John Burn-Murdochdigs into one of the strangest paradoxes in tech today: 𝐀𝐈 𝐢𝐬 𝐬𝐮𝐩𝐞𝐫 𝐜𝐚𝐩𝐚𝐛𝐥𝐞 — 𝐲𝐞𝐭 𝐰𝐨𝐫𝐤𝐞𝐫𝐬 𝐬𝐞𝐞𝐦 𝐮𝐧𝐭𝐨𝐮𝐜𝐡𝐞𝐝.
GPT-style models pass elite exams and write essays indistinguishable from postgraduate work. But when it comes to job losses? Most roles are surprisingly stable — even in sectors we assumed would be first in line for disruption. So… 𝐰𝐡𝐚𝐭’𝐬 𝐠𝐨𝐢𝐧𝐠 𝐨𝐧?
𝐀 𝐧𝐞𝐰 𝐝𝐚𝐭𝐚-𝐝𝐫𝐢𝐯𝐞𝐧 𝐚𝐧𝐚𝐥𝐲𝐬𝐢𝐬 𝐬𝐡𝐞𝐝𝐬 𝐥𝐢𝐠𝐡𝐭. John Burn-Murdochanalyzed U.S. employment trends across roles deemed high-risk for automation (based on Brookings & OpenAI research). Here’s what he found:
- 𝐍𝐨 𝐜𝐫𝐚𝐬𝐡 𝐲𝐞𝐭: Roles like accounting clerks, legal secretaries, and underwriters? No major drops in employment — yet.
- 𝐁𝐮𝐭 𝐬𝐨𝐦𝐞 𝐞𝐱𝐜𝐞𝐩𝐭𝐢𝐨𝐧𝐬: Writers and software developers are showing clear signs of AI-related job decline — diverging from industry trends around them.
𝐖𝐡𝐲 𝐭𝐡𝐞 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐜𝐞? A new study by AI research lab METR gives us a fresh lens. Turns out, 𝐀𝐈 𝐞𝐱𝐜𝐞𝐥𝐬 𝐚𝐭 “𝐧𝐞𝐚𝐭, 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝” 𝐭𝐚𝐬𝐤𝐬 — not necessarily ones that are difficult or technical. That means jobs like writing articles or coding cleanly defined programs are far more automatable than you might think. On the flip side, even simple jobs like travel agents, executive assistants and bookkeeping clerks require juggling unpredictable goals, unclear inputs, human interaction, and multitasking. In other words: 𝐡𝐮𝐦𝐚𝐧 “𝐦𝐞𝐬𝐬𝐢𝐧𝐞𝐬𝐬” 𝐛𝐮𝐲𝐬 𝐭𝐢𝐦𝐞.
𝐀 𝐧𝐞𝐰 𝐥𝐨𝐠𝐢𝐜 𝐨𝐟 𝐯𝐮𝐥𝐧𝐞𝐫𝐚𝐛𝐢𝐥𝐢𝐭𝐲. It’s not about how smart the task is — but how structured it is. Ironically, Silicon Valley’s obsession with efficiency and predictability may have made some tech jobs 𝐞𝐚𝐬𝐢𝐞𝐫 𝐟𝐨𝐫 𝐀𝐈 𝐭𝐨 𝐫𝐞𝐩𝐥𝐚𝐜𝐞.
In particular, jobs that are: linear, repeatable and freelance/contract-based are most at risk — especially when AI tools can be slotted in without HR friction.
𝐀 𝐠𝐫𝐢𝐦 𝐟𝐮𝐭𝐮𝐫𝐞 𝐟𝐨𝐫 𝐝𝐚𝐭𝐚-𝐝𝐫𝐢𝐯𝐞𝐧 𝐜𝐨𝐥𝐮𝐦𝐧𝐢𝐬𝐭𝐬? If your job involves gathering data, writing up clean summaries, and staying within a fixed word count… well, AI is coming for you — fast.
𝐓𝐡𝐞 𝐛𝐢𝐠 𝐥𝐞𝐬𝐬𝐨𝐧? AI isn’t just about automating low-skill work. It’s about automating 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐰𝐨𝐫𝐤. And that flips many of our assumptions on their head.
