AI Hype vs Reality (2026): We Scored 3649 Jobs
We scored 3649 roles covering 🇺🇸 168.7M US workers. 26% of the US workforce — 44.3M people — are in roles genuinely at risk from AI. 33% are not. The hype is real for some jobs, and overblown for most.
Investors poured $202B into AI in 2025, a 75% year-on-year increase (Crunchbase). Microsoft, Alphabet, Amazon, and Meta committed $320B to AI infrastructure. Headlines predict millions of jobs erased within years. The World Economic Forum forecasts 92 million roles displaced by 2030 (WEF). That’s the hype side.
Then there’s the data. We scored every role against current AI capabilities across seven dimensions using the JobZone scoring framework. The average score is 45.1/100, meaning the typical role sits in the YELLOW zone, where AI transforms tasks but doesn’t replace the worker. The data tells a more specific story than the headlines.
This article breaks down exactly where the AI hype holds up, where it falls apart, and what the data means for your career. We cover 28 industries, the 20 safest and 20 most exposed roles, the gap between AI investment and measurable productivity returns, and a practical career action plan backed by the numbers.
📊 Is AI Hype Justified? What 3649 Jobs Show
The World Economic Forum’s 2025 Future of Jobs report predicts 92 million roles displaced by 2030, offset by 170 million new roles created, a net gain of 78 million (WEF). Goldman Sachs estimated 300 million jobs globally “exposed” to AI automation. These are macro forecasts. Our data tests them at the role level.
Across 3649 assessed roles covering 🇺🇸 168.7M US workers, 33% of the US workforce (56.2M people) are in GREEN zone roles where AI cannot perform the core work. 40% (68.1M) sit in YELLOW, where AI transforms tasks but doesn’t replace the worker. 26% (44.3M) are in RED zone roles, where AI can already handle a significant portion of core workflows.
In role terms, that’s 1769 roles (48%) in GREEN, 1364 (37%) in YELLOW, and 516 (14%) in RED.
The RED zone accounts for 26% of assessed workers. That’s significant. Real people in real jobs facing real displacement. But it is far from the universal disruption narrative. The majority of workers sit in roles that AI transforms or cannot touch at all.
🇺🇸 56.2M US workers · 1769 roles
AI cannot perform core work
🇺🇸 68.1M US workers · 1364 roles
AI transforms, doesn’t replace
🇺🇸 44.3M US workers · 516 roles
AI handles significant core tasks
📢 The Hype Claims
- • “92M jobs displaced by 2030” (WEF)
- • “300M jobs exposed globally” (Goldman Sachs, 2023)
- • $202B invested in AI in 2025 alone
- • $320B planned AI infrastructure by Big Tech
- • 37% of leaders plan to replace workers with AI by end of 2026
- • Headlines predicting mass unemployment
📉 What the Data Shows
- • 33% of 🇺🇸 US workers assessed are protected
- • 1769 of 3649 roles are in GREEN zone (48%)
- • Average score: 45.1/100
- • AI contributed to 4.5% of 2025 US job losses (~55,000 jobs, Challenger)
- • Yale Budget Lab: no widespread AI job displacement 2022-2025
- • Risk is targeted, not universal
Key Finding: The Hype Is Both Right and Wrong
The data shows AI displacement is real, but concentrated. 🇺🇸 44.3M US workers (26% of the assessed workforce) are in 516 roles that face genuine risk. That’s not trivial. But it’s not the sweeping “AI will replace everyone” narrative dominating headlines. 33% of US workers are in roles with structural barriers AI cannot overcome: physical presence, licensing, trust, and real-time judgement. The hype gets the direction right and the magnitude wrong.
What does the data tell us? The WEF’s “92M displaced” figure assumes AI capability continues accelerating at its current pace. Our role-level data shows why that prediction overshoots: it treats all jobs as software problems. 33% of 🇺🇸 US workers are in roles that require a human body, a legal licence, or a trust relationship that operates outside software entirely. No amount of AI progress changes the fact that a plumber needs hands, a surgeon needs a licence, and a therapist needs the patient to believe they’re heard. The hype collapses when it hits the physical world.
The 🇺🇸 US Workforce Reality Check
Predictions like “92 million displaced” or “300 million exposed” become less alarming when you see the actual breakdown. Our data covers 🇺🇸 170.5M US workers (100% of the total US workforce). Here’s how they distribute across zones:
33% of assessed US workers in roles AI cannot perform. These workers are in physical, licensed, or trust-based roles. Their jobs are safe on any foreseeable timeline.
40% in roles where AI changes the toolkit. The worker stays. The tasks shift. These workers need to adapt their skills, not change their career.
26% in roles where AI can handle significant core tasks. This is real displacement risk, concentrated in specific sectors and actionable.
The US workforce data adds context the hype lacks. 🇺🇸 44.3M US workers in at-risk roles is significant. These are real livelihoods. But it’s 26% of the assessed US workforce, not 100%. The other 73% (🇺🇸 124.4M US workers) are in roles where the hype narrative simply doesn’t match the data. The hype makes a 26% problem sound like a 100% problem. That’s the gap between hype and reality.
🏭 AI Hype vs Reality by Industry
The AI hype treats “jobs” as one category. The data tells a different story: risk varies dramatically by sector. Some industries match the hype narrative closely. Others don’t match it at all. The gap between the highest-scoring and lowest-scoring domains is 31.9 points, a gap that blanket predictions ignore entirely.
