AI is already changing how knowledge work gets done. Tasks that took weeks now take days. Systems that could only handle short, isolated jobs are running workflows of several hours with hardly any oversight. That part is real and measurable.

Everything beyond that is guesswork, and I've stopped pretending otherwise.

Forecasting technology has always been a humbling business. We can see a year or two ahead, roughly: the tools that exist, the trends, what the early adopters are doing. After that, visibility drops off fast. Anyone who claims to see the full picture from here is selling something.

The Steam Engine, Not the Internet

If you want a historical parallel, I don't think the internet or the smartphone is the right one. The steam engine and the mechanization of agriculture feel closer. Those didn't just make existing work faster; they redefined what work was, and reshaped entire labor markets in the process.

I'll admit every generation believes its own technology is the big one, so take the comparison with a grain of salt. But the pattern fits uncomfortably well. Work done behind a computer, especially routine cognitive work, is the most exposed. Roles built on judgment, accountability, and deep context are holding up better, for now. That boundary keeps moving though, and in my experience it moves faster than most people expect.

Building in Low Visibility

For entrepreneurs this is an uncomfortable position. You're building a business inside a transformation whose shape you can't see. There's no playbook, and the people selling playbooks are part of the problem.

My working conclusion, for now: stop trying to be right about the future and get faster at noticing when you're wrong. Short feedback loops beat good predictions in a climate like this. Experiment early, measure honestly, change course without ego when the data tells you to.

The future with AI is unclear. I've made my peace with that. It beats false certainty, and it leaves room to actually pay attention.