As AI handles more of the execution, what work should belong to agents vs. humans—and why does that distinction matter?
Over the past few years, we've all witnessed and experienced a slow but significant shift in how we work. I've upskilled repeatedly, and my analytics stack has kept rebuilding itself beneath me. When I first read about the power of generative AI, my immediate thought was, "I'll never have to spend hours on StackOverflow debugging code again."
Today, I use AI models at work to write, design, code, brainstorm, complete tasks at breakneck speed, draft proposals for executives, craft analytical stories, test hypotheses, and prepare me for important meetings and tough conversations—and that's just the start.
But the more I integrate AI to make my life 'easy,' the more I realize there's tough, indispensable work that humans do that AI cannot replicate. Generative AI cannot set aspirations, make decisions in difficult times, build trust among stakeholders, or hold itself accountable for outcomes—good or bad. That work remains deeply human.
As agentic AI absorbs more of the analytics workflow, the scarcest resource in your tech stack won't be speed or execution quality. It will be the ability to generate original thought, apply business judgment, and decide what deserves attention in the first place.
Independent Thought Remains a Competitive Advantage
I've come to believe that AI won't replace human jobs, but it will help us execute work faster than ever. That raises the real question: What do we do with the time AI gives back to us? Do we use it to think more deeply, explore new ideas, and tackle harder problems? Or do we gradually stop doing those things ourselves?
Compared to earlier this year, I've noticed a shift in my own behavior. I increasingly catch myself asking Copilot to interpret data patterns before I even look at them myself. I prefer convenience over challenge. Why spend twenty minutes exploring a dataset when an AI can lay the groundwork in seconds? The convenience is undeniable.
Yet, this growing dependency makes me nervous—and not because the agent sometimes gets it wrong (often, it doesn't). What unsettles me is how easy it becomes to skip the most valuable part of analysis: forming my own initial point of view. Without that effort, I'd essentially be receiving someone else's version of the insight, dressed up as my own.
That's my big anxiety around the AI-and-jobs conversation: not becoming obsolete, but becoming unoriginal.
Imagine going through an entire career without regularly having a single original thought. I don't want to lose the intellectual friction that produces genuine insight. That friction—the struggle to understand, question, and synthesize—is where true analytical value lies. And as of 2026, with agentic AI increasingly embedded in everything from data preparation to dashboard generation, preserving that skill is more critical than ever.
In practical terms, this means being deliberate about how we delegate. Use AI for execution but reserve the thinking for yourself. Ask the agent to prepare, but always review the raw data first. Let AI handle the routine, but never outsource the curiosity that drives better questions. The agents will get faster; the differentiator will be who asks the better question—and who has the judgment to act on it.