Editorial disclosure
This article reflects the independent analysis and professional opinion of the author, informed by published research, public statements, and industry data. No company reviewed or influenced this content prior to publication. All quotes are sourced and attributed. Where predictions are speculative, they are labeled as such.
On March 13, 2026, Bill McDermott went on CNBC and said something that should have stopped every parent, every college senior, and every career counselor in their tracks.
Unemployment for new college graduates, the CEO of ServiceNow predicted, "could easily go into the mid-30s in the next couple of years."
Not 10%. Not 15%. The mid-30s. A six-fold increase from today's 5.7%.
"So much of the work is going to be done by agents," McDermott continued. "So it's going to be challenging for young people to differentiate themselves in the corporate environment."
Here is the part that should make you uncomfortable: McDermott's company partnered with Anthropic on January 28, 2026 to sell AI agents to enterprises. He is predicting mass unemployment for new graduates while his company sells the tools that cause it. He is simultaneously the fire and the fire alarm.
Is his 35% number hyperbolic? Probably. But the direction is not. And he is far from the only voice saying it.
The Chorus of CEOs Saying the Quiet Part Out Loud
Dario Amodei, the CEO of Anthropic -- the company that builds Claude, one of the most capable AI systems on earth -- predicted in May 2025 that AI could automate 50% of all entry-level white-collar jobs within one to five years. At Davos in January 2026, he reiterated the warning, calling the coming disruption "unusually painful." Unemployment, he said, could reach 10-20%.
Now here is the twist that should keep you up at night. Anthropic's own CPO, Mike Krieger (co-founder of Instagram), admits they rarely hire fresh graduates. Anthropic has no internship program. The company building the AI that displaces entry-level workers does not hire entry-level workers itself.
Sit with that for a moment.
Two Anthropic researchers, Sholto Douglas and Trenton Bricken, went further. In June 2025, they predicted a "pretty terrible decade" -- a period of "job losses without immediate quality-of-life improvements from other AI advances." The current algorithms, they said, are "sufficient to automate white-collar work" with enough data. We are not waiting for a breakthrough. The technology is here. The deployment is what is catching up.
And Anthropic's CEO says 90% of the company's own code is now AI-generated.
Jensen Huang, CEO of NVIDIA, told an audience at the Milken Institute that "every job will be affected, and immediately." He also said something that landed like a grenade in every computer science department: kids should not learn to code. "Nothing would give me more joy than if none of them are coding at all."
He is wrong about the prescription -- understanding code still matters enormously -- but he is revealing about the direction. The CEO of the company whose GPUs power every AI system on earth is telling you that the skill that defined an entire generation of tech careers is becoming a commodity.
He did offer one useful line: "You're not going to lose your job to an AI, but you're going to lose your job to someone who uses AI." Remember that. We will come back to it.
Mark Zuckerberg told Joe Rogan in January 2025 that Meta would have an AI that could "effectively be a sort of mid-level engineer" by the end of that year. Mid-level engineers at Meta earn mid-six-figure salaries. That is not a research prediction. That is a cost-cutting target announced on the most popular podcast in the world.
(He later walked it back. Less than five months later, he revised the timeline, saying coding agents that replace most human programmers are "no longer within reach" until mid-2026 at the earliest. The goal did not change. Only the date.)
Tobi Lutke, CEO of Shopify, leaked a memo that has become the new hiring standard in Silicon Valley: "Before asking for more Headcount and resources, teams must demonstrate why they cannot get what they want done using AI."
That is not a suggestion. It is a gate. Every new hire at Shopify must now clear a bar that did not exist two years ago: prove that an AI agent cannot do this job. Lutke said he has seen employees use AI to "get 100X the work done." If that is even half true, why would anyone hire ten people when one person with AI tools can match them?
Satya Nadella confirmed that AI now writes 20-30% of Microsoft's internal code. Microsoft will still grow headcount, he said -- but "that headcount we grow will grow with a lot more leverage than the headcount we had pre-AI." Translated from CEO-speak: fewer humans, more output per human.
