'AI Won't Replace You. Someone Using AI Will.' -- Is This Actually True?
In 2023, Harvard Business School professor Karim Lakhani published an influential piece in the Harvard Business Review with a thesis that became a mantra: "AI won't replace humans -- but humans with AI will replace humans without AI."
Three years later, the phrase has saturated LinkedIn posts, conference keynotes, and corporate memos. It reframes AI as a tool and positions the worker as the one in control.
But is it accurate? Research from the Federal Reserve Bank of Dallas, Stanford University, and PwC tells a more complicated story -- one where the platitude holds for some workers and completely fails others.
The Dallas Fed finding: AI is doing both things at once
In February 2026, Dallas Fed economist J. Scott Davis published a study examining employment and wage trends in AI-exposed industries since ChatGPT's release in late 2022.
Total U.S. employment has increased about 2.5% since ChatGPT's debut, yet employment in the 10% of sectors most exposed to AI has declined by roughly 1%. Jobs in computer systems design and related services have fallen by 5%.
At the same time, wages in the most AI-exposed industries have grown 8.5% -- outpacing the national average of 7.5%. In computer systems design specifically, pay has risen 16.7%.
Fewer workers. Higher wages. AI is simultaneously eliminating jobs and making the remaining ones more valuable.
The explanation hinges on a distinction between codified and tacit knowledge. Codified knowledge -- textbook-derived, teachable information -- is what AI replicates effectively. Tacit knowledge -- judgment, intuition, the ability to navigate ambiguity -- is what it cannot.
The median experience premium across occupations is 40%, but it ranges from under 10% for fast food workers and ticket agents to over 100% for lawyers, insurance underwriters, and credit analysts. Occupations with higher experience premiums show a stronger positive relationship between AI exposure and wage growth.
AI substitutes for workers whose value comes from what they learned in school. It complements workers whose value comes from what they learned on the job.
Stanford's "canaries in the coal mine"
The Dallas Fed data is corroborated by a large-scale empirical study from Stanford's Digital Economy Lab. In August 2025, Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen analyzed individual-level payroll data from ADP -- the largest payroll provider in the U.S.
- Early-career workers (ages 22-25) in the most AI-exposed occupations experienced a 13% relative decline in employment, even after controlling for firm-level shocks
- Among young software developers specifically, the decline was nearly 20%
- Employment for experienced workers in the same roles remained stable or grew
The declines are concentrated in roles where AI automates rather than augments. The tasks being automated are precisely the ones that used to train new workers: summarizing information, drafting documents, cleaning data, writing routine code. The entry-level knowledge-work job is disappearing.
The broken career ladder
This is where the framing fractures.
In a functioning labor market, vacancy chains create upward mobility. A senior employee leaves, a mid-level worker moves up, a junior is hired to fill the gap. AI disrupts this chain at the bottom. Companies are not firing entry-level workers en masse. They are not hiring them.
Entry-level job postings in the U.S. declined approximately 35% from January 2023 to June 2025. A Stanford analysis found a 67% decrease in U.S. entry-level tech job postings between 2023 and 2024. In the UK, tech graduate roles fell 46% in 2024, with projections for a further 53% drop by 2026, according to the Institute of Student Employers.
The WEF's Future of Jobs Report 2025 projects a net gain of 78 million jobs by 2030 (170 million created, 92 million displaced). But these are not one-for-one exchanges. The jobs being created require different skills, in different industries, often in different geographies.
Even Anthropic CEO Dario Amodei -- whose company builds the AI models driving this shift -- warned in mid-2025 that AI could eliminate roughly 50% of entry-level white-collar jobs within five years, and that governments and AI companies need to stop "sugarcoating" the impact.
The upskilling paradox
The standard response to AI displacement is: upskill. Learn to use the tools. Become the "someone using AI" who replaces the someone who is not.
But the data on who gets to upskill undermines this.
