If you are under 35, work in an office, or are counting on an entry-level job to launch your career — this week's data is not theoretical. It's about you. Goldman Sachs economist Elsie Peng published a U.S. Daily Note last week built on actual payroll records — not models, not surveys, not speculation — and found AI is eliminating approximately 16,000 net American jobs every single month. Right now. Not in 2030. Not when AI 'gets good enough.' Now. That same week, Stanford University released its 2026 AI Index — 423 pages of employment surveys, Gallup polling, and Pew Research cross-referenced by one of the most credible non-partisan research institutions in the country. Two independent data streams. The same week. The same conclusion. Source: Goldman Sachs U.S. Daily Note (economist Elsie Peng), April 6, 2026; Stanford HAI AI Index Report 2026, April 13, 2026.

What they found will directly affect you if you are under 35, work in an office, or are counting on an entry-level job to launch your career. That 16,000 per month compounds to 192,000 American jobs erased in the past year alone — from actual payroll records. Entry-level software developer employment among workers aged 22–25 is already down nearly 20% since 2024, with a separate Stanford Digital Economy Lab study tracking the same pattern since 2022 (when generative AI tools entered mainstream use). Gen Z excitement about AI collapsed from 36% to 22% in a single year. Gen Z anger rose from 22% to 31%. Only 10% of Americans — per Pew Research, March 2026 — say they are more excited than concerned about AI. And the United States recorded the lowest government AI regulation trust of any country surveyed by Stanford: 31%. Here is what both reports actually found, which occupations are at highest risk right now, and — specifically — what to do before the window closes. Source: Goldman Sachs U.S. Daily Note, April 6, 2026; Stanford HAI AI Index 2026, April 13, 2026; Stanford Digital Economy Lab, 2025; Pew Research, March 2026.

What Goldman Sachs Actually Found — And Why This Report Is Different

Goldman Sachs has published AI-and-jobs research before. In 2023, its economists projected that AI could eventually affect 300 million jobs globally and automate up to 25% of current work tasks. Those reports were widely covered and widely ignored because they described future risk, not current reality. The April 2026 Goldman report, authored by economist Elsie Peng, is different in a specific and important way: it measures what already happened over the past twelve months in actual payroll data — not what might happen over the next decade. The methodology separates AI's two competing employment effects. AI substitution — when AI replaces human workers outright, handling tasks those workers previously performed — destroyed approximately 25,000 gross jobs per month in the past year. AI augmentation — when AI makes existing workers more productive, potentially enabling companies to grow faster and hire more — added back approximately 9,000 jobs per month. The net: 16,000 U.S. jobs erased per month, every month, for the past year. That is approximately 192,000 net jobs per year. For reference, the U.S. economy typically needs to add 100,000 to 150,000 jobs per month just to absorb new workforce entrants. AI is now running in the opposite direction for a measurable slice of the workforce. Source: Goldman Sachs U.S. Daily Note, economist Elsie Peng, April 6, 2026.

The sectoral breakdown matters. Goldman's data shows AI substitution is hitting hardest in data entry, customer service, legal support, billing, content moderation, basic financial analysis, and entry-level software development. These are precisely the job categories that have historically been the on-ramps to careers for new college graduates, recent bootcamp completers, and young workers without decades of experience. They are also precisely the categories that AI tools — particularly agentic AI, which can execute multi-step tasks autonomously — are now best at handling. That is not a coincidence. It is a structural problem for a specific generation. Source: Goldman Sachs analysis, April 2026; TechRepublic analysis of Goldman findings, April 2026.