Sam Altman, CEO of OpenAI, said in January 2025 that customer support jobs would be "totally, totally gone." He predicted that AI agents would "join the workforce" and materially change company output. He later acknowledged that some companies are "AI washing" -- using AI as an excuse for layoffs that would have happened anyway -- but maintained that "current jobs are going to get disrupted as AI can do more and more."
Geoffrey Hinton, the Turing Award winner often called the "Godfather of AI," was the most blunt of all. In a discussion with Senator Bernie Sanders, he said: "It seems very likely to a large number of people that we will get massive unemployment caused by AI."
Then he said the part that no sitting CEO will say: "Rich people are going to use AI to replace workers...it's going to create massive unemployment and a huge rise in profits. That is the capitalist system."
Hinton does not expect new jobs to come close to the number eliminated. He predicted that 2026 specifically would see AI "be able to replace many other jobs" beyond call centers.
The Numbers Behind the Rhetoric
Predictions from CEOs are one thing. Data is another. Here is what the data says, and it is worse than the headlines suggest.
Entry-level job postings have collapsed. Software development and data analysis roles for entry-level candidates have plummeted 67%. Not dipped. Not declined. Plummeted. In the UK, tech graduate roles fell 46% in 2024, with a further 53% drop projected by 2026.
The hiring rate for new graduates is in freefall. The Gusto New Grad Hiring Report found that the Class of 2025 was hired at a rate of 4.8% -- 44% lower than the Class of 2022. If you graduated in 2022, you entered one job market. If you graduate in 2026, you enter a fundamentally different one.
Underemployment has hit a five-year high. The Federal Reserve Bank of New York reports that 42.5% of recent graduates are underemployed -- working in jobs that do not require a college degree. That is the highest rate since the depths of the pandemic in 2020. And this is before widespread AI agent deployment. The agents are barely here yet, and the job market is already deteriorating.
Oxford Economics found that 85% of the rise in U.S. unemployment since mid-2023 stems from new graduates struggling to find entry-level roles. Not from mass layoffs of experienced workers. The pain is concentrated at the bottom of the ladder.
Big tech has cut entry-level hiring by half. Entry-level hiring at major tech companies dropped over 50% in three years. A Harvard study found that after late 2022, AI-adopting companies hired five fewer junior workers per quarter than they had before. That is not a rounding error. That is a structural shift.
"Entry-level" no longer means entry-level. Over 60% of positions labeled "entry-level" in software and IT now require three or more years of experience. The label is a lie. The door says "entry" but the lock requires a key you can only get by already being inside.
How It Is Actually Happening
The numbers are abstract until you see the mechanisms. Here is how companies are actually replacing entry-level workers with AI.
Klarna replaced 700 customer service agents with an AI chatbot in 2024. The bot handled 2.3 million conversations across 23 markets in 35 languages -- 66% of all customer service interactions. CEO Sebastian Siemiatkowski celebrated it publicly. Then customer complaints increased. Satisfaction dropped. He admitted: "We went too far." Klarna is now rehiring humans.
This is actually the scariest version of the story -- not the one where AI works perfectly, but the one where companies fire people for AI, the AI underperforms, and they hire back at lower wages with worse terms because the labor market has already cratered. The overshoot-then-correct cycle may define the next five years.
IBM CEO Arvind Krishna paused hiring for back-office roles in 2023, targeting roughly 7,800 positions for AI replacement over five years. Employment verification, employee transfers, the kind of mundane administrative work that has historically been a new graduate's first corporate job -- all slated for full automation.
Chegg, the homework help platform, lost 99% of its stock value -- from a $14.7 billion market cap to roughly $156 million -- after students switched to ChatGPT. It then laid off 45% of its workforce, citing "new realities of AI." An entire company built on helping students learn was destroyed by the technology students now use instead.
The consulting firms are gutting the base of their pyramids. KPMG cut entry-level hiring by 29%. Deloitte by 18%. EY by 11%. Starting salaries have been frozen for three consecutive years. McKinsey's internal AI tool Lilli and BCG's Deckster can now perform roughly 80% of junior analyst work -- the research, the data crunching, the slide generation that thousands of new hires used to do as their first real job. The classic consulting pyramid, where a mass of junior analysts supported a few senior partners, is losing its base.
