Source: arXiv:2506.00058 — "Prompt Engineer: Analyzing Skill Requirements in the AI Job Market" (LinkedIn job postings scraped April 7, 2025)
The takeaway: "Prompt engineer" was never really a job category—it's a transitional label, like "computer operator" in 1985. Eventually, we surmise, everyone will use AI the way everyone uses computers now. The jobs that are emerging are building the systems, not using them as consumers (which everyone will likely be).
The Split
Two Markets, Two Stories
Here's the twist: AI specialist roles (ML engineers, LLM engineers, AI researchers) are still growing. General software hiring is down significantly—the exact causes are debated (interest rates? AI productivity? post-ZIRP correction?), but the divergence is real.
The numbers: Q2 productivity rose 3.3% while hours worked grew just 1.1%. General software postings dropped 25% month-over-month in September. Meanwhile, ML engineer salaries hit $168K average. Many factors are at play, but the split highlights a trend.
The Adoption
88% of Companies Now Use AI
McKinsey's 2025 Global Survey shows AI adoption jumped from ~50% (where it hovered for years) to 88% in just two years.
The gap: Nearly everyone is using AI, but almost nobody has figured it out. That 1% maturity figure explains the hiring recalibration—companies realized they need fewer hires and more execution.
The Risk
Who's Actually Getting Displaced?
Stanford researchers analyzed millions of ADP payroll records—the largest real-time study of AI's labor impact. The finding that made headlines: young workers in AI-exposed jobs are seeing employment decline while older workers in the same roles are growing.
Source: Stanford Digital Economy Lab — "Canaries in the Coal Mine?" (Brynjolfsson, Chandar, Chen — ADP payroll data, Aug 2025)
The pattern: As Erik Brynjolfsson offered as a possible explanation: "Older workers have tacit knowledge—tricks of the trade learned from experience that may never be written down anywhere. They have knowledge that's not in the LLMs, so they're not being replaced by them." It could also be the data is incomplete, as older workers migrate to fractional work or retire.
Check Your AI Exposure
Based on Stanford's analysis of millions of payroll records. AI will touch every role—the question is how the transition plays out for yours.
Get personalized insights + monthly AI job market updates.
The Skills
What AI Roles Actually Require
The jobs are on the creator side—building, deploying, and maintaining AI systems. Machine learning engineer salaries jumped 53% in 15 months while general software engineer pay crept up only 4%.
75%+want domain specialists, not generalists
40%ML engineer posting growth (2024→2025)
Source: 365 Data Science — Analysis of 1,000+ AI engineer postings (2025)
Hiring insight: The roles that matter: MLOps, model deployment, LLM fine-tuning, production systems. "I can write good prompts" isn't on anyone's shortlist. The premium is on people who can ship.
The Bottom Line
Entry-level job postings dropped 35% from January 2023 to June 2025 (Revelio Labs). AI was a key contributor. But the story isn't "AI is taking all the jobs"—it's that the kind of work that matters is shifting.
Declining
Routine codified tasks Entry-level "book learning" roles "AI user" as a specialty
Growing
System builders & deployers Domain expertise + AI skills Tacit knowledge roles
Hiring AI talent? Looking for your next role?
Fluency Digital specializes in placing senior AI talent at companies building real systems—not hype. Let's talk about what the market actually looks like.
About this data: All statistics are from peer-reviewed research, published reports, or primary data sources linked throughout. The Stanford/ADP study represents the largest real-time analysis of AI's employment impact using actual payroll data from millions of workers. LinkedIn job posting data comes from academic research analyzing 20,000+ listings. We update this page as new data becomes available.