The table below shows every domain we’ve assessed, ranked by average JobZone Score. Higher scores mean more protection from AI. Lower scores mean the hype is closer to reality.
| Domain | Avg Score | GREEN / RED |
|---|---|---|
| Trades & Physical | 60.5 | 87% / 3% |
| Veterinary & Animal Care | 59.8 | 89% / 2% |
| Military | 57.6 | 75% / 0% |
| Healthcare | 57.5 | 78% / 6% |
| Sports & Recreation | 56.2 | 84% / 0% |
| AI | 56.0 | 72% / 3% |
| Social Services | 55.8 | 79% / 0% |
| Religious & Community | 54.4 | 87% / 3% |
| Public Safety | 53.0 | 69% / 4% |
| Utilities & Energy | 50.6 | 60% / 5% |
| Other | 50.5 | 61% / 6% |
| Education | 49.1 | 57% / 4% |
| Cybersecurity | 49.0 | 56% / 8% |
| Agriculture | 48.1 | 52% / 4% |
| Transportation | 46.4 | 53% / 13% |
| Engineering | 46.0 | 51% / 5% |
| Government & Public Admin | 42.4 | 43% / 16% |
| Retail & Service | 40.8 | 31% / 15% |
| Science & Research | 40.7 | 24% / 5% |
| Legal & Compliance | 39.7 | 31% / 19% |
| Library, Museum & Archives | 39.4 | 31% / 10% |
| Creative & Media | 37.2 | 33% / 26% |
| Development | 36.0 | 29% / 23% |
| Cloud & Infrastructure | 35.1 | 24% / 28% |
| Real Estate & Property | 34.5 | 14% / 17% |
| Manufacturing | 31.1 | 13% / 35% |
| Business & Operations | 29.6 | 11% / 36% |
| Data | 28.6 | 5% / 35% |
The pattern is clear: industries built on physical work, licensing, and human trust score highest. Industries built on digital workflows and routine text processing score lowest. Someone in Trades & Physical faces a fundamentally different reality than someone in Data.
Workforce Impact by Domain
Role counts only tell part of the story. A domain with 30 assessed roles might represent hundreds of thousands of workers. The cards below show the zone breakdown for the safest and most exposed domains, with workforce distribution visible in each bar.
Safest Domains (Hype Overstates Risk)
Most Exposed Domains (Hype Matches Reality)
Hype Matches Reality (Higher Risk)
- • Data — avg 28.6/100
- • Business & Operations — avg 29.6/100
- • Manufacturing — avg 31.1/100
- • Real Estate & Property — avg 34.5/100
- • Cloud & Infrastructure — avg 35.1/100
Digital-first, routine, text-based work. The hype is justified here.
Hype Overstates Risk (Protected)
- • Trades & Physical — avg 60.5/100
- • Veterinary & Animal Care — avg 59.8/100
- • Military — avg 57.6/100
- • Healthcare — avg 57.5/100
- • Sports & Recreation — avg 56.2/100
Physical presence, licensing, trust. The hype oversells risk here.
Insight: AI Risk Is Sector-Specific, Not Universal
The 31.9-point gap between the highest-scoring and lowest-scoring domains is the single most important number in this article. A blanket “AI will replace X% of jobs” prediction ignores this variation entirely. Any credible assessment of AI’s impact on employment must be industry-specific. The hype isn’t, and that’s why it misleads.
For the full role-level breakdown within any domain, click the domain name in the table above. Each domain page shows every assessed role with its score, zone, and individual assessment. The workforce stacked bar below shows the aggregate picture across all assessed domains.
Why Industries Split This Way
The domain-level data reveals a structural divide in the economy that the AI hype narrative ignores. The divide isn’t between “high-skill” and “low-skill” work. It’s between work that happens in the physical world and work that happens in software.
Physical-World Industries
Healthcare, construction, trades, emergency services, and manufacturing all score well above the database average. The common thread: the worker must be physically present, the environment is variable, and the work product is a physical outcome (a healed patient, a wired building, a manufactured component). AI cannot be physically present, cannot handle variable environments, and cannot produce physical outcomes.
These industries also face the most acute worker shortages. The same barriers that block AI also limit the supply of qualified humans.
Software-World Industries
Administration, finance operations, content production, data processing, and customer service score below average. The common thread: the worker sits at a computer, the environment is standardised, and the work product is digital (a report, a dataset, an email, a summary). AI operates natively in this environment. It processes text, analyses data, and generates digital output faster and at lower cost.
This doesn’t mean all roles in these domains are at risk. Senior analysts, strategists, and client-facing professionals still combine judgement and trust that AI can’t replicate.
This divide explains why the AI hype feels both right and wrong depending on who you ask. A software engineer sees AI changing everything. A nurse sees it changing nothing meaningful. Both are right about their own domain. The hype goes wrong when it generalises from one to the other.
🔴 Where the AI Hype Is Real
The hype isn’t wrong about everything. These 20 roles score lowest in our framework. AI can already perform a significant portion of their core tasks. The displacement risk is genuine, not speculative. The average score across these roles is 3.3/100.