Goldman Sachs CEO David Solomon demonstrated that AI can now draft 95% of an IPO prospectus in minutes -- a task that previously required a six-person team two weeks. "The last 5% now matters," he said, "because the rest is now a commodity."
Across all of tech in 2025, there were 783 layoff events affecting 245,953 people. In just the first three months of 2026: 166 layoffs, 55,775 people. And 55% of hiring managers expect more layoffs this year, with 44% citing AI as the top driver.
The Global Picture
The macro-level numbers are staggering.
Goldman Sachs estimates that 300 million full-time jobs globally are exposed to automation. McKinsey says generative AI could automate 60-70% of current work activities before 2030. The Anthropic Economic Index, released March 5, 2026, found that computer programmers face 74.5% observed AI exposure -- the highest of any profession measured. Customer service representatives: 70.1%. Data entry: 67.1%.
The World Economic Forum offers a counterpoint: by 2030, they project 170 million new jobs created and 92 million displaced -- a net gain of 78 million. That sounds reassuring until you realize the new jobs will not be in the same cities, the same industries, or for the same people as the ones that vanish. A displaced financial analyst in New York is not going to become a wind turbine technician in West Texas without years of retraining and a willingness to uproot their life.
Korn Ferry surveyed 1,670 global talent leaders and found that 37% of companies plan to replace entry-level roles with AI. For back-office roles, it is 58%. Only 11% of talent acquisition leaders believe their executives are prepared to manage this transition.
The ladder is being pulled up. And the people pulling it do not have a plan for what comes next.
The Uncomfortable Question Nobody Is Answering
Here is the question I cannot get out of my head:
If the people building AI agents do not hire entry-level workers, and the companies buying AI agents are cutting entry-level workers, who exactly is supposed to train the next generation?
Entry-level jobs have never just been about labor. They are the training ground. The place where a new petroleum engineer learns that the model and the well behave differently. Where a junior analyst learns that the spreadsheet is only as good as the assumptions behind it. Where a new consultant learns that the client's real problem is never the one in the brief.
Korn Ferry warned explicitly: cutting entry-level roles creates a "long-term leadership vacuum." These are not just cost line items. They are the pipeline for future managers, directors, VPs, and executives. Companies saving money today are borrowing against leadership they will desperately need in ten years.
The oil and gas industry knows this problem intimately. We call it the Great Crew Change -- the wave of retirements that hollowed out mid-career expertise because the industry stopped hiring during the downturn. We are about to repeat that mistake, industry-wide, at a global scale. Except this time, it is not a cyclical downturn causing it. It is a permanent technological shift.
The Oil and Gas Angle: Closer Than You Think
If you work in energy, you might be reading this thinking: this is a tech and finance problem. We are different. We have wells and rigs and field operations. AI cannot replace that.
You are half right. And the half you are wrong about matters.
SLB launched its Tela platform in November 2025 -- an agentic AI assistant for upstream energy. The agents interpret well logs, predict drilling issues, and optimize equipment. The SPE's own Journal of Petroleum Technology described these systems as capable of reasoning, adapting, and collaborating -- "like a junior engineer who understands the tools, knows when something looks wrong, and can explain the process."
Read that sentence again. Like a junior engineer. That is not my description. That is the Society of Petroleum Engineers'.
Production surveillance AI is already handling the majority of generative AI use cases in the industry -- artificial lift optimization, gas lift management, field production monitoring. The 7 AM well-check that used to be a new hire's first job? AI agents are doing it. The data cleaning and reformatting that occupied the first year of every petroleum engineer's career? The SPE calls it "grunt work" and says it can be delegated to agents.
ExxonMobil reduced well planning time from nine months to seven months with AI tools. That is two months of junior engineering work that evaporated.
