AI-Proof: Will AI Take My Job?

Inspiration

Like millions of workers worldwide, I've been watching AI capabilities explode—ChatGPT writes essays, Cursor codes entire codebaes, Nano Bana creates state of the art images. Every week brings another headline about AI disrupting jobs. But when I tried to figure out if my career was actually at risk, all I found was a vague advice about "staying adaptable."

I wanted real answers based on real data. Which specific jobs are vulnerable? Which are safe? What does the evidence actually say? how to transition my career? That frustration led me to build AI-Proof—a tool that gives people clarity instead of anxiety.

What it does

AI-Proof analyzes automation risk across 1100+ occupations using 10 years of employment data, academic research, and job market projections through 2034.

The key innovation is making complex labor market data conversational. Using Hex's AI capabilities, anyone can ask questions like "What are the safest jobs paying over $100k?" or "Should I transition from accounting to data analysis?" and get instant, evidence-based answers—no technical skills required.

Core features:

• AI Exposure Score (0-100) for each occupation based on automation probability, job growth trends, and wage vulnerability
• 12+ interactive visualizations showing risk patterns, salary correlations, industry breakdowns, and employment trajectories
• Natural language interface powered by Hex Thread and semantic modeling
• Dynamic filtering by salary range, risk level, growth rate, and job category
• Personalized recommendations suggesting safer career transitions based on your current role

How we built it

Data Foundation: I started by merging 14 datasets from the Bureau of Labor Statistics, O*NET occupation database, and Oxford's automation research. The biggest challenge was standardizing occupation codes across different classification systems (SOC, NAICS) and reconciling naming inconsistencies.

Risk Algorithm: The AI Exposure Score went through multiple iterations. My formula combines three factors:

$$\text{AI Risk Score} = (0.5 \times \text{Automation Probability}) + (0.3 \times \text{Job Decline Factor}) + (0.2 \times \text{Wage Vulnerability})$$

This captures that a declining, low-wage, highly automatable job is far riskier than an automatable role that's actually growing.

Hex Platform: I built the entire analysis in Hex, leveraging:

• Python notebooks for data processing and risk calculations
• Plotly for interactive visualizations
• Semantic modeling to define metrics like "safe high-paying jobs" that Hex Thread can query
• Input widgets (dropdowns, sliders, filters) for dynamic exploration
• Thread AI for natural language conversations with the data

Visualization Strategy: I created charts that tell different parts of the story—distribution histograms, scatter plots showing salary-risk tradeoffs, heatmaps by industry, timeline projections, and quadrant analyses showing growth vs. risk.

Challenges we ran into

Data consistency nightmare: Matching occupation names across datasets was brutal. "Software Developers, Applications" vs "Application Software Developer" vs "Software Developer" needed fuzzy matching and manual validation for high-employment roles.

The risk paradox: Early analysis showed high-risk jobs paying MORE on average. This seemed wrong until I realized: unstable jobs pay risk premiums. That counterintuitive insight became a key finding—safety and salary don't always correlate.

Hex learning curve: I'd never used Hex before this hackathon. Understanding how semantic models power Thread queries took experimentation. The documentation helped, but real learning came from breaking things and fixing them.

Balancing depth vs. accessibility: I kept adding "just one more analysis" until the dashboard felt overwhelming. Had to ruthlessly cut features and focus on core insights. Simplicity is harder than complexity.

Time pressure: With the deadline approaching, I had to choose between adding advanced features or polishing what existed. I chose polish—better to ship something excellent than something half-finished.

Accomplishments that we're proud of

Made data accessible: Non-technical people can now explore 1,100+ occupations and get personalized insights without writing code. That's powerful democratization of labor market intelligence.

Found surprising insights:

• Higher salary ≠ safer job (some $150k roles have 70%+ risk)
• Medium-risk jobs aren't "doomed"—they're evolving (software developers will use AI, not be replaced by it)
• Skilled trades paying $80k are often safer than white-collar jobs paying $120k

Built something genuinely useful: This isn't just a school project—people actually need this. Students choosing careers, workers considering pivots, policymakers planning retraining programs.

Learned an entire platform in days: Going from "What is Hex?" to building a production-quality dashboard with semantic modeling and AI features in 10 hours feels like a real accomplishment.

Created clarity from chaos: Turned 14 messy datasets and conflicting research into one coherent, interactive story about the future of work.

What we learned

Technical: Hex's semantic layer is genuinely transformative—it's not just natural language queries, it's encoding domain knowledge so AI can reason intelligently about your specific data. I also learned that great data products are about removing friction, not adding features.

Domain insights: The job market is way more nuanced than "AI will/won't replace X." Most occupations are transforming, not disappearing. The key question isn't "Is my job safe?" but "How will my job change, and am I ready?"

Data storytelling: I learned to lead with the human question ("Will I be okay?") before showing methodology. People don't care about your clustering algorithm—they care about their future.

Scope management: Perfect is the enemy of shipped. I had ideas for 20 more features, but delivering 12 polished charts beats 30 half-baked ones every time.

Personal growth: Building something thousands of people could use—going from my own career anxiety to creating a tool that helps others—taught me that the best projects solve problems you actually have.

What's next for AI-Proof: Will AI Take My Job?

Short term:

• Add skills gap analysis showing which specific skills to learn for career transitions
• Integrate real-time job posting data from LinkedIn/Indeed APIs to track market sentiment shifts
• Create shareable "Career Risk Reports" people can download as PDFs

Medium term:

• Geographic analysis showing automation vulnerability by state/city (some regions will be hit harder)
• Company-level risk scoring (is your employer in a vulnerable industry?)
• Historical case studies showing how past automation waves affected similar occupations

Long term:

• Partner with career counseling platforms and universities
• Build a recommendation engine matching people to retraining programs
• Create an API so other tools can access AI-Proof's risk scores

Vision: I want AI-Proof to become the trusted, evidence-based source people turn to for career decisions in the AI era. Not fear-mongering, not false reassurance—just clear, actionable data about the future of work.

The future isn't predetermined. With the right information, people can prepare, adapt, and thrive.

---

Built with Hex, Python, Plotly, and determination.
Data sources: U.S. Bureau of Labor Statistics, O*NET Database, Oxford/Frey & Osborne Automation Research

Built With

• hex