Ramp data confirms the trend from the employer side: freelance marketplace spending as a share of total company spend has collapsed since AI tools became available. Companies that previously hired freelancers for writing, design, and development are now using AI tools directly. Entry-level project share on Upwork has fallen from 15% to below 9%.

What the Freelance Data Predicts for Employed Roles

Freelance platforms act as a leading indicator for the employed workforce because they remove the buffers that slow displacement in traditional employment. When a company needs a blog post, it can now use ChatGPT instead of hiring a freelance writer. The gig disappears instantly. For employed writers, the same pressure exists but plays out over months and quarters as employers restructure teams and reassign responsibilities.

The category-level data is instructive. Writing dropped 30%, software development 21%, and graphic design 17%. The ranking maps almost exactly to our RED zone rankings for the corresponding employed roles. Content writers, junior developers, and production designers all sit in the RED or YELLOW Urgent zones in our database. The freelance data validates our scoring framework with real-world market behaviour.

Ramp’s data adds the employer perspective: companies are actively redirecting budget from freelance marketplaces to AI tools. The freelance spend share has collapsed. This is not a temporary dip — it represents a permanent shift in how companies source routine content, code, and design work. For freelancers in these categories, the displacement is not a forecast. It is the current market reality.

🎓 The Entry-Level Squeeze

If AI is going to replace anyone first, it’s junior workers. Entry-level roles involve the most structured, repeatable tasks with the least institutional knowledge requirements. Stanford, Harvard, and Indeed all show measurable declines in entry-level postings since 2022. The data is consistent across sources.

The entry-level data is unambiguous. Stanford measures a 16% employment decline in AI-exposed entry roles. Harvard tracks a 7.7% decline in junior positions across affected fields. Metaintro shows entry-level postings have fallen 29 percentage points since January 2024. Indeed reports that entry-level listings have dropped from 16% to 10% of all postings.

The entry-level squeeze is particularly significant for understanding AI displacement risk because junior roles are the pipeline for senior roles. If companies stop hiring entry-level workers because AI handles the tasks, the entire career ladder breaks. Workers who would have started in data entry, progressed to analysis, and moved into management never get the first rung.

Anthropic CEO Dario Amodei warns that 50% of entry-level white-collar roles could be eliminated within 1–5 years. Whether that specific timeline holds, the direction is confirmed by every data source tracking the entry-level market. For career starters in digital-first, administrative, or analytical roles, the competition is no longer just other graduates — it’s AI.

The Career Ladder Problem

Entry-level displacement has a compounding effect that goes beyond the immediate job losses. Senior roles draw their pipeline from junior positions. If companies stop hiring entry-level accountants, they eventually face a shortage of experienced accountants. If junior analyst positions disappear, there are no mid-level analysts five years from now.

This creates a paradox: companies cutting junior roles for short-term AI savings may face talent shortages for senior roles in the medium term. Some organisations are already recognising this — redesigning entry-level positions around AI supervision and exception handling rather than eliminating them entirely. But many are not.

💵 Salary Impact: Who Earns Less, Who Earns More

AI displacement does not affect all salary bands equally. The pattern is counterintuitive: mid-range white-collar salaries are most at risk, not the lowest-paid roles. Physical, low-wage jobs (trades, care work) have strong AI defences. High-skill knowledge work with creative or strategic components retains value. The squeeze is in the middle — routine office work that pays $35K-$65K.

The salary data reveals a split market. Workers who can use AI tools effectively earn more — PwC reports a 26% wage premium for AI-skilled workers, and up to 56% in AI-enhanced roles. Workers whose roles are replaced by AI see the opposite: their labour market value decreases as the supply of candidates exceeds the shrinking number of positions.

The pattern creates an hourglass effect: high-skill creative and strategic roles gain value from AI augmentation. Physical and trades roles retain stable wages due to structural protections. The middle band of routine office work — data entry, bookkeeping, basic analysis, customer service — faces both job losses and wage stagnation.

For workers in RED zone roles earning $35K–$65K, this is the critical insight: the salary band most at risk is the one that can least afford disruption. These are not high-earning knowledge workers with savings to retrain. They are mid-range administrative and clerical workers for whom displacement means immediate financial pressure.

The Hourglass Economy

AI is reshaping the salary distribution into an hourglass. At the top: high-skill strategic and creative roles that use AI as a multiplier earn more than ever. At the bottom: physical and hands-on roles that AI cannot perform retain stable demand and wages. The middle is being hollowed out — routine office work that lives entirely in software.

The $35K–$65K band represents the core of the at-risk workforce: administrative assistants, junior analysts, bookkeepers, customer service representatives, and data processing workers. These roles are large in employment volume, concentrated in the RED and YELLOW Urgent zones, and face the most immediate displacement pressure. Workers in this band who can either move up (adding strategic skills) or move sideways (adding physical or licensed skills) improve their position significantly.

🏢 What Employers Are Doing Now

Companies aren’t waiting for AI to mature before acting. Survey data from 2025-2026 shows employers are already cutting headcount, restructuring roles, and redirecting budgets — often based on AI’s potential, not its proven performance. The displacement is as much about employer sentiment as it is about technology.

The employer data tells a story of rapid action. Companies are not waiting for AI to prove itself before restructuring. They’re cutting headcount based on projected AI capability — and the majority of employers surveyed plan further cuts in 2026.

The Forrester finding is notable: 55% of companies that made AI-driven workforce cuts now regret the decision. This suggests the first wave of displacement was hasty — driven more by AI hype than by measured performance. But the regret is unlikely to reverse the trend. Subsequent rounds of cuts will be more targeted, better measured, and harder to argue against.