But here is the moat. The Bureau of Labor Statistics projects petroleum engineering jobs to grow 2.4% by 2033, with only a 30% probability of automation overall. That is dramatically lower than the 74.5% exposure facing programmers or the 70% facing customer service. Why? Because you cannot send Claude to a wellsite. You cannot ask an AI agent to troubleshoot a stuck packer in a West Texas well at 2 AM. Physical infrastructure, field judgment, and the sheer complexity of subsurface uncertainty create barriers that pure software jobs do not have.
The SPE put it well: "The rise of AI does not signal the demise of petroleum engineering. It signals its transformation. AI will not eliminate petroleum engineers, but it will eliminate those unwilling to learn and evolve."
The transformation is real. SLB and Halliburton have both launched internal AI bootcamps. The AI in oil and gas market is projected to reach $7.5 billion by 2030, growing at 12.8% annually. The companies that figure out how to combine domain expertise with AI capability will dominate. The ones that do not will be acquired by ones that do.
The Honest Counter-Arguments
Before we pivot to strategy, let me be straight about the counter-arguments, because intellectual honesty matters more than clicks.
The Yale Budget Lab studied nearly three years of labor market data since ChatGPT's launch and found "no clear evidence that the U.S. workforce has undergone widespread displacement." Occupational mix has remained "strikingly stable." Historical precedent supports this -- computers did not transform office workflows for nearly a decade after public release. AI may follow the same slow-then-sudden pattern.
The AI productivity paradox is real. Thousands of CEOs admitted that AI has had no measurable impact on productivity yet. Only one in four AI projects delivers on promised ROI. Only 16% are scaled across the enterprise. Robert Solow's 1987 observation -- "You can see the computer age everywhere but in the productivity statistics" -- may be repeating.
Klarna's reversal is a cautionary tale for the AI-replaces-everything crowd. They fired 700, quality suffered, customers complained, and they rehired. The "AI layoff trap" may claim many companies that cut too fast and too deep.
At Davos in January 2026, few C-suite leaders agreed with Amodei's aggressive timeline. Practical implementation is much slower than theoretical capability. Goldman Sachs has called the impact "mild and short-lived." A LinkedIn executive pushed back on the imminent replacement narrative entirely.
These are fair points. I take them seriously.
But here is why they do not make me feel better:
The Yale study's own authors note that "widespread effects will take longer than 33 months to materialize." They are not saying it will not happen. They are saying it has not happened yet. The hiring freeze is already here -- 67% fewer entry-level software postings, 44% lower new grad hiring rates, 42.5% underemployment. The Anthropic Economic Index found "suggestive evidence that hiring of younger workers has slowed" even as overall unemployment remains low. The pain is real. It is just concentrated where it is hardest to see: among people who never got a job in the first place, rather than people who lost one.
The question is not whether AI will displace entry-level work. It is already doing so. The question is how fast, how far, and whether new opportunities emerge quickly enough to absorb the displaced. Betting your career on the optimistic timeline is a gamble. Having a strategy for the pessimistic one is just prudent.
So What Do You Actually Do?
This is where most articles about AI and jobs end with vague platitudes: "upskill," "be adaptable," "embrace lifelong learning." Those are not strategies. They are bumper stickers.
Here is what I would actually tell a new graduate sitting across from me today.
1. Become AI-Native, Not AI-Replaceable
The single most important career decision you can make in 2026 is whether you learn to direct AI tools or compete against them. There is no third option.
Tobi Lutke's question at Shopify is now the question every hiring manager is asking: "What would this look like if AI did it?" If the answer is "basically the same," you are replaceable. If the answer is "AI does the 80% and the human does the 20% that actually matters," you are the human they keep.
The engineer who can orchestrate AI agents -- telling them what to analyze, validating their outputs, connecting their results to business decisions -- is worth ten engineers who cannot. Josh Bersin calls this the "Superworker" -- someone who operates as an orchestrator of unlimited resources, combining human judgment with AI efficiency. The irony is that younger workers, who grew up with technology, are the most capable of becoming Superworkers. They are being cut first from a workforce they are best positioned to lead.
Concrete action: Pick an AI coding assistant (Claude, Cursor, Copilot) and use it on a real project. Not a tutorial. A real project with messy data and unclear requirements. Learn where the AI excels and where it fails. That knowledge is your edge.