John Burn-Murdoch digs into one of the strangest paradoxes in tech today: 𝐀𝐈 𝐢𝐬 𝐬𝐮𝐩𝐞𝐫 𝐜𝐚𝐩𝐚𝐛𝐥𝐞 — 𝐲𝐞𝐭 𝐰𝐨𝐫𝐤𝐞𝐫𝐬 𝐬𝐞𝐞𝐦 𝐮𝐧𝐭𝐨𝐮𝐜𝐡𝐞𝐝.
GPT-style models pass elite exams and write essays indistinguishable from postgraduate work. But when it comes to job losses? Most roles are surprisingly stable — even in sectors we assumed would be first in line for disruption. So… 𝐰𝐡𝐚𝐭’𝐬 𝐠𝐨𝐢𝐧𝐠 𝐨𝐧?
𝐀 𝐧𝐞𝐰 𝐝𝐚𝐭𝐚-𝐝𝐫𝐢𝐯𝐞𝐧 𝐚𝐧𝐚𝐥𝐲𝐬𝐢𝐬 𝐬𝐡𝐞𝐝𝐬 𝐥𝐢𝐠𝐡𝐭. John Burn-Murdoch analyzed U.S. employment trends across roles deemed high-risk for automation (based on Brookings & OpenAI research). Here’s what he found:
- 𝐍𝐨 𝐜𝐫𝐚𝐬𝐡 𝐲𝐞𝐭: Roles like accounting clerks, legal secretaries, and underwriters? No major drops in employment — yet.
- 𝐁𝐮𝐭 𝐬𝐨𝐦𝐞 𝐞𝐱𝐜𝐞𝐩𝐭𝐢𝐨𝐧𝐬: Writers and software developers are showing clear signs of AI-related job decline — diverging from industry trends around them.
𝐖𝐡𝐲 𝐭𝐡𝐞 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐜𝐞? A new study by AI research lab METR gives us a fresh lens. Turns out, 𝐀𝐈 𝐞𝐱𝐜𝐞𝐥𝐬 𝐚𝐭 “𝐧𝐞𝐚𝐭, 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝” 𝐭𝐚𝐬𝐤𝐬 — not necessarily ones that are difficult or technical. That means jobs like writing articles or coding cleanly defined programs are far more automatable than you might think. On the flip side, even simple jobs like travel agents, executive assistants and bookkeeping clerks require juggling unpredictable goals, unclear inputs, human interaction, and multitasking. In other words: 𝐡𝐮𝐦𝐚𝐧 “𝐦𝐞𝐬𝐬𝐢𝐧𝐞𝐬𝐬” 𝐛𝐮𝐲𝐬 𝐭𝐢𝐦𝐞.
𝐀 𝐧𝐞𝐰 𝐥𝐨𝐠𝐢𝐜 𝐨𝐟 𝐯𝐮𝐥𝐧𝐞𝐫𝐚𝐛𝐢𝐥𝐢𝐭𝐲. It’s not about how smart the task is — but how structured it is. Ironically, Silicon Valley’s obsession with efficiency and predictability may have made some tech jobs 𝐞𝐚𝐬𝐢𝐞𝐫 𝐟𝐨𝐫 𝐀𝐈 𝐭𝐨 𝐫𝐞𝐩𝐥𝐚𝐜𝐞.
In particular, jobs that are: linear, repeatable and freelance/contract-based are most at risk — especially when AI tools can be slotted in without HR friction.
𝐀 𝐠𝐫𝐢𝐦 𝐟𝐮𝐭𝐮𝐫𝐞 𝐟𝐨𝐫 𝐝𝐚𝐭𝐚-𝐝𝐫𝐢𝐯𝐞𝐧 𝐜𝐨𝐥𝐮𝐦𝐧𝐢𝐬𝐭𝐬? If your job involves gathering data, writing up clean summaries, and staying within a fixed word count… well, AI is coming for you — fast.
𝐓𝐡𝐞 𝐛𝐢𝐠 𝐥𝐞𝐬𝐬𝐨𝐧? AI isn’t just about automating low-skill work. It’s about automating 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐰𝐨𝐫𝐤. And that flips many of our assumptions on their head.