The Stanford AI Index 2026: What 423 Pages of Annual Data Shows About Where We Actually Are

Stanford University's Institute for Human-Centered Artificial Intelligence has published its AI Index annually since 2017. It is the closest thing the field has to a neutral annual audit — not an advocacy document, not a company report, not a think-tank brief with a funding agenda. The 2026 edition, released April 13 and running 423 pages, cross-references model performance benchmarks, global investment flows, employment and wage data, scientific output, policy and regulatory counts, public opinion polling across dozens of countries, and documented AI incidents from the past year. This year's findings arrive at a moment when two narratives are competing loudly in American public life: 'AI is creating a productivity boom that will ultimately create more jobs than it destroys' (the expert consensus position — Stanford documents that 73% of surveyed experts expect AI to have a positive impact on jobs) and 'AI is taking jobs, reducing wages, and making my career prospects worse' (the position held by a rapidly growing share of younger Americans — with only 23% of the general public sharing the expert optimism, a 50-point gap). What the Stanford data shows is that both groups are selectively correct — and that the timing of each effect is what neither side is honest about. Source: Stanford HAI AI Index Report 2026, April 13, 2026.

• AI adoption is accelerating faster than any technology in history. Generative AI reached 53% global population adoption within just three years — faster than the personal computer, faster than the internet, faster than smartphones. Important US-specific context: Stanford specifically notes that the United States ranks 24th globally in adoption at only 28.3% — meaning the global 53% figure, while accurate, overstates American adoption by nearly double. The US ranks well below many Asian and emerging market countries on generative AI use. The estimated consumer value of generative AI tools to Americans reached $172 billion annually by early 2026, with the median value per user tripling between 2025 and 2026. This is genuine and significant — but the US is not the most AI-enthusiastic nation, which matters for understanding where job displacement pressure is highest relative to adoption benefits. Source: Stanford HAI AI Index 2026, April 13, 2026.
• Employment impact is already measurable for younger workers — and the pattern is specific. Stanford confirmed that entry-level jobs in software development and customer support have been reduced, while mid-career and senior positions have held steady or increased. Among software developers specifically, employment among workers aged 22–25 fell nearly 20% since 2024. The same pattern appears in customer service, with entry-level roles disappearing while senior positions hold. These are not projections — they are current labor market measurements. Source: Stanford HAI AI Index 2026, April 13, 2026.
• The report also notes a nuance that complicates a simple AI-causes-unemployment story: unemployment is rising across many occupations, and workers least exposed to AI have seen unemployment rise more than workers most exposed to AI. This means the job market is broadly weakening, and AI's contribution is one force among several — including the tariff recession concerns, Federal Reserve policy, and post-pandemic structural shifts — affecting the labor market simultaneously. Attributing everything to AI overstates the case; attributing nothing to AI understates what Goldman's more granular analysis shows. Source: Stanford HAI AI Index 2026; IEEE Spectrum analysis, April 14, 2026.
• AI productivity gains are real and being captured — but not uniformly. Stanford's 2026 AI Index documents that AI is boosting productivity by 14% in customer service and 26% in software development, yet these gains are not appearing uniformly across tasks requiring judgment or ambiguity. The gains are concentrated in the productivity of existing experienced workers, not in entry-level replacement at the scale that would explain a full 192,000 annual job reduction on its own. This aligns precisely with Goldman's framework: augmentation benefits flow to mid-career workers; substitution risk concentrates at entry level. A third of organizations surveyed by McKinsey (cited in the Stanford Index) expect AI to shrink their workforce in the coming year, particularly in service and supply chain operations — the same entry-level-heavy categories already showing decline in the payroll data. Source: Stanford HAI AI Index 2026, April 13, 2026; SignalFire analysis (cited in Stanford HAI AI Index 2026), April 2026; McKinsey & Company, 2025 survey, cited in Stanford 2026 AI Index.