2. Build Skills AI Cannot Replicate (Yet)
Not all skills are equally automatable. The Anthropic Economic Index shows a clear pattern: routine cognitive work is most exposed. Work that requires physical presence, relationship trust, ethical judgment, or cross-domain synthesis is least exposed.
Field judgment. You cannot send an AI agent to a wellsite. The engineer who can read a situation -- the vibration that does not match the data, the crew member who is not following procedure, the weather that is about to shut down operations -- has a skill that no model can replicate from a data center.
Client relationships and trust. Goldman Sachs' David Solomon is right that AI can draft 95% of an IPO prospectus. But the client does not hire Goldman for the prospectus. They hire Goldman for the relationship, the judgment, and the trust built over years of deals. The same is true in every professional services field.
Cross-domain synthesis. The ability to connect geology, economics, operations, and regulatory reality into a coherent decision -- this is where human engineers create value that AI cannot. AI can optimize within a domain. Humans integrate across domains.
Ethical judgment and safety decisions. When a well control situation develops, the decision to shut in or continue is not a probability calculation. It is a judgment call with lives at stake. AI can inform that decision. It cannot make it.
3. Stack Your Skills in a T-Shape (or M-Shape)
The T-shaped professional -- deep expertise in one domain, broad literacy across several -- has always been valuable. In the AI era, it is becoming essential.
The vertical bar: deep domain knowledge in petroleum engineering, geology, drilling operations, or production optimization. The kind of knowledge that takes years of field experience and cannot be compressed into a prompt.
The horizontal bar: AI and data literacy, business acumen, communication, project management. The ability to translate between the technical and the strategic.
Hybrid workers with this profile are paid 20-40% more than traditional counterparts and constitute 12% of all job openings. They do not face replacement by AI in the coming years. Why? Because AI automates narrow, pattern-based work. T-shaped professionals do the opposite: they connect disparate knowledge into novel solutions.
Some are now talking about the M-shaped professional -- deep expertise in multiple disparate fields. Code plus psychology plus design. Reservoir engineering plus data science plus business development. The rarer the combination, the harder it is to automate.
4. Build in Public
Credentials are becoming less differentiating. Every applicant has a degree. Increasingly, every applicant can use AI to produce polished work samples. What separates candidates now is evidence of doing.
Contribute to open-source projects. The petro-mcp project -- an open-source MCP server for petroleum engineering data -- is one example. Contributing to tools like this demonstrates domain knowledge, technical capability, and initiative in a single artifact that any hiring manager can review.
Publish analysis using public data. Take a public dataset -- production data from state regulatory commissions, EIA data, well completion records -- and produce an analysis that demonstrates both domain knowledge and analytical skill. Post it on LinkedIn. Write about what you found. Show your work.
Document your learning. The graduate who can show a portfolio of AI-augmented projects -- with honest discussion of where the AI helped and where human judgment was required -- stands out from the one who just lists coursework.
5. Target Companies Where AI Augments, Not Replaces
Not every company is racing to cut headcount. Some -- particularly mid-size operators, field-heavy service companies, and organizations in the early stages of digital transformation -- need humans more than ever. They need humans who can also work with AI.
Mid-size oil and gas operators (50-500 employees) are in a sweet spot. They are too small for fully automated AI systems but large enough to benefit from AI-augmented workflows. They need engineers who can do field work and build dashboards. Who can run a rod pump and interpret the anomaly detection model.
Field-heavy roles -- drilling, completions, production operations, well intervention -- remain stubbornly resistant to automation. The physical world is messy, unpredictable, and dangerous. It requires humans.
Use jobs.petropt.com to find roles that match your actual skill profile. The platform shows competition levels for open positions so you can focus your energy where you have the best odds, not just where the most postings are.
6. Move Fast
When fewer jobs exist, speed becomes a competitive advantage. The 24-hour application window matters more than ever. A posting that receives 500 applications in a week gives early applicants a structural advantage -- many recruiters review in order received and stop when they have enough candidates.