Only 26% of organizations offer formal AI upskilling programs -- down from 35% a year earlier, per LinkedIn's 2026 Workplace Learning Report. AI training budgets were cut by an average of 18% in the second half of 2025, even as AI tool spending increased by 23%, according to Gartner research cited in the same report.
The access gap is stark. 72% of C-suite executives use AI daily, per BCG's AI at Work 2025 survey. Only 18% of individual contributors do. Workers at smaller companies or in slower-moving industries often lack any exposure to AI tools.
PwC's 2025 Global AI Jobs Barometer found that workers with AI skills command a 56% wage premium over peers without them. But the bottleneck is access, not motivation -- when employers offer training, 70% of workers complete it (WEF Future of Jobs Report 2025).
IDC projects that skills shortages will cost the global economy $5.5 trillion by 2026. The IMF warns that AI's growth benefits in advanced economies may be more than double those in low-income countries, where infrastructure and training access lag further.
What the platitude gets right -- and what it misses
"AI won't replace you -- someone using AI will" is accurate in a narrow, mechanical sense. PwC's data confirms it: wages are rising twice as quickly in AI-exposed industries, and productivity growth has nearly quadrupled in the most AI-exposed sectors since 2022.
But the statement carries an implicit assumption: that becoming "someone who uses AI" is a matter of individual choice. That the playing field is level.
The data says otherwise. Whether you become someone who uses AI depends on:
- Whether your employer invests in training (most do not, and the percentage is declining)
- Whether you are early or late in your career (AI augments the experienced and substitutes for the junior)
- Whether your industry is moving fast or slow (the gap between AI-forward and AI-lagging sectors is widening)
- Whether you are in an advanced economy or a developing one (only 26% of jobs in low-income countries are even exposed to AI, per the IMF)
- Whether entry-level roles still exist to give you the experience that eventually makes AI a complement rather than a competitor
The platitude puts the burden entirely on the individual worker. The research distributes it across employers, policymakers, and institutions.
The pipeline problem
There is a structural risk that even thoughtful analyses underweight.
If AI continues to eliminate entry-level positions at current rates, the pipeline of experienced workers thins within a decade. Today's senior engineers, analysts, and consultants gained their tacit knowledge by spending years in the junior roles now being automated. Without those roles, the next generation cannot accumulate the experience that makes AI a complement.
The experienced workers AI augments today were trained by the entry-level jobs AI is eliminating. If no one solves this, the experience premium that currently protects senior workers erodes as organizations lose the capacity to develop talent internally.
Some companies recognize this. IBM has tripled its entry-level hiring for 2026. But the short-term incentives -- lower headcount, higher margins, AI-driven productivity gains -- all point toward continued contraction.
Practical takeaways
For job seekers navigating this market, the macro trends matter less than the micro decisions.
1. Tacit knowledge is your moat. The Dallas Fed research is clear: AI substitutes for codified knowledge and complements tacit knowledge. Seek roles and projects that build judgment -- cross-functional experience, client-facing work, ambiguous problem-solving -- rather than tool proficiency alone.
2. Do not wait for employer-provided training. With only 26% of organizations offering formal AI programs and that number declining, self-directed learning is not optional. Free resources from Google, IBM, Harvard, and Stanford exist at every skill level.
3. Volume still matters in the job search. The labor market's structural shift means covering more ground is a necessity, not a luxury. With entry-level postings down 35%, efficiency in applications is a competitive advantage.
4. The platitude is half-right. AI will not replace most experienced workers soon. But for early-career workers and those without access to AI training, the threat is not a future hypothetical -- it is the current job market.
The honest version of the platitude: "AI won't replace experienced workers who have access to AI tools and the institutional support to use them effectively. Everyone else should be concerned."
Less catchy. More accurate.
Nox is an AI agent that finds and applies for jobs on your behalf -- discovering roles, scoring them against your preferences, and submitting tailored applications through ATS platforms. If the job market demands volume and speed, Nox handles both.