The Regret Factor

55% of companies that made AI-driven workforce cuts now regret the decision (Forrester, 2026). This is a critical data point. It means the first wave of displacement was driven more by narrative than by performance. Companies cut roles because AI was expected to fill the gap, not because it already had.

For workers, this regret creates a temporary reprieve. Companies that moved too fast are now rehiring, restructuring, or pausing further cuts until AI proves its value. But this window is limited. The second wave of cuts — informed by actual AI deployment data rather than hype — will be more targeted and harder to reverse.

The practical implication: workers in RED zone roles who still have their jobs are in a grace period. The question is whether they use that time to reskill, pivot, or add structural protections to their career profile. The employer data suggests the next round of restructuring will be more measured, more effective, and more permanent.

🚦 What To Do If Your Job Is on the List

If your role is on the at-risk list, the data says you have a window — not an eternity. The roles most resistant to AI share traits you can develop: physical skills, licensing, human trust, and strategic judgement. The dividing line is training, not talent.

The reskilling data shows both the urgency and the opportunity. 60% of workers will need reskilling by 2027 (WEF). 120 million workers globally need retraining within 3 years. Yet most employees have received zero AI training. The gap between what’s needed and what’s happening is enormous.

Sector-Specific Pathways

The right reskilling path depends on your starting point. Workers in different at-risk sectors face different transition options based on their existing skills, location, and financial situation. Below are pathways grounded in the data:

The common thread across all effective transition paths: move from digital-only output to roles that involve physical presence, licensing, trust, or unpredictable judgement. The further your new role sits from pure software work, the more structural protection it has.

One final data point: IDC estimates $5.5 trillion in economic value is at risk from the AI skills gap. Companies need people who understand AI — not to build it, but to deploy, manage, and govern it responsibly. Workers in at-risk roles who add AI literacy to their existing domain knowledge become significantly more valuable, even if their original role is automated.

🛡️ Protected Alternatives: 1769 GREEN Zone Roles

Not every career path leads through the RED zone. The GREEN zone roles at the top of our index share structural barriers AI cannot overcome. Physical presence, regulatory licensing, human judgement under uncertainty, and interpersonal trust create layers of protection. Many of these roles are in critical shortage — they’re not just safe, they’re in growing demand.

🇺🇸 56.2M US workers (33%) sit in GREEN zone roles. These positions score above 67 on the JobZone Score, meaning AI faces significant structural barriers to performing their core tasks. The top 20 most protected roles — the inverse of the at-risk list — are shown below.

The BLS data reinforces the pattern: the sectors most resistant to AI are also the ones projecting the strongest growth. Healthcare and construction — both structurally protected by physical presence and licensing — lead US job growth projections through 2033. For workers in at-risk roles looking for alternatives, these sectors offer both AI safety and job availability.

Healthcare dominates the protected category for good reason. Every healthcare role requires at least two of the three core barriers: physical presence with patients, regulatory licensing, and trust-based human relationships. A nurse must be at the bedside, licensed to practice, and trusted by the patient. AI can assist with diagnostics, scheduling, and documentation — but it cannot replace the human delivering care.

Trades and physical roles are structurally protected for the simplest reason: the work happens in unpredictable physical environments that AI and robotics cannot navigate. An electrician rewires a building from the 1940s with non-standard layouts. A plumber fixes pipes in a cramped crawl space. A construction worker builds on a site that changes every day. These are problems AI can model but cannot solve without a human body.

Cybersecurity is a special case: it is a digital-first sector that nonetheless scores well on our index. The reason is adversarial unpredictability. Cyber threats evolve constantly, attack patterns are novel by design, and the consequences of failure are severe. AI helps with threat detection and log analysis, but the strategic response — containment, forensics, communication — requires human judgement under uncertainty. ISC2 reports a 4.8 million global workforce gap in cybersecurity, making it one of the strongest alternative sectors for at-risk workers.

The Four Structural Barriers

Roles that resist AI displacement share four structural barriers. The more barriers a role has, the higher it scores. GREEN zone roles typically have three or four:

For the full list of protected roles, see Jobs That AI Cannot Replace and Most AI-Proof Jobs.

📝 All 200 RED Zone Roles

For reference, here is every role in our database that scores below 33 on the JobZone Score. Each links to its full assessment with individual dimension scores, task analysis, and AI capability mapping.

Every role scoring below 33 on the JobZone Score, sorted by score (most vulnerable first). Search all 3649 roles →

Each role links to its full assessment page where you can see the individual dimension scores, task-by-task AI capability analysis, and a detailed breakdown of why the role received its score. The assessment pages also show related roles in the same domain and specialism, so you can compare your position against similar ones.

Note that our database is continuously expanding. We add new roles as we complete assessments, which means the RED zone count, percentage breakdowns, and domain averages update automatically. The data on this page reflects the current state of the database at the time of your visit.

✅ The Bottom Line

AI displacement is real, measurable, and concentrated. It is not evenly spread across the economy. It targets digital-first, pattern-based, unregulated work — and the roles that match that profile are already feeling the pressure. The question is not whether these jobs will change, but how quickly — and whether the workers in them can pivot before the window closes.

Key Takeaways

For the inverse view — the jobs most resistant to AI — see Jobs That AI Cannot Replace. For the timeline angle — which jobs AI will replace soonest — see What Jobs Will AI Replace First? For salary-optimised safe careers, see High Paying AI-Proof Jobs. For the full picture — all zone breakdowns, expert positions, and evidence — see Will AI Replace Humans?