Why Gen Z Is First in Line — The Structural Problem Nobody Is Explaining Clearly

The Goldman Sachs and Stanford data both arrive at the same uncomfortable structural finding: AI is hitting the jobs that Gen Z was supposed to get first. Not the jobs Gen Z will eventually have — the jobs Gen Z needs right now, in their first two to five years of working life, to build the experience and credentials that lead to everything else. Understanding why requires being clear about what AI is currently best at, and matching that against where in the job market young people under 30 are concentrated. Gen Z workers disproportionately hold roles in data entry, customer service, legal support, billing, content moderation, basic financial analysis, and entry-level software development — the exact categories Goldman Sachs identifies as the hardest hit by AI substitution. This is not because Gen Z is less capable than previous generations at the same career stage. It is because that is structurally where early-career workers are. Entry-level roles, by definition, involve more routine, more repetition, and more clearly defined tasks than senior roles — which are precisely the tasks large language models and agentic AI are most effective at automating. Source: Goldman Sachs, April 2026; Stanford HAI AI Index 2026.

The Gallup data commissioned by the Walton Family Foundation and GSV Ventures, conducted in February and March 2026 across 1,572 Americans aged 14–29, makes the emotional reality concrete. The share of Gen Z respondents who describe themselves as excited about AI fell from 36% in 2025 to 22% in 2026 — a 14-point drop in a single year. The proportion feeling hopeful dropped from 27% to 18%. The proportion feeling angry rose from 22% to 31%. Gallup's senior education researcher Zach Hrynowski attributed the rising anger directly to AI dimming entry-level job prospects, and specifically noted that the oldest Gen Zers — those most exposed to the actual job market — are the angriest. The technology has not become worse. The benchmarks are impressive. The demos are compelling. What has changed is that the young Americans closest to the economic impact of AI deployment are no longer experiencing it primarily as a tool that helps them — they are experiencing it as a force that is competing with them for the jobs they were counting on to launch their careers. Source: Gallup/Walton Family Foundation/GSV Ventures survey, February–March 2026; cited in Stanford HAI AI Index 2026.

Profession-by-Profession: Who Is Most at Risk Right Now

The Stanford AI Index and Goldman Sachs data allow a reasonably clear current-risk picture by occupation category. Note that 'at risk' here means entry-level and junior positions, not career tracks overall. Senior roles in these same fields are generally not showing the same displacement pattern.

Sources: Stanford HAI AI Index 2026 (which measures the decline in entry-level software developer employment from 2024); Stanford Digital Economy Lab, 2025 study led by Erik Brynjolfsson (which tracked the same pattern from 2022, when generative AI tools entered mainstream use — both baselines confirm the same structural shift); Goldman Sachs analysis, April 2026; McKinsey Global Institute occupational analysis, referenced in Stanford Index; U.S. Bureau of Labor Statistics occupational data, 2024–2026.

The 73% vs. 23% Gap: Why AI Experts and Ordinary Americans Are Both Correct — And Why You Should Care Which Side You're On

The Stanford AI Index 2026 documents a striking divide: 56% of surveyed AI experts believe AI will have a positive impact on the United States over the next 20 years, and a separate expert survey within the Index found 73% expect AI to ultimately create more jobs than it destroys globally. Only 23% of the American public shares that broader optimism — a 50-point gap that Stanford documents is widening. The AI discourse tends to treat this as a communications problem — experts need to explain better, or the public is being irrationally fearful. Stanford's data suggests a different and more uncomfortable interpretation: they are both right, and they are describing different time horizons. The expert position — that AI will ultimately create more economic value than it destroys, and that productivity gains typically diffuse into broader employment over 10–20 years — is grounded in solid economic theory and the historical pattern of previous technological transitions. The industrial revolution destroyed specific jobs and created industries. Computerization did the same. There is strong reason to think AI will follow the same arc over a long enough time horizon. The public position — that AI is hitting entry-level workers now, that the promised 'future jobs' are abstractions while the current job losses are in the paycheck data — is also correct. Goldman Sachs counted 192,000 net job losses in the past year alone. Entry-level hiring at the top 15 U.S. tech companies fell 25% from 2023 to 2024, and the decline continued through 2025. The 23-year-old applying for jobs in 2026 is not wrong to feel the gap between 'AI will create new jobs' (future) and 'my job offers are down' (present). They are experiencing the leading edge of the transition, not the eventual equilibrium. Source: Stanford HAI AI Index 2026; Pew Research, March 2026; SignalFire analysis (tracking 650M LinkedIn profiles, cited in Stanford HAI AI Index 2026), April 2026.