| # | Role | Score |
|---|---|---|
| 1 | File Clerks (Mid-Level) | 1.5 /100 |
| 2 | Micro-Task Worker (Online) (Mid-Level) | 1.7 /100 |
| 3 | Data Entry Keyer (Mid-Level) | 2.3 /100 |
| 4 | Word Processor and Typist (Mid-Level) | 2.6 /100 |
| 5 | Vulnerability Tester / Scanner Operator (Entry-Level) | 2.7 /100 |
| 6 | Telephone Operator (Mid-Level) | 3.0 /100 |
| 7 | Virtual Assistant (Entry-to-Mid Level) | 3.2 /100 |
| 8 | Live Chat Support Agent (Entry-to-Mid Level) | 3.4 /100 |
| 9 | Telemarketer (Mid-Level) | 3.4 /100 |
| 10 | Medical Transcriptionist (Mid-Level) | 3.6 /100 |
| 11 | Toll Collector (Mid-Level) | 3.6 /100 |
| 12 | Machine Feeders and Offbearers (Mid-Level) | 3.6 /100 |
| 13 | Procurement Clerks (Mid-Level) | 3.6 /100 |
| 14 | Correspondence Clerk (Mid-Level) | 3.6 /100 |
| 15 | Desktop Publisher (Mid-Level) | 3.7 /100 |
| 16 | Office Machine Operator, Except Computer (Mid-Level) | 3.9 /100 |
| 17 | OnlyFans Chatter / Ghostwriter (Entry-to-Mid Level) | 4.0 /100 |
| 18 | Meter Reader (Mid-Level) | 4.1 /100 |
| 19 | Medical Scribe (Mid-Level) | 4.3 /100 |
| 20 | Insurance Claims and Policy Processing Clerk (Entry-to-Mid) | 4.4 /100 |
What do they share? Every role on this list operates in a digital-first environment where AI tools already handle core workflows. The work is routine, text-based, and rule-following. No licensing requirement protects them. No physical presence is needed. No trust relationship anchors them. The hype, for these specific roles, is justified.
Why These Roles Cluster Together
The bottom 20 roles aren’t random. They cluster around four shared traits that make them vulnerable to AI displacement. Understanding these traits matters more than memorising a list, because any role with these characteristics faces the same pressure, whether it’s on this list or not.
Shared Traits of High-Risk Roles
Digital-first environment
Work happens entirely on screen. No site visits, no patient contact, no physical materials. Everything the worker touches is data, text, or pixels, exactly the medium AI operates in natively.
Routine, pattern-based tasks
Rule-following that AI excels at. If the work can be described as “look at input X, apply rule Y, produce output Z,” AI can learn it. Consistency is the enemy. Machines love patterns.
No licensing requirement
No regulatory barrier to AI adoption. Nobody needs to change a law, pass an exam, or hold a credential. The AI can start doing the work tomorrow without permission from any authority.
Text or data output
The deliverable is exactly what AI produces best: reports, summaries, translations, data entry, correspondence, or analysis. No physical artefact. No human-to-human interaction as the product.
The pattern across these roles is clear: they’re not at risk because they’re low-skilled. Many require expertise, judgement, and years of training. They’re at risk because their work happens entirely in software, and that’s where AI is strongest. A senior data entry specialist and an experienced translator face the same structural problem: their output is digital, their process is repeatable, and no law prevents an AI from doing it.
If your role is on this list, the displacement risk is real. The timeline is years, not decades. But this list also shows the limits of hype: these are 20 roles out of 3649 assessed, not the universal job apocalypse the headlines suggest. See the full at-risk list →
The Hype Is Real Here
For these 20 roles, every data point supports the hype narrative. AI tools can already handle the core tasks. Employers are actively exploring automation. The scores are low because the barriers are absent. This is not speculation. It’s measurement. The question for workers in these roles isn’t whether AI will affect their work, but when and how to prepare.
🔍 Is the AI Hype Real for YOUR Job?
Statistics tell you about averages. Your career isn’t an average. Search below to see exactly where your role sits: the score, the zone, the specific tasks AI can and can’t perform, and what protective barriers exist. Every one of our 3649 assessments is individually researched, not generated from a template.
Each result links to a full assessment page with detailed scoring breakdown, AI capability analysis, and career recommendations.
Can’t find your exact title? Try a related term. Our database covers 3649 roles across 28 industries. For a complete search with filters by zone, domain, and specialism, use the full search tool.
What Each Assessment Includes
Every role assessment is individually researched and scored. When you click through to a role page, you’ll find:
JobZone Score
The overall 0-100 score measuring AI displacement risk
Zone Classification
GREEN (safe), YELLOW (transforming), or RED (at risk)
Task-by-Task Analysis
Which specific tasks AI can perform and which it cannot
Protective Barriers
Physical, licensing, trust, and judgement barriers identified
Career Recommendations
Specific guidance based on your role’s score and zone
Related Roles
Adjacent roles in similar domains with their scores
The assessments are the foundation behind every number in this article. When we say 26% of 🇺🇸 US workers are at risk, that’s based on 516 individually scored assessments covering 44.3M people, not a statistical model. When we say 33% are protected, that’s 1769 individually verified roles with identified structural barriers. The data is specific enough to be actionable for your career.
🛡️ Where the AI Hype Falls Apart
Media coverage implies AI will replace “everyone.” These 20 roles prove otherwise. They score highest in our framework because they combine multiple structural barriers that AI cannot overcome, not on any current timeline. The average score across these roles is 83.6/100.
What protects them? Physical presence, regulatory licensing, human judgement under uncertainty, and interpersonal trust. A plumber needs hands. A doctor needs a licence. A firefighter needs split-second decisions in chaos. A therapist needs the patient to believe they’re heard. These aren’t software problems. They’re human problems that happen in the physical world.
Why These Roles Are Structurally Protected
The top 20 roles cluster around four protective traits. These traits don’t just resist current AI. They resist any foreseeable AI advancement. Understanding why tells you more about AI’s limits than any headline prediction.
Protective Traits That Block AI
Physical presence required
The body is the barrier. No API replaces hands. A plumber fixes a pipe in a different position every time. A surgeon operates on a body that’s never identical to the last. The physical world resists standardisation, and AI needs standardisation.
