Further to https://pjlee.net/blog/ai-agents-very-important-in-2025

my experience of intensive use of AI during 2025, and especially so far during 2026 (because the pace of improvement has accelerated!) is that AI is coming for white-collar jobs faster than predicted.  Blue-collar jobs too, but it will take longer to mass manufacture hundreds of millions of robots.  

Producing AI agents is much much quicker. 

So within perhaps 2 years at most, AI looks likely to be able to do almost all white collar work.  There will be exceptions, but those look likely to be temporary: e.g.

• people with specialist knowledge of a company's clients, data and systems (the AI will catch up eventually)
• people with roles mandated by regulation (regulators will be slow to act, but eventually will realise that AI will eventually be capable of doing a better job).

Many employers will be slow to react at first.  But the pressure not to recruit new humans (unless a strong business case can be made that the work can't be done by existing employees augmented with AI) will grow.  Employers that fail to adapt are likely to find that competitors outpace them.

So what should you do? 

Use AI to increase your productivity by a multiple.

That's what I have been working on over the past year. I have achieved it in several areas (most spectacularly in coding) and have been building a system that increases my productivity every day.

There are big societal issues coming too: as I put in a newsletter to International Qualified Actuaries Group members recently:

"We're heading toward an agentic world where AI agents can do most white-collar work within 2 years. Productivity becomes very high, products become cheap — but far fewer people work, including actuaries. Regulatory inertia will protect some temporarily, but not indefinitely.

The disruption goes deeper: What do pensions look like when lifespans extend but careers shorten? What's insurance's role with abundant goods but scarce employment? How do we finance universal basic income when traditional tax bases collapse? What's the point of professional exams when AI can do the work?"