Every week, a new wave of articles warns us that AI is killing jobs. The argument is always the same: automation replaces human work, and workers lose. It is a tidy story. It may also be the wrong one.
The real problem may be quieter, slower, and more damaging. AI is not destroying jobs. It is destroying the willingness of organizations to grow people. And that distinction matters enormously — for firms, for the economy, and especially for anyone entering the workforce today.
The Jevons Paradox is not the point
Most commentators reach for the Jevons Paradox when discussing AI and jobs. Jevons observed in the nineteenth century that more efficient steam engines led to more coal use, not less — because efficiency unlocked demand. Applied to AI: if AI makes workers more productive, we will want more output, not fewer workers.
That is a reasonable argument in some contexts. But it sidesteps the real question. This is not a story about wanting more output. This is a story about who would pay to grow people — and right now, the answer is: nobody.
The training problem nobody wants to own
Here is the economic reality that most commentators might be missing. Training a junior employee is expensive. It takes senior time, patience, and a long horizon. The company that invests in training a twenty-two-year-old today may collect the benefit in seven years — long after the person who did the training has moved on, and long after the manager who approved the budget has left for another firm.
AI changes this calculus sharply. With AI tools, a small team of experienced people can produce what used to require ten. The temptation to stop hiring and training juniors is not irrational — it is the logical response to short-term incentives. Every individual manager, evaluated on this quarter’s output, makes the same rational choice. Stop building the bench. Use the tools. Ship faster.
The result is a collective action problem. Every firm does what makes sense for them individually, and the system as a whole would stop producing the experienced mid-level talent it will need in a decade.
A few firms will win big — Later
There is an investment angle here worth watching. A small number of firms with genuinely long-time horizons will continue to train juniors, precisely because everyone else has stopped.
In five to seven years, when the hollowing out becomes visible, experienced mid-career professionals will be scarce. You can poach a few senior people. You cannot manufacture an entire generation of capable thirty-year-olds who simply were never trained. The firms that built bench strength quietly during the AI adoption frenzy will collect a meaningful scarcity premium. The firms that cut training entirely will find themselves unable to grow — not because they lack capital or technology, but because they lack people who know how to do things.
This is not speculative. It is a predictable consequence of the incentive structure described above. The only uncertainty is timing.
So, what should a young person do?
If the market has stopped training you, the question becomes: how do you train yourself? And here, of all places, the Bhagavad Gita offers the clearest answer available.
Chapter 4, verse 34:
तद्विद्धि प्रणिपातेन परिप्रश्नेन सेवया |
उपदेक्ष्यन्ति ते ज्ञानं ज्ञानिनस्तत्वदर्शिन: ||
Seek this knowledge through humble surrender, sincere inquiry,
and devoted service — the wise who have seen the truth will teach you.
Shankaracharya, in his commentary on this