The Map That Shook Silicon Valley
On March 15, 2026, Andrej Karpathy — former head of AI at Tesla, OpenAI co-founder, and one of the most respected voices in machine learning — quietly published an interactive map on his website. Within hours, it went viral. Within days, the GitHub repository was deleted. But the damage was done.
The map scored 342 occupations across the U.S. economy on a 0-to-10 scale of AI exposure. Not theoretical exposure. Not "someday maybe" exposure. Exposure based on what today's large language models can already do. The weighted average across all jobs? 4.9 out of 10. Nearly half the workforce sitting in the danger zone.
If you're a Hong Kong SME owner reading this, you might think: "That's America's problem." You'd be wrong. And by the time you realise it, your competitors will already have acted.
What Karpathy Actually Found
Karpathy scraped the Bureau of Labor Statistics Occupational Outlook Handbook and used an LLM to assess each role's automation potential. The results were displayed as a treemap — rectangles sized by total employment, coloured by AI exposure score.
The pattern was unmistakable: jobs that process, generate, or transform text, data, and code scored highest. Accountants, legal assistants, data analysts, marketing coordinators, customer service representatives — the backbone of every SME's back office.
42% of all jobs scored 7 or higher, representing 59.9 million workers and $3.7 trillion in annual wages. These aren't factory floor roles. These are the people sitting in your office right now, doing the work that keeps your business running.
The Hong Kong Translation
Hong Kong's economy is overwhelmingly service-based. Financial services, professional services, trading, logistics coordination — these are precisely the text-heavy, data-heavy roles that Karpathy's analysis flagged as most exposed.
According to the Deloitte-HKU AI Adoption Index 2026, only 8% of Hong Kong SMEs have reached "transformative" digital maturity. Meanwhile, 52% cite "lack of AI skills" as their primary bottleneck. The gap between knowing AI is coming and actually doing something about it is where businesses go to die.
Here's the uncomfortable reality: while your competitors in Shenzhen, Singapore, and Seoul are deploying AI staff at scale, only 27% of Hong Kong SMEs even plan to increase their AI spending. The city that built its reputation on being fast, agile, and commercially savvy is falling behind.
Why Your Back Office Is Ground Zero
Think about what your admin team does every day. They process invoices. They respond to customer enquiries. They compile reports. They schedule meetings. They reconcile spreadsheets. They draft emails. They update databases.
Every single one of those tasks scores 7+ on Karpathy's exposure scale. Not because the people doing them are bad at their jobs — but because these tasks are fundamentally about processing structured information, which is exactly what AI does best.
The question isn't whether AI can do this work. It already can. The question is whether you'll be the one deploying it, or whether your competitor across the street will do it first and undercut your margins by 30%.
The "But My Business Is Different" Trap
Every SME owner who's ever gone bankrupt said the same thing: "My business is different." It's the most expensive sentence in business history.
Karpathy's data doesn't care about your industry vertical. It cares about task composition. If your employees spend more than 50% of their time on text-based, data-based, or communication-based tasks, your business is exposed. Period.
A restaurant owner who thinks AI doesn't affect them? Their bookkeeping, supplier communications, staff scheduling, customer feedback management, and marketing are all high-exposure tasks. A property developer? Contract review, compliance documentation, market analysis, tender preparation — all scoring 7+.
The exposure isn't industry-specific. It's task-specific. And every business has these tasks.
What Smart Hong Kong Businesses Are Doing Right Now
The businesses that will survive this transition aren't waiting for a government grant or a consultant's recommendation. They're running a simple playbook:
Step 1: Audit your task exposure. List every role in your company. For each role, break down the daily tasks. Score each task honestly: could an AI handle this today? Not perfectly — but well enough to save 60-70% of the time?
Step 2: Start with the highest-exposure, lowest-risk roles. Customer service responses, data entry, report generation, email drafting — these are safe starting points where AI can deliver immediate ROI without disrupting critical operations.
Step 3: Deploy AI staff alongside human staff. This isn't about firing anyone. It's about giving your team AI colleagues that handle the repetitive work, freeing your humans to do what AI can't — build relationships, make judgment calls, and drive creative strategy.
The Clock Is Already Ticking
Gartner predicts that by the end of 2026, 40% of enterprise applications will feature task-specific AI agents, up from less than 5% in 2025. That's not a gradual shift. That's a cliff.
The Hong Kong government's 2026 budget has shifted from general relief to targeted high-tech transformation funding. The HKTDC and Microsoft are running AI adoption programmes specifically for SMEs. The infrastructure is being built. The incentives are being offered. The window is open.
But windows close. And when 40% of your competitors have AI agents handling their operations while you're still manually processing invoices, no amount of government funding will close that gap.
Your Move
Karpathy deleted his GitHub repository, but the data is already out there. The scores are already assigned. The exposure is already real. The only variable left is what you do about it.
You can wait and see. You can tell yourself your business is different. You can hope that AI is overhyped and this will all blow over. Or you can look at the map, find your industry, check your exposure score, and start building your AI workforce today.
懂AI，更懂你 — UD has spent 28 years helping Hong Kong businesses navigate technological shifts. This one is the biggest yet. And we're here to walk you through every step.
🤖 Ready to Check Your AI Exposure?
UD's AI Staff Solution helps Hong Kong SMEs deploy AI employees
for customer service, admin, HR, marketing, and more —
zero MPF, zero turnover risk, 24/7 operation.