The Gender Angle: Women Are Facing a Harder Hit Than Men

A separate finding from a joint Brookings Institution and Centre for the Governance of AI study, referenced in the Stanford 2026 Index, identifies a gender dimension that is getting relatively little coverage in mainstream AI reporting: women will be disproportionately harder hit by AI-related job displacement than men. The reason is structural rather than anything about capability or adaptability. Women are significantly overrepresented in administrative, clerical, and support roles — exactly the categories experiencing the highest current AI substitution rates. Legal secretaries, administrative assistants, billing coordinators, data entry specialists, and customer support roles are all female-majority occupations in the United States, and all are in Goldman Sachs's highest-risk categories. This does not mean women cannot adapt — the data on women in STEM and women-led AI adoption is positive in both directions. But it does mean the transitional pain described above is not evenly distributed, and policy responses that ignore the gender dimension of early AI displacement will miss where the harm is actually concentrated. Source: Brookings Institution / Centre for the Governance of AI study, cited in Stanford HAI AI Index 2026; TechRepublic analysis, April 2026.

What Is Actually Growing: The Jobs That Are Benefiting

The picture is not uniformly bleak, and presenting it that way would be as misleading as dismissing the displacement data entirely. Stanford's 2026 Index confirms several sectors where AI is expanding employment rather than contracting it, and these are worth being specific about because they represent real current opportunity — not projected future possibility.

• AI-specialized engineering and research roles — The demand for machine learning engineers, AI safety researchers, model trainers, and AI infrastructure engineers has expanded substantially. The US remains the world's leading source of AI researchers and developers by a wide margin. Entry into these roles typically requires graduate-level technical background or equivalent experience, which means they are not accessible to most current entry-level workers without significant reskilling. Source: Stanford HAI AI Index 2026.
• Mid-career and senior positions in AI-disrupted fields — Stanford's employment graphs show a consistent pattern: while entry-level headcount is falling, mid-career and senior positions in the same fields are holding steady or growing. This means the career ladder still exists — but the first rung is being pulled up. Workers who have already crossed into mid-career territory are experiencing AI as augmentation, not substitution. The implication for younger workers is that the years 2026–2028 may require more lateral moves, more reskilling, and more patience before the career path they expected becomes available again. Source: Stanford HAI AI Index 2026.
• Skilled trades — Electricians, plumbers, HVAC technicians, construction trades, and similar physically demanding, contextually complex jobs are showing virtually no AI displacement. Goldman Sachs's substitution data does not include physical trades in its high-risk categories for a simple reason: AI cannot wire a panel, diagnose a leak inside a wall, or replace an HVAC compressor on a commercial roof. The US already has a well-documented shortage of skilled trades workers. That shortage is deepening as fewer young people enter trades programs. This is one of the highest-certainty employment paths available to workers who want to avoid AI displacement risk entirely. Bureau of Labor Statistics projects 11% growth for electricians through 2033 and 6% for HVAC technicians — both faster than average. Source: Bureau of Labor Statistics Occupational Outlook Handbook, 2024–2026 edition.
• Healthcare (non-administrative) — Clinical roles from nursing to physical therapy to specialist medicine are facing zero meaningful AI substitution risk at the direct-care level. Administrative healthcare roles are more exposed. Stanford's data shows AI-related publications in the natural, physical, and life sciences all increased 26% to 28% year over year — meaning AI is accelerating research, but human clinical care demand is growing with an aging population. Source: Stanford HAI AI Index 2026; Bureau of Labor Statistics.
• AI-adjacent creative and strategic roles — Brand strategists, senior editors, UX researchers, and creative directors report the opposite of displacement: AI has made them more productive, enabling them to execute at a higher level and produce work that would previously have required larger teams. Stanford's 2026 AI Index confirms AI is boosting productivity by 14% in customer service and 26% in software development contexts — the same gains apply in creative and strategic roles where human judgment guides and AI executes. The key differentiator is that these roles require consistent contextual judgment that AI supports but cannot replace. Source: Stanford HAI AI Index 2026, April 13, 2026.