Regulatory licensing
Legal frameworks that change at legislative speed, years to decades, not months. No jurisdiction licenses an AI to prescribe medication, sign off on electrical work, or fly a commercial aircraft. Even if AI could do these things, the law prevents it.
Judgement under uncertainty
Real-time decisions in unpredictable, high-stakes environments. A firefighter assessing a collapsing building. A detective reading a suspect. A paramedic triaging multiple casualties. These aren’t pattern-matching problems. They’re judgement calls in chaos.
Interpersonal trust
The human IS the service, and trust can’t be faked. People won’t accept AI therapy, AI teaching their children, or AI delivering a terminal diagnosis. The relationship itself is the deliverable, and relationships require a human on both sides.
How Barriers Stack
Roles with one barrier have moderate protection. Roles with two are strongly protected. Roles with three or four are effectively immune to AI displacement on any foreseeable timeline. Most of the top 20 roles have at least two. The highest-scoring roles have three or four. A registered nurse, for example, requires physical presence (patient contact), licensing (state board), trust (patient relationship), and judgement (clinical decisions under pressure). That’s four barriers. AI would need to overcome all four simultaneously to displace the role.
These roles aren’t niche outliers. They represent 33% of the assessed 🇺🇸 US workforce (56.2M US workers). Many face critical shortages, which means the same barriers that block AI also limit the supply of qualified humans. For workers in GREEN zone roles, the labour market dynamics are the best possible position: protected from automation AND in high demand.
The hype narrative requires you to believe AI will simultaneously solve robotics (physical presence), change legislation (licensing), develop consciousness (trust), and handle novel scenarios (judgement). The data says it hasn’t done any of these things. The gap between the bottom 20 roles (avg 3.3/100) and the top 20 roles (avg 83.6/100) is the distance between hype and reality. See the full protected list →
GREEN Zone: Where Hype Fails
- • Location: Must be physically present
- • Regulation: Licensed or certified
- • Tasks: Variable, contextual, judgement-based
- • Human element: Trust, empathy, physical skill IS the service
- • Demand trend: Growing, with persistent shortages globally
- • Avg score: 83.6/100 (top 20 roles)
RED Zone: Where Hype Holds
- • Location: Entirely screen-based
- • Regulation: No licensing required
- • Tasks: Repeatable, pattern-matching, rule-following
- • Human element: Output is data, text, or process
- • Demand trend: Shrinking as AI takes over tasks
- • Avg score: 3.3/100 (bottom 20 roles)
The contrast between these two profiles is the clearest rebuttal of blanket AI hype. Every trait that makes a role AI-resistant (physical presence, licensing, trust, variable conditions) is also a trait that creates persistent demand and above-average wages. Every trait that makes a role AI-exposed (digital-first, routine, unlicensed) is a trait that makes it easier to automate and puts downward pressure on wages. The data doesn’t just tell you which roles are safe. It tells you which direction to move.
⚖️ The Case For & Against the AI Hype
Before we reach a verdict, both sides deserve a fair hearing. The AI hype has genuine evidence behind it, and genuine evidence against it. The strongest position is to understand both before deciding where the truth lies.
🔴 The Case That the Hype Is Real
The Money Is Real
$202B invested in AI in 2025, a 75% year-on-year increase (Crunchbase). Microsoft, Alphabet, Amazon, and Meta committed $320B to AI infrastructure. OpenAI and Anthropic alone captured 14% of global venture investment. When this much capital flows in one direction, it creates its own gravity. Companies spending billions on AI infrastructure have economic incentive to find ways to use it, including workforce reduction.
Real Displacement Is Already Happening
AI contributed to approximately 55,000 US job losses in 2025, 4.5% of total layoffs (Challenger). That percentage is growing. Customer service, content writing, translation, and data entry roles are seeing measurable reductions. Our own data confirms it: 516 roles in the RED zone aren’t speculative. They’re already being automated. 37% of business leaders plan to replace workers with AI by end of 2026.
Capability Is Accelerating
GPT-4 to GPT-5 represented a meaningful jump in reasoning capability. Multimodal models now process text, images, audio, and video. Agentic AI systems can execute multi-step workflows autonomously. The capabilities that scored a role at 15/100 last year may push it to 8/100 next year. The direction matters, and the pace is steep.
Corporate Announcements Are Explicit
Major companies are publicly stating AI replacement plans. BT announced 55,000 job cuts with AI replacing 10,000 roles. Klarna reduced customer service headcount by 700 through AI agents. IBM’s CEO paused hiring for roles AI could fill. These aren’t predictions. They’re corporate strategies being executed.
🟢 The Case That the Hype Is Overblown
The Productivity Paradox
A February 2026 NBER study surveying nearly 6,000 senior executives found that 89% reported no measurable impact on firm-level productivity, despite widespread AI adoption (Fortune / NBER). $202B in investment hasn’t produced productivity gains that the people running companies can measure. If AI were truly replacing workers at scale, it would show up in productivity data. It hasn’t.
No Widespread Job Losses Detected
Yale’s Budget Lab studied US labour data from 2022-2025 and found no evidence of widespread AI-driven job losses (Yale Budget Lab). Unemployment remains low. Labour force participation is stable. If AI were replacing millions of workers, the macroeconomic data would show it. It doesn’t.
AI Tools Often Slow People Down
A METR study found that experienced open-source developers were 19% slower when using AI coding tools (METR). Harvard Business Review reported that 41% of workers encountered AI-generated “workslop” costing approximately 2 hours of rework per incident (HBR). The gap between AI’s demo capabilities and real-world reliability remains wide.
