Canaries In The AI Coal Mine
What are some early signs of AI’s impact on the workplace?
There has been a number of detailed studies on the effect of AI in the workplace.
Most notably from Yale and Stanford.
LLM adoption at work for over 18’s reached 46% by June/July 2025.
In the footer of this post there is a list of resources.
Empirical evidence has struggled to keep pace with technological advancement, leaving many fundamental questions unanswered.
There are five points which emerged over the last few months:
- The workplace is complex and implementing AI for Work is not a straight forward process. It is not a one-size-fits-all approach and understanding the flow of work is important.
- Tasks are not work. AI is good at performing tasks; even if the tasks are long and complex. Work is a sequence or orchestration of tasks. Work has noice, disruptions and signal detection in the noise is hard.
- Workers know what good AI looks like; MIT coined the term, Shadow AI Economy, where workers bring their own AI and use it extensively while official implementation flounder.
- There exists an Human / AI Collaboration spectrum which can act as a guide. Ranging from augmentation to automation.
- For tasks, the Decision Level (Problem-solving & diagnosis) needs to be separated from the Action Level (Implementation & execution). AI can take care of the decision level, while the human oversees and controls the action level
Workers aged 22 to 25 have experienced a 6% decline in employment from late 2022 to July 2025.
Now, only a few days ago, Stanford released a new study with six interesting findings on the early effects of AI in the labour market…
One
They found a substantial decline in employment for early-career workers in occupations most exposed to AI, such as software development and customer support.
Two
A economy-wide employment continues to grow, but employment growth for young workers has been stagnant.
Young workers suffer most because AI replaces “book knowledge” from schooling but not “tacit knowledge” from experience, especially in non-college jobs up to age 40.
Three
Entry-level employment has declined in applications of AI that automate work, with muted effects for those that augment it.
Four
The employment declines remain after conditioning on firm-time effects, with a 13% relative employment decline for young workers in the most exposed occupations.
Five
The labor market adjustments are more visible in employment than in compensation.
Six
The patterns hold in occupations unaffected by remote work and across various alternative sample constructions.
More Detail on the Study
Generative AI has sparked intense debates, some view it as a productivity booster, pessimists fear job losses, and skeptics see little change.
Past technologies displaced some roles, improved others, and altered many, hinting at AI’s early effects.
AI has progressed swiftly —
A key worry: AI may replace young entry-level workers in fields like software and customer service.
Researchers examined ADP payroll data from millions of workers through July 2025 to track shifts. Findings show sharp drops for 22–25-year-olds in AI-exposed jobs like developers and service reps, while experienced workers or low-exposure roles like nursing aides held steady or grew.
Overall jobs rose, but young workers’ growth halted since late 2022, with 6% declines in vulnerable positions versus 6–9% gains for older ones.
Losses focused where AI automates tasks (replacing humans) rather than augments (assisting), per Claude AI data.
Declines persisted after adjusting for company-wide shocks like recessions.
Impacts hit job counts more than wages, with minimal salary differences by age or exposure. Results remained consistent excluding tech or remote jobs, surging from late 2022 with AI tool proliferation.
Insights indicate AI’s impact on entry-level roles, potentially evolving and not just from COVID education gaps.
Young workers suffer most because AI replaces “book knowledge” from schooling but not “tacit knowledge” from experience, especially in non-college jobs up to age 40.
Historically, technologies have affected different tasks, occupations, and industries in different ways, replacing work in some, augmenting others, and transforming still others.
The second key fact is that overall employment continues to grow robustly, but employment growth for young workers in particular has been stagnant since late 2022.
In jobs less exposed to AI, young workers have experienced comparable employment growth to older workers.
In contrast, workers aged 22 to 25 have experienced a 6% decline in employment from late 2022 to July 2025 in the most AI-exposed occupations, compared to a 6–9% increase for older workers.
These results suggest that declining employment in AI-exposed jobs is driving tepid overall employment growth for 22- to 25-year-olds as employment for older workers continues to grow.
Chief Evangelist @ Kore.ai | I’m passionate about exploring the intersection of AI and language. Language Models, AI Agents, Agentic Apps, Dev Frameworks & Data-Driven Tools shaping tomorrow.