Petro-Jobs tracks competition levels and posting velocity so you can identify high-value, low-competition opportunities before they are flooded. In a market where applications per role have jumped 30% year-over-year while postings decline 15%, this kind of intelligence is not optional. It is survival.
The Industries That Are Still Hiring
It is worth knowing where the growth is, because it is not where most graduates are looking.
Healthcare and social assistance leads all projections. The Bureau of Labor Statistics projects 2.3 million new jobs by 2033 -- one-third of all projected growth. Entry-level healthcare roles (certified medical assistants, dental assistants, home health aides) remain stable because they require physical presence and human empathy.
Skilled trades -- electricians, HVAC technicians, welders, maintenance specialists -- face chronic labor shortages. You cannot automate a pipe fitting. Infrastructure spending is accelerating demand.
Clean energy and EVs are creating entirely new job categories. Solar installation, wind turbine maintenance, battery systems engineering, and EV infrastructure are growing faster than the workforce to fill them.
Construction is adopting digital twins, drones, and BIM but still needs people on site. The technology augments rather than replaces.
The pattern is clear: mission-driven work in the physical world is insulated in ways that pure knowledge work is not. That does not mean knowledge work is dead. It means the bar for knowledge workers just got dramatically higher.
What Companies Owe the Next Generation
I want to end with something that is not usually in career advice articles, because it is not directed at job seekers. It is directed at the companies doing the hiring -- and the not-hiring.
You cannot strip-mine the entry-level workforce for a decade of AI-driven cost savings and then wonder why you have no leadership pipeline in 2035. You cannot eliminate the training ground and then complain that candidates lack experience. You cannot require three years of experience for an entry-level role and then lament the skills gap.
The IMF noted in January 2026 that the skills being demanded are evolving faster than education systems can adapt. The WEF projects that 39% of existing skill sets will be transformed or obsolete by 2030. If companies eliminate the entry-level positions where those new skills are developed, they are not just cutting costs. They are cutting the future.
Bill Gates put it simply: AI is "improving at a rate that surprises me." He also noted that software developers and energy sector experts are among the safer professions. Not safe. Safer. The distinction matters.
The Bottom Line
Here is what I believe, having spent years at the intersection of AI and energy systems:
This is a genuinely difficult moment for new graduates. The difficulty is not exaggerated. The data supports it. The CEOs confirming it have every reason to know -- they are the ones making the decisions.
But difficulty creates opportunity for those who adapt. The petroleum engineer who understands both the drill bit and the algorithm will be the most valuable person in the room. The analyst who can direct AI agents while also reading a client's body language will outperform ten people who can only do one or the other. The graduate who builds in public, moves fast, and targets the right opportunities will find them -- even in this market.
The world is changing. The entry-level on-ramp is being redesigned while people are still trying to merge onto it. Some of the old paths are closing. But the people who understand what is happening -- who see the shift clearly and position themselves for it rather than against it -- will not just survive. They will define what comes next.
The tools are changing. The field is not going away. The question is whether you will be the one directing the tools or the one replaced by them.
Start now. Not tomorrow. Now.
For energy industry job opportunities matched to your skills, visit jobs.petropt.com. For open-source petroleum engineering AI tools, see petro-mcp on GitHub. For more on breaking into the industry, read our guides on breaking into oil and gas in 2026, the petroleum engineering skills gap, and the state of O&G hiring in 2026.
Find Your Next Role
AI-Matched Oil & Gas Jobs
Petro-Jobs uses AI to match your skills with upstream oil & gas positions. Upload your resume and get matched in minutes.
Related Articles
- State of Oil & Gas Hiring 2026 -- Data-driven salary benchmarks, hiring trends, and where the jobs are.
- Breaking Into Oil & Gas in 2026: An Honest Guide -- The complete career playbook for entering the industry.
- From Spreadsheets to AI Agents: Future-Proof Your PE Career -- A level-by-level technical roadmap for building the skills that matter.
- Petro-Jobs: AI Job Matcher for Entry-Level Petroleum Engineers -- How our AI-powered job matching tool works.