Enterprise Adoption Lags Hype
Only 1% of companies report reaching AI maturity (Deloitte State of AI 2026). Only 21% have mature AI governance. Fewer than 30% of CEOs are satisfied with GenAI ROI (Gartner). 71% of organisations are using GenAI in at least one function (McKinsey), but using a tool and replacing a worker are different things.
The Financial Model Is Unsustainable
OpenAI reported $12B in losses for Q3 2025, with projected cumulative losses of $115B by 2029 (Fortune). The leading AI company is losing money at massive scale. If the business model behind AI can’t sustain itself, the infrastructure behind the hype narrative has a structural weakness.
⚖️ The Verdict: What 3649 Roles Tell Us
Both sides have legitimate evidence. Our data resolves the tension by showing the answer is position-dependent. The hype is right about 🇺🇸 44.3M US workers (26%) in 516 roles where AI can already perform core tasks. The hype is wrong about 56.2M US workers (33%) in 1769 roles where structural barriers prevent displacement regardless of AI capability.
The 68.1M US workers in 1364 YELLOW zone roles (40%) are where the detail matters most. AI transforms these roles by changing how the work gets done, augmenting some tasks, and automating others. But the worker remains. The hype calls them “replaced.” The data calls them “transformed.” The distinction matters if it’s your career.
The AI hype gets the direction right (AI is changing work) and the magnitude wrong (it’s not replacing most workers). 🇺🇸 124.4M of 168.7M assessed US workers (73%) are in roles where AI augments or cannot touch the core work. That’s not a collapse. It’s a transition.
⚠️ The AI Productivity Paradox
If AI is as powerful as the hype claims, it should be showing up in productivity data. Is AI overrated? The evidence says yes, at least so far. $202B invested in 2025, yet the measurable returns tell a different story. This section presents six data points that collectively paint a picture of heavy investment with modest impact.
6,000 Executives: AI Had No Impact on Productivity
A February 2026 NBER study surveying nearly 6,000 senior executives found that 89% reported no measurable impact on firm-level productivity, despite widespread AI adoption. This is the largest executive survey on AI productivity to date, and the result is stark: the people running companies can’t measure a productivity gain from AI. They’re spending the money. They’re deploying the tools. The needle isn’t moving.
Source: Fortune / NBER (Feb 2026)
AI Tools Made Experienced Developers 19% Slower
A METR study found that experienced open-source developers were 19% slower when using AI coding tools. The overhead of reviewing, correcting, and integrating AI-generated code offset any speed gains. This counter-intuitive finding challenges the assumption that AI tools automatically improve productivity. For experienced workers who already know what they’re doing, AI can add friction rather than remove it.
Source: METR (Jul 2025)
41% of Workers Encountered AI “Workslop”
Harvard Business Review reported that 41% of workers encountered AI-generated “workslop”: low-quality output requiring approximately 2 hours of rework per incident. The term captures a real phenomenon: AI produces output that looks complete but contains errors, hallucinations, or misaligned content that a human must then fix. The rework cost often exceeds the time saved by using AI in the first place.
Source: HBR (Sep 2025)
<30% of CEOs Happy With AI ROI
Gartner’s 2025 Hype Cycle found that fewer than 30% of CEOs are satisfied with their returns on GenAI investments. The gap between investment and measurable returns is widening. Companies are spending more, expecting more, and getting less than promised. Gartner’s Hype Cycle itself placed GenAI in the “Trough of Disillusionment”, the phase where reality catches up with expectations.
Source: Gartner (Aug 2025)
Only 1% of Companies at AI Maturity
Deloitte’s State of AI 2026 report found that just 1% of organisations consider themselves AI-mature. Only 21% have mature AI governance. Only 20% report talent readiness. 71% are using GenAI in at least one function (McKinsey), but adoption and impact are different things. Using ChatGPT for meeting summaries is not the same as automating a workforce.
Source: Deloitte State of AI 2026
OpenAI: $12B Loss in Q3 2025
Even the leading AI company is losing money at scale. OpenAI reported $12B in losses for Q3 2025, with projected cumulative losses of $115B by 2029. The financial sustainability of the AI industry itself is an open question. The hype assumes AI companies will continue investing indefinitely. The balance sheets suggest the runway is finite, and the pressure to demonstrate ROI is growing.
Source: Fortune (Nov 2025)
50% of Organisations to Require “AI-Free” Skills Assessments
Gartner predicts that by 2026, 50% of organisations will require “AI-free” skills assessments during hiring. This backlash signal is significant: employers are realising that AI-assisted candidates may not have the skills their credentials suggest. The overcorrection is itself evidence that the initial hype overshot reality.
Key Finding: Massive Investment, Modest Returns
The pattern across these data points is consistent: AI adoption is widespread, but measurable productivity gains are elusive. $202B in investment hasn’t translated to productivity improvements that CEOs, researchers, or workers can reliably measure. The hype outpaces the ROI. This doesn’t mean AI is useless. It means the gap between what AI companies promise and what enterprises experience is wide. The paradox: everyone is using AI, nobody can prove it’s working at scale.
What does the productivity paradox mean for the hype-vs-reality question? It means the timeline matters enormously. AI may eventually deliver on the hype, but the current data says “not yet.” Workers in RED zone roles should prepare for displacement, but the idea that AI is already replacing workers en masse is contradicted by the productivity data. The hype is running ahead of the reality by years, possibly by a decade.
📈 Is AI Hype Dying?