5 Survival Moves — What to Actually Do if You Are in a High-Risk Category

Generic 'learn AI skills' advice is everywhere and not particularly useful. What follows is specific to the risk categories identified above, grounded in what is actually changing in the labor market in 2026.

Move 1: Build the skills AI cannot replicate — client judgment, complexity management, and creative decision-making

AI is extremely capable at executing defined tasks. It is not capable of navigating ambiguous client relationships, managing conflicting stakeholder priorities, or making creative decisions in contexts where the 'right' answer is genuinely contested. These are the skills that distinguish mid-career workers from entry-level workers in every field — and Stanford's data confirms that mid-career roles are not seeing displacement at the same rate. The strategic priority for workers in high-risk entry-level categories is to accelerate the accumulation of judgment and complexity management experience as fast as possible — even if it means volunteering for harder, messier projects that do not fit the original job description. If your current role is mostly defined and repeatable, you are on the wrong side of the risk curve. Seek the adjacent projects where the answer is unclear. Source: occupational data analysis, Stanford HAI AI Index 2026.

Move 2: Become the person who operates the AI — not the person who competes with it

Goldman Sachs's augmentation data shows that AI is adding 9,000 jobs per month. Those are not new AI engineering jobs — they are existing roles that have been restructured around AI tool usage. The workers in those roles are operating AI, not competing with it. In customer service, that means moving from handling tickets to designing the AI system that handles tickets, training the AI on edge cases, and managing escalations the AI cannot resolve. In legal support, it means moving from document review to AI review management — supervising and quality-checking the AI's review outputs. In data entry, it means working as the data quality specialist who audits AI-processed records. The path is consistent: move up the stack from executing tasks to managing AI execution of those tasks. This is achievable without a PhD or a career change; it typically requires six to twelve months of deliberate reskilling within your existing field. Source: occupational restructuring patterns, Goldman Sachs analysis, April 2026.

Move 3: Build a skills portfolio that demonstrates AI proficiency in your specific domain

The Stanford AI Index reports that entry-level hiring at the top 15 tech companies fell 25% from 2023 to 2024 — but that number conceals a split. The drop was concentrated in roles that did not require demonstrated AI tool proficiency. Entry-level candidates who arrived with demonstrated ability to use Cursor for coding, Claude for document analysis, or domain-specific AI tools relevant to the role were notably more successful in competitive hiring processes according to multiple tech recruiting analyses. This is not about adding 'Prompt Engineering' to a LinkedIn headline — it is about being able to demonstrate in a portfolio or interview that you can complete the actual tasks the role requires significantly faster and better using AI. Show your work. Bring examples. Quantify the output. Source: SignalFire analysis (tracking 650M LinkedIn profiles, cited in Stanford HAI AI Index 2026); independent recruiter surveys, Q1 2026.

Move 4: If you are a student or recent graduate, seriously consider the trades

This recommendation will be uncomfortable for many readers, and it is worth addressing that discomfort directly. There is a cultural expectation in American education that a four-year college degree leads to a white-collar professional career. That path has enormous value — a college education is not a bad investment. But the specific on-ramp that path used to provide — entry-level white-collar work as the first professional step — is where AI displacement is concentrated hardest and fastest right now. Skilled trades offer a different path: median wages for electricians ($61,590/year), plumbers ($59,880/year), and HVAC technicians ($57,300/year) are not dramatically lower than entry-level professional salaries, the jobs cannot be automated at scale for the foreseeable future, and the shortage of workers means both job availability and wage growth are strong. Bureau of Labor Statistics projects 11% growth for electricians through 2033 and 6% for HVAC technicians — both faster than average. Apprenticeship programs are often paid. This is not a consolation prize. Source: Bureau of Labor Statistics Occupational Outlook Handbook, 2024–2026 edition.

