Living Alongside AI

I have been using AI tools to build Majhi OS — an autonomous hiring operations system — for the better part of two years. AI is inside the product. AI accelerates how I build it. AI expands what is possible to build with a small team.

None of that is the most interesting thing I want to say about living alongside AI.

The interesting thing is what I have learned about what AI cannot do — and why that turns out to matter more than what it can.

Two failure modes

There are two ways to get your relationship with AI wrong.

The first is resistance. People in this camp treat AI as a threat to their relevance, avoid learning to use it, and find their productivity relative to peers declining as a result. This is the failure mode that gets talked about most.

The second is less discussed, and I think more dangerous for the people it affects: deference. Deference is using AI for so much, so fast, that you stop building the judgment that AI is supposed to be leveraging. You outsource not just execution but thinking. Over time, you become a fast executor of AI outputs without the independent capacity to evaluate whether those outputs are any good.

Both failure modes end in the same place. The first gets there slowly, through avoidance. The second gets there faster and more invisibly, through over-reliance.

What AI is

AI is best understood as leverage on your existing capability. If you have the judgment to recognize a good outcome — if you know what you're trying to accomplish and can evaluate quality — AI dramatically accelerates your ability to produce and iterate toward it. What took a week takes an afternoon. What required a specialist takes a conversation.

That compression is real. I use it constantly. The question it surfaces is not "should I use AI?" but "what is it leveraging?"

If AI is leveraging genuine judgment — your ability to evaluate quality, to recognize what the situation requires, to make calls under ambiguity — the output improves and compounds. You get better at the things that matter because the friction that used to absorb your time is removed.

If AI is substituting for judgment you haven't yet developed, the output gets faster and shallower simultaneously. You produce more, but you are not the author of what you produce in any meaningful sense. The loop that builds skill — doing something, seeing the outcome, adjusting — is short-circuited before it completes.

AI is leverage on your existing capability. The question is not whether to use it. The question is what it is leveraging.

The identity question

I think the most useful question to ask about your relationship with AI is not about productivity. It is an identity question: if AI could do everything you currently do, what would you actually be for?

This question is clarifying in an uncomfortable way for most people. The honest answer reveals which parts of your current work are about executing — following established processes, producing outputs in known formats, processing information flows — and which parts are about being someone: having particular knowledge, relationships, judgment, or context that is specifically yours.

AI is displacing execution. It is not displacing being someone.

The lawyers who will be most valuable are not the ones who are faster at the things AI can now do faster. They are the ones who have the judgment to know when the AI output is wrong, the relationships that make clients trust them with decisions that go beyond the research, and the understanding of what a client actually needs versus what they asked for. Those things take years to build. AI does not compress them.

The same pattern holds in every field I have observed closely enough to understand.

What I keep

I can be specific about what this looks like in my own work.

I use AI extensively for drafting, synthesis, and the first pass at analysis. It saves significant time and often surfaces connections I would have taken longer to reach independently.

But the things that determine whether Majhi Group closes a mandate or whether Majhi OS wins a client are not things AI produces. The judgment about what a client actually needs. The insight that a search has stalled for a specific reason that isn't visible in the data. The read on what a particular executive's motivations are and how to align with them. The call about whether a situation requires directness or patience.

AI does not produce those things. I have tried to outsource them and found that the output degrades in ways that are immediately visible to anyone who knows what good looks like in those situations.

What I keep is exactly the work that required the most investment to develop. That is not a coincidence. The work that compounds — that builds something genuinely difficult to replicate — is work that cannot be easily compressed.

The practical implication

The practical implication is simpler than it sounds: use AI aggressively for everything it can do, and protect ruthlessly the time you spend on the things it cannot.

Most people get this backwards. They are slow and reluctant to use AI on execution tasks that AI handles well, and they allow AI to crowd out the deep thinking, the direct contact with clients and colleagues, the reading and reflection that builds the judgment they will eventually need AI to leverage.

The people who will be most valuable in a world where AI is pervasive are those who know clearly what they are for — and have built that over years of practice. AI becomes their force multiplier. Everyone else is running faster toward the same cliff edge.

Living alongside AI, done well, is not complicated. Know what you're for. Build that. Use AI for everything else.

See also: What the History of Automation Actually Tells Us About AI, AI and Human Potential, What AI Agents Actually Are

Sources

McKinsey Global Survey on AI — The State of AI, 2024

MIT Sloan Management Review — New Frontiers in Reskilling and Upskilling

Harvard Business Review — Don't Let AI Destroy the Skills That Make Your Company Competitive