Decision Fatigue & AI
AI allows for faster results, but there is still a hard human limit
During my time working with AI to produce code, design documents, and create tests, I've discovered some things.
GIGO (Garbage In, Garbage Out) still applies. AI is a magnifier. It will magnify bad software development practices and good ones. "Move fast and break things" means things will be broken much faster. Tests, docs, plans; all are necessary context. Giving your agents some kind of access to a ticket system ( even git bug ) massively improves their performance. Tests are basically 'free' now as are doc comments.
Most people don't know how to actually USE AI. What helped 6 months ago now mostly hinders SoTA models. Too many skills or MCPs, or a large Claude or Agents file actually degrades performance. AI is moving faster than the search engines, which return stale results about best practices that now can actively harm performance.
Decision fatigue is a well known phenomena. And I am finding its possible to hit that limit much quicker with AI. Humans need time and discernment to choose and guide outcomes from AI. If we are reduced to mere 'prompt' or 'loop kickoff' machines, if we give metrics that require us to run these tools 24/7 or making managerial decisions every minute of every work day, workers will burn out.
I am now doing analysis, product design, and research. I can have AI help me do a lot of that. But then there are a lot of documents generated. Things I need to read, understand, verify, agree with, push back on, record, plan, and finally decide on. Jeff Bezos famously said he can make 3 big decisions a day.
Workers are gonna be making more managerial style decisions. And I don't know if businesses are ready for that. Are they gonna give workers the kind of space needed for this, or is it gonna boil down to metrics yet again and not outcomes?
And I think this will likely move to ALL industries eventually. As robotics become more capable, perhaps, eventually, even the lawn care folks will need folks to be managers of robot teams.
Systems thinking, management theory, those things will become even more critical as AI continues to mature.