A few days ago, I came across a research paper from Stanford that completely changed how I think about AI’s impact on our industry. The study, “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence” by Brynjolfsson, Chandar, and Chen, doesn’t just theorize about AI’s future impact – it analyzes real employment data from millions of workers to show what’s happening right now.
The findings hit particularly close to home for me. As I’ve been experimenting with AI tools in my verification workflow – using them to help generate testbenches, debug complex scenarios, and streamline documentation – I’ve been wondering about the broader implications for our industry. This study provides some concrete answers, and they’re not what I expected.
The Numbers Don’t Lie
The researchers analyzed millions of payroll records from ADP (the company that handles payroll for about half of American workers) alongside millions of conversations with Claude to understand how companies are actually implementing AI. What they found was striking:
Employment for software developers aged 22-25 has dropped nearly 20% since late 2022, while developers 26 and older have continued to see employment growth.
This isn’t limited to software development either. The pattern appears across customer service, data analysis, and other roles that are highly exposed to AI capabilities. But here’s the key insight that caught my attention: it’s not just about AI adoption but it’s about how companies choose to implement it.
The Critical Distinction: Automation vs. Augmentation
The Stanford team identified two fundamentally different approaches to AI implementation:
Automation: AI completely replaces human work. When companies use this approach, they see significant job losses, especially among younger workers.
Augmentation: AI enhances human capabilities rather than replacing them. Companies using this approach see much less employment disruption.
| Aspect | Automation | Augmentation |
|---|---|---|
| Definition | AI completely replaces human work | AI enhances human capabilities |
| Employment Impact | Significant job losses, especially for young workers | Minimal employment disruption |
| Human Role | Eliminated or drastically reduced | Enhanced and empowered |
| Implementation Goal | Replace workers to reduce costs | Make workers more productive |
| Risk Level | High for job displacement | Low for job displacement |
Real-World Outcomes (Based on Stanford Study)
| Metric | Automation | Augmentation |
|---|---|---|
| Junior Developer Employment | ~20% decrease since 2022 | Minimal impact |
| Senior Developer Employment | Variable impact | Continued growth |
| Productivity Gains | Short-term cost reduction | Long-term capability enhancement |
| Quality Outcomes | Inconsistent, lacks human judgment | Higher quality through human oversight |
| Innovation Potential | Limited to AI capabilities | Combines AI efficiency with human creativity |
Why Are Young Engineers Hit Hardest?
The age disparity in the data raises an important question: why are junior engineers more vulnerable? Based on my experience working with both junior and senior engineers, watching training programs, and observing our industry dynamics, I see several factors:
Lower replacement costs: From a purely economic perspective, replacing a junior engineer is less costly than replacing someone with years of domain expertise.
Simpler initial tasks: Junior engineers typically start with more straightforward tasks: bug fixes, basic features, routine maintenance. These are often the easiest for AI to handle.
Domain knowledge gap: Experienced engineers don’t just know how to write code or design circuits – they understand why certain approaches work and others don’t. AI still struggles with this contextual understanding.
In the verification world, I’ve seen this play out clearly. A junior engineer might be tasked with writing basic UVM testbenches, while a senior engineer is designing the overall verification strategy and handling complex corner cases. Guess which tasks are easier for AI to automate?
What This Means for Hardware Engineers
The implications for our field are significant. Hardware engineering has always been somewhat protected from automation due to the physical constraints and domain expertise required. But as AI tools become more sophisticated, we need to think strategically about our role.
The vulnerability areas: Basic scripting tasks, simple testbench development, routine debugging, and documentation writing are all areas where AI can potentially replace human effort entirely.
The protected areas: System-level thinking, complex problem-solving, understanding hardware-software interactions, and making trade-off decisions based on real-world constraints remain firmly in human territory.. for now.
| Task Category | Automation Approach | Augmentation Approach |
|---|---|---|
| Testbench Development | AI writes complete testbenches, replaces verification engineer | AI generates initial structure, engineer designs strategy and handles complex scenarios |
| Bug Analysis | AI automatically fixes all bugs without human review | AI identifies potential issues, engineer analyzes root causes and implements fixes |
| Documentation | AI generates all documentation, no human input | AI drafts initial docs, engineer reviews, validates, and adds domain expertise |
| Code Review | AI approves/rejects code changes autonomously | AI flags potential issues, human engineer makes final decisions |
| Verification Planning | AI creates complete verification plans | AI suggests test scenarios, engineer designs overall strategy |
My Recommendations: Adapt, Don’t Compete
Based on this research and my own experience, here’s what I think we should focus on:
For Junior Engineers
1. Embrace AI as a productivity multiplier Instead of trying to compete with AI at writing basic Verilog or Python scripts, learn to use it as a sophisticated assistant. I’ve found that AI can help me generate initial testbench structures much faster, but I still need to understand the underlying concepts to validate and refine the output.
2. Develop uniquely human skills Focus on areas where human judgment is irreplaceable: understanding customer requirements, making architectural decisions, debugging unexpected system behaviors, and communicating complex technical concepts to diverse stakeholders.
3. Become an AI power user Don’t just be a passive consumer of AI tools. Learn to craft effective prompts, validate AI-generated code thoroughly, and understand the limitations of different AI approaches.
For Engineering Managers
Think augmentation first, automation second
Before asking “How can AI replace this person’s work?” ask “How can AI make this person 2x more productive?” In verification, this might mean using AI to generate test cases while engineers focus on coverage analysis and complex scenario development.
Invest in upskilling your team The Stanford study suggests that companies using AI for augmentation see better outcomes. This requires training your team to work effectively with AI tools rather than simply replacing them.
The Broader Picture
What makes this study particularly compelling is its methodology. The researchers didn’t just rely on surveys or company statements about AI usage. They analyzed actual conversations with Claude to classify different types of AI implementation, providing unprecedented insight into how AI is actually being used in the workplace.
The data shows clear patterns: industries with higher AI exposure see more dramatic employment effects for young workers. Figure 3 in the study demonstrates this relationship convincingly across different sectors.
Looking Ahead
This research represents a wake-up call for our industry. The employment effects of AI aren’t some distant future concern… they’re measurable impacts happening right now. The key insight is that we have agency in how this plays out.
For hardware engineers, the message is clear: adapt your skills to work alongside AI rather than compete against it. Focus on the uniquely human aspects of engineering while leveraging AI to handle routine tasks more efficiently.
As I continue to experiment with AI tools in my own verification work, I’m constantly reminded that the technology is only as good as the human using it. The engineers who thrive in this new landscape will be those who learn to dance with AI, not those who try to ignore or compete against it.
What’s your experience with AI tools in hardware engineering? Are you seeing similar patterns in your workplace? I’d love to hear your thoughts and experiences in the comments below.