Move 5: Stop treating AI as something happening to your career and start using it to accelerate the one you want

This is the hardest shift — psychological as much as practical. The Stanford data shows that 73% of AI experts believe AI will ultimately be positive for employment, and the historical pattern of technological transitions supports that view. The workers who come out ahead in technology transitions are almost always the ones who engage with the new technology directly and early, rather than the ones who resist it or wait for stability. Practically, this means: use the AI tools in your field aggressively, not defensively. If you are in customer service, build and share a documented workflow of how you use AI to handle fifty tickets in the time it used to take to handle ten. If you are in entry-level development, build a portfolio project using Claude Code or Cursor that demonstrates what one person can produce with AI assistance. The workers who become the 'AI operators' described in Move 2 all started by using the tools themselves, developing expertise, and making that expertise visible to employers and clients. Source: occupational adaptation analysis, HBR, March 2026; Goldman Sachs augmentation data, April 2026.

The Trust Collapse: Why Only 31% of Americans Trust Their Government to Handle This

One of the most striking — and least discussed — findings in the Stanford 2026 AI Index is the American public's trust in government to regulate AI. The US recorded the lowest government AI regulation trust of any country in Stanford's survey: 31%. Singapore ranked highest, at 81%. The UK, Germany, France, Japan, Canada, and Australia all substantially outrank the United States on this measure. The US public also reported the lowest expectation that AI would improve their jobs: only 33% of Americans expect AI to make their jobs better, compared to a global average of 40% and compared to China, where 83% of respondents believe AI products and services offer more benefits than drawbacks. This is a compounding policy problem. AI displacement is creating genuine harm for entry-level and younger workers. The workers experiencing that harm have extremely low confidence that the government will do anything useful about it. The most prominent policy response from AI's own leaders — Sam Altman's April 2026 call for a 'New Deal' including a 4-day workweek and wealth taxes on AI gains — was immediately criticized by policy experts as regulatory nihilism dressed as reform, according to Fortune's reporting. Whether effective policy materializes is genuinely uncertain. The trust data suggests American workers are not counting on it. Source: Stanford HAI AI Index 2026; Fortune, April 2026.

The US vs. China Dimension: Stanford's Most Alarming Chart

The 2026 Stanford AI Index contains a chart that is getting significant attention in policy circles and almost no attention in consumer coverage: the compression of the US–China AI performance gap. In late 2023, the best American AI models led Chinese models by 17.5 percentage points on MMLU, 24.3 percentage points on MATH, and 31.6 percentage points on HumanEval. By March 2026, the US leads China by a total of 2.7 percentage points on the benchmarks Stanford tracks — and the US and Chinese models have traded places at the top of performance rankings multiple times since early 2025. The Anthropic top model currently leads by that 2.7-point margin; before it, DeepSeek models briefly held the lead. This matters for American workers in AI-exposed fields for a specific reason: the competitive pressure on American AI companies from near-parity Chinese models is driving the same cost-efficiency push that is driving the substitution Goldman Sachs measured. When Cursor — a $29.3 billion company — runs on Chinese model Kimi K2.5 because it performs at similar quality for one-sixth the cost, that cost pressure flows through to every downstream decision about whether to hire a person or use the AI. The geopolitical AI race and the American job market are not separate stories. Source: Stanford HAI AI Index 2026; CNBC HumanX Conference reporting, April 11, 2026; MIT Technology Review, April 13, 2026.

The Honest Outlook for 2026–2027: What the Data Actually Supports

Responsible coverage of this topic requires separating what the data currently supports from what is extrapolation. Here is what is actually evidence-backed as of April 2026, and what is not.