Is AI hype dying? The short answer: no. But it’s changing shape. Gartner placed GenAI in the “Trough of Disillusionment” in its 2025 Hype Cycle (Gartner) , the phase where reality catches up with expectations. That’s not death. It’s maturation.
The funding tells one story: $202B invested in AI in 2025, up 75% year-over-year. The impact data tells another: AI contributed to just 4.5% of total US job losses in 2025, approximately 55,000 positions. That’s real displacement, but modest against predictions of millions. The gap between investment and impact is the defining feature of the current moment.
Yale’s Budget Lab studied US labour data from 2022-2025 and found no evidence of widespread AI-driven job losses (Yale Budget Lab). That doesn’t mean displacement isn’t happening. It means it’s concentrated in specific roles and sectors, not sweeping through the labour market. Our data confirms this: 516 of 3649 roles are in the RED zone, concentrated in digital-first industries. That’s targeted, not universal.
Signs the Hype Is Cooling
- • GenAI enters “Trough of Disillusionment” (Gartner 2025)
- • <30% of CEOs satisfied with GenAI ROI (Gartner)
- • Only 1% of companies at AI maturity (Deloitte 2026)
- • 50% of orgs will require “AI-free” skills assessments by 2026 (Gartner)
- • 41% of workers encountering workslop (HBR)
- • Experienced devs 19% slower with AI tools (METR)
- • 89% of executives report no measurable productivity impact (NBER)
Signs It’s Still Growing
- • $202B+ invested in 2025 (75% YoY increase)
- • $320B planned AI infrastructure (Big Tech)
- • 71% of organisations using GenAI, up from 65% in 2024 (McKinsey)
- • 37% of business leaders plan to replace workers with AI by end of 2026
- • OpenAI, Anthropic alone captured 14% of global venture investment
- • Multimodal and agentic AI capabilities accelerating
- • 516 roles in our data face genuine, measurable displacement risk
The pattern matches every major technology cycle: initial euphoria, over-investment, disappointment when returns don’t materialise immediately, then a longer, steadier phase of genuine adoption. The internet in 1999. Mobile in 2008. Cloud computing in 2015. Each followed the same arc. AI is following it now.
Understanding the Gartner Hype Cycle
Gartner’s Hype Cycle maps technology adoption through five phases. Understanding where AI sits in this cycle matters because it frames what comes next, not just where we are now.
Innovation Trigger
A breakthrough generates interest. For GenAI: ChatGPT’s launch in November 2022. 100 million users in two months. Media frenzy. “This changes everything.”
Peak of Inflated Expectations
Hype exceeds capability. Every company announces an “AI strategy.” Predictions of millions of jobs eliminated. $202B invested in a single year. Headlines predict mass unemployment within 2-3 years.
Trough of Disillusionment ← GenAI is here (Gartner 2025)
Reality catches up. ROI disappoints. Early adopters encounter workslop, integration challenges, and hallucinations. CEOs report no measurable productivity gains. The narrative shifts from “AI replaces everything” to “AI is harder than we thought.”
Slope of Enlightenment
Practical use cases emerge. Companies learn which problems AI actually solves. Displacement happens in specific roles (our RED zone) while other roles (our GREEN zone) are confirmed safe. Expectations align with capabilities.
Plateau of Productivity
Technology becomes mainstream and useful in ways that are measurable. Job displacement has occurred in the roles it was always going to affect. New roles have emerged. The “hype vs reality” question is settled by observable outcomes.
The Trough of Disillusionment is not failure. It’s recalibration. Every significant technology passes through it. The internet lost 78% of its market value between 2000-2002 before becoming the backbone of the modern economy. Cloud computing was dismissed as “just someone else’s computer” before becoming $600B+ in annual revenue. The question isn’t whether AI survives the trough (it will). The question is which specific predictions survive it. Our data suggests the targeted displacement predictions survive. The universal replacement predictions do not.
How Past Tech Hype Cycles Played Out
| Technology | Peak Hype | Trough |
|---|---|---|
| Internet/Dotcom | 1999-2000 | 2001-2003 |
| ATMs | 1970s-80s | Mid-1980s |
| Industrial Robots | 1980s | Late 1980s |
| Cloud Computing | 2011-2014 | 2015-2017 |
| GenAI | 2023-2024 | 2025-? (current) |
The historical pattern is instructive. Every technology hype cycle predicted sweeping job losses. Every one delivered targeted displacement in specific roles while creating or transforming others. The magnitude of the hype always exceeds the magnitude of the outcome. AI is following the same path: real impact in specific digital-first roles, negligible impact in physical, licensed, and trust-based work.
For the full breakdown of displacement data, see our AI and job loss statistics and AI job loss predictions. For the financial bubble perspective, see AI bubble analysis.
The Hype Isn’t Dying. It’s Maturing.
The data suggests we’re past peak hype and entering a phase of realistic assessment. AI investment is still growing, but scrutiny of actual returns is intensifying. Job displacement is real but targeted: 516 of 3649 roles in our data, concentrated in digital-first industries. The correction isn’t about whether AI works. It’s about recalibrating expectations for how fast and how broadly it replaces human work. The hype isn’t dying. It’s becoming specific. And specific is what careers need.
✅ What This Means for Your Career
The AI hype is both right and wrong, and which it is depends on your specific role. 🇺🇸 44.3M US workers (26%) are in 516 roles that face genuine displacement risk. 56.2M (33%) have structural protection that no amount of AI capability can overcome. The remaining 68.1M (40%) are transforming. AI changes how the work gets done, not whether a human does it.
The data points to three concrete actions, depending on where you sit.
If You’re in a GREEN Zone Role
Your role is structurally protected. The barriers that block AI — physical presence, licensing, trust, judgement — aren’t going away. Focus on mastering AI tools that augment your work. A nurse who uses AI for documentation is more productive, not more replaceable. The combination of human skill plus AI productivity is the most valuable position in the labour market.
If You’re in a YELLOW Zone Role
Your role is transforming. The tasks AI can do will shift to AI. The tasks that require human judgement, creativity, or relationship management will stay with you. The workers who thrive are those who understand which parts of their role are AI-proof and double down on those skills. Upskilling in AI tools makes you more valuable, not less.
If You’re in a RED Zone Role
The displacement risk is real. The timeline is years, not decades, and you have time to plan. The most effective transition path is toward roles with structural barriers. Consider the GREEN zone roles listed below. Many have faster entry routes than people assume: trade apprenticeships (earn while you learn), nursing certifications (12-18 months), cybersecurity credentials (3-6 months for CompTIA Security+).
Top 5 AI-Safe Roles Worth Exploring
These are the five highest-scoring roles in our database, with the strongest structural protection from AI displacement. Each links to a full assessment with detailed scoring.
| Role | Score |
|---|---|
| Electrical Power-Line Installer and Repairer (Mid-Level) | 91.6/100 |
| Signalling Tester In Charge / STIC (Mid-Level) | 87.7/100 |
| Model Alignment Researcher (Mid-Level) | 86.1/100 |
| AI Safety Researcher (Mid-Senior) | 85.2/100 |
| Foster Carer (Mid-Level) | 84.5/100 |
Top 5 Roles to Assess for Transition
If your role matches one of these, the data says the displacement risk is highest. This isn’t about panic. It’s about having the information to plan.
| Role | Score |
|---|---|
| File Clerks (Mid-Level) | 1.5/100 |
| Micro-Task Worker (Online) (Mid-Level) | 1.7/100 |
| Data Entry Keyer (Mid-Level) | 2.3/100 |
| Word Processor and Typist (Mid-Level) | 2.6/100 |
| Vulnerability Tester / Scanner Operator (Entry-Level) | 2.7/100 |
Concrete Steps
- 1. Assess your position. Search the database for your job title. Every assessment shows your score, your zone, the specific tasks AI can and can’t perform, and what protective barriers exist. Know where you stand before deciding what to do.
- 2. Understand the traits, not just the score. Roles protected by physical presence, licensing, or trust are structurally safe. Roles that are digital-first and routine are structurally exposed. The traits matter more than the specific job title. They tell you whether your next role will also be safe.
- 3. If you’re exposed, plan a transition. Don’t panic. Plan. Explore adjacent GREEN zone roles. A data entry clerk has transferable organisational skills for healthcare administration. A translator has communication skills for international education. The goal is to move toward roles with structural barriers.
- 4. Upskill strategically. If you’re in YELLOW or GREEN, learn to use AI tools in your field. The combination of domain expertise plus AI fluency is the most competitive position in the market. AI augments your value rather than threatening it.
- 5. Read the career guide. For a deeper analysis of what protects careers from disruption and specific transition pathways, read the AI-Proof Career Guide. It covers reskilling timelines, salary expectations, and entry routes for every major GREEN zone sector.
Transition Pathways Into AI-Safe Sectors
If the hype is real for your role, the next question is practical: how do you move toward structural protection? The GREEN zone sectors have faster entry routes than most people assume. Many don’t require a four-year degree.
Healthcare (1-4 years)
Fast entry: Licensed Practical Nurse (LPN) in 12-18 months. Certified Nursing Assistant (CNA) in 4-12 weeks. Medical assistant in 9-12 months. Home health aide in weeks. From office roles: Organisation, scheduling, patient communication, record management, and compliance skills all transfer directly. Healthcare administration is a natural bridge for workers with data management or customer service backgrounds.
Trades (6 months - 4 years)
Fast entry: Construction labourer (immediate). Solar installer (6-12 months). Apprenticeships: Electrician (4-5 years), plumber (4-5 years), HVAC (3-5 years). Earn while you learn. Apprentices are paid from day one. No student debt. Apprentice wages start at $15-20/hr and rise to $30-45/hr as a journeyman. Wind turbine technician and solar installer are the two fastest-growing occupations in the US economy (BLS).
Cybersecurity (3-12 months)
Fast entry: CompTIA Security+ certification (3-6 months study). Entry-level SOC analyst roles accept certifications without degrees. Pathway: Security+ → SOC Analyst → specialisation (cloud security, penetration testing, incident response). From non-IT: Analytical thinking, process documentation, and compliance experience transfer. StationX offers structured training for career changers.
Education (1-2 years for alternative routes)
Fast entry: Teaching assistant (immediate). Alternative certification (1-2 years while teaching). Substitute teaching (bachelor’s degree in any subject). Advantage: Industry experts are actively recruited (retired engineers teaching physics, former accountants teaching maths) because of STEM teacher shortages. UNESCO estimates the world needs 44 million additional teachers by 2030.
The Common Thread
Every GREEN zone pathway leads to a credential (licence, certification, or apprenticeship completion) that AI cannot hold. The credential is the legal barrier that prevents displacement regardless of AI capability. Earning one is the single most effective career protection investment you can make, and it’s measured in months, not decades. For the full analysis of transition pathways, salary data, and entry requirements, read the AI-Proof Career Guide.
The only way to cut through the hype is to check the data for your role. Search for your job title to see exactly where you stand. No fearmongering, no false reassurance. Just data and what it means for your career.
The AI Hype Reality Check
When evaluating whether AI hype applies to your career, ask four questions:
- Does your work require a human body at a specific location? (Physical barrier)
- Does your work require a licence, certification, or legal authority? (Regulatory barrier)
- Does your work depend on human relationships, empathy, or trust? (Psychological barrier)
- Does your output require novel judgement in unpredictable situations? (Cognitive barrier)
Two or more “yes” answers = the hype doesn’t apply to you. Zero = it does. Check your role →
📊 What 3649 Roles Tell Us About AI Hype
We scored 3649 roles covering 🇺🇸 168.7M US workers. 26% of the US workforce faces genuine risk from AI. 33% does not. 40% is being transformed. The AI hype gets the direction right (AI is changing work) and the magnitude wrong (it’s not replacing most workers).
The AI Hype Verdict in Numbers
The $202B invested in 2025 is real money with real consequences. But 89% of executives can’t measure a productivity gain. Yale’s Budget Lab finds no widespread job displacement. Gartner places GenAI in the Trough of Disillusionment. The financial model behind the leading AI company is losing $12B per quarter. The hype is running ahead of the reality by years.
The evidence points to a specific conclusion: AI displacement is real, targeted, and concentrated in digital-first roles with no structural barriers. It is not the universal job apocalypse the hype narrative claims. 🇺🇸 124.4M of 168.7M assessed US workers (73%) are in roles where AI augments or cannot touch the core work.
The 31.9-point gap between the safest and most exposed domains tells the whole story. The hype is a blunt instrument applied to a precise problem. AI displacement is sector-specific, role-specific, and trait-specific. It follows predictable patterns: digital-first work with no licensing, no physical presence, no trust relationship, and routine output is exposed. Everything else is not. That’s what 3649 individually assessed roles show.
For workers in RED zone roles, the hype is real enough to act on. The timeline is years, not decades. Transition pathways exist in healthcare, trades, cybersecurity, and education, all sectors with persistent shortages, above-average wages, and structural protection from AI.
For workers in GREEN and YELLOW zone roles, the hype is noise. AI changes your tools, not your employment. The competitive advantage goes to those who master AI as an augmentation tool within their structurally protected role. The combination of domain expertise plus AI fluency is the strongest position in the labour market.
The question “is AI hype real?” has a data-backed answer: it depends on your role. Search all 3649 assessed roles to find where yours sits. For deeper analysis, read the AI-Proof Career Guide.
For related data:
- • Jobs AI Cannot Replace — 1769 GREEN zone roles ranked by score
- • AI and Job Loss Statistics — displacement data and trends
- • Jobs Most at Risk From AI — 516 RED zone roles ranked
- • Will AI Replace Humans? — the full spectrum analysis
- • AI Job Loss Predictions — institutional forecasts vs data
- • AI Statistics — full data across 28 categories
This page is updated as new role assessments are added and external research data refreshes. The structural patterns (physical work, licensing, and trust as AI barriers) don’t change. The specific numbers and workforce figures are updated regularly. Bookmark this page and return to check for new data.
What's your AI risk score?
We're building a free tool that analyses your career against millions of data points and gives you a personal risk score with transition paths. We'll only build it if there's demand.
No spam. We'll only email you if we build it.
The AI-Proof Career Guide
We've found clear patterns in the data about what actually protects careers from disruption. We'll publish it free — but only if people want it.
No spam. We'll only email you if we write it.
About This Data
This article draws on two data sources: our JobZone scoring framework (3649 roles assessed across 28 domains) and externally-sourced research with full citations. All internal data is dynamic, queried at render time from our database of individually assessed roles.
The JobZone Score (0–100) measures each role against current AI capabilities across seven dimensions: task automation potential, physical requirements, licensing barriers, judgement complexity, trust requirements, AI growth correlation, and evidence of current AI capability. Scores below 25 are RED (high displacement risk), 25–47 are YELLOW (transforming), and 48+ are GREEN (protected).
Workforce estimates are based on BLS employment data, covering 🇺🇸 170.5M US workers (100% of the total US workforce). External research data points are hardcoded with source citations and represent point-in-time findings from the cited publications.
External sources cited: World Economic Forum (Future of Jobs 2025), NBER (CEO productivity study, Feb 2026), Yale Budget Lab (AI labor market impact), METR (developer productivity study), Harvard Business Review (workslop), Gartner (2025 Hype Cycle & Strategic Predictions 2026), Deloitte (State of AI 2026), McKinsey (State of AI), Crunchbase (AI funding trends), Fortune (OpenAI financials), Goldman Sachs, Challenger (job cut data), and CBS News.
Methodology note: The JobZone Score is a measurement tool, not a prediction. It measures each role against current AI capabilities as documented in research literature, product demonstrations, and enterprise deployment data. Scores represent the state of AI today, not a forecast of future capability. As AI capabilities change, scores are updated to reflect new evidence. The zone boundaries (GREEN 48+, YELLOW 25-47, RED below 25) are calibrated against observed real-world displacement patterns. For the full methodology including scoring dimensions, calibration process, and validation approach, see the methodology page.
About the Authors
Nathan House
AI and cybersecurity expert with 30 years of hands-on experience. Nathan founded StationX (500,000+ students) and built JobZone Risk to ensure people invest their career development in the right direction.
StationX HAL
Custom AI infrastructure built by Nathan House for StationX. HAL co-develops JobZone Risk end-to-end: the scoring methodology, the assessment pipeline, every role assessment, and the statistical analysis that powers these articles — all directed by Nathan.