The most important distinction the evidence supports: the disruption is targeted and just beginning, as MIT Technology Review summarized the Stanford Index findings. 'Targeted' means it is not hitting all workers equally — it is concentrated in entry-level, routine white-collar roles among younger workers. 'Just beginning' means the trend is measurable now, but the Goldman Sachs and Stanford data both contain firm survey evidence that executives expect it to accelerate through 2026 and 2027 as agentic AI deployment broadens. Preparing for that acceleration now — rather than waiting for more data — is the rational response. Source: MIT Technology Review analysis of Stanford HAI AI Index 2026, April 13, 2026.

What to Do This Week — A Concrete Action List

• Audit your job description against a current AI tool — today. Open ChatGPT, Claude, or Gemini. Paste in your job description (or the core tasks you did yesterday). Ask: 'What percentage of these tasks could an AI complete at acceptable quality right now?' If the honest answer is more than half, you are in a substitution-risk role and this is the week to start planning your transition — not because disaster is imminent, but because the Goldman Sachs data shows the displacement is already happening and executives expect it to accelerate. This is not cause for panic. It is cause for deliberate action this week rather than next year. Source: risk assessment framework, Goldman Sachs analysis, April 2026.
• Spend 30 minutes this week learning the AI tools in your specific field. Not general ChatGPT usage — field-specific. Legal professionals: Harvey AI or Casetext. Customer service: Intercom Fin or Zendesk AI. Developers: Cursor, Claude Code, GitHub Copilot. Financial analysts: Bloomberg AI tools, Notion AI for reporting. Marketing: Jasper, Claude for content strategy. This is the most direct path from 'AI competition risk' to 'AI operator advantage.'
• Update your resume and LinkedIn to reflect AI tool proficiency in your domain. This is now a standard screening criterion at many employers, and its absence signals a gap. Include specific tools, not vague 'AI familiarity' language. Be concrete: 'Used Claude and Cursor to reduce code review time by 60%' outperforms 'Familiar with AI tools' in every screening context.
• If you are a student deciding between programs: weight AI-exposure risk heavily. Computer science and data science degrees remain high-value with the caveat that entry-level roles are contracting — factor in time to specialization. Engineering, healthcare, and skilled trades are lower-risk career entries. The humanities and social sciences are facing both AI content competition and entry-level market contraction — if pursuing these, the path to employment needs to be extremely deliberate about differentiation.
• Read the Stanford AI Index. It is free, publicly available, and is the most comprehensive neutral annual data source on where AI is actually moving the labor market. The 2026 edition is at hai.stanford.edu. Reading the Executive Summary takes under an hour and gives you better data than 99% of the AI hot takes you will encounter this year.

The week that began April 13, 2026 delivered the most comprehensive simultaneous data drop on AI's actual impact on American workers that we have seen in the technology's history. Goldman Sachs measured the job losses from payroll records. Stanford measured the sentiment from surveys. A separate Stanford Digital Economy Lab study tracked the employment collapse in entry-level software development all the way back to 2022. The three data streams arrived at the same place from different directions: the disruption is real, it is concentrated in early-career and entry-level workers, it is accelerating, and it is hitting hardest in a generation that had every reason to expect better. The long-term outlook is not hopeless — economic history is on the side of the experts who believe productivity gains diffuse broadly over time. But the people absorbing the transitional cost right now are real people in their 20s who did everything right — went to college, learned the skills, took the entry-level jobs that were supposed to lead somewhere — and are now competing with software for positions that used to be theirs. That is not a prediction. It is what Goldman Sachs found in the payroll data last week. Acknowledging it is not pessimism. It is the precondition for doing anything useful about it — and the five moves above are that something useful. Sources: Goldman Sachs, April 2026; Stanford HAI AI Index 2026; Stanford Digital Economy Lab, 2025; Gallup/Walton Foundation, February–March 2026; Pew Research, March 2026; Bureau of Labor Statistics, 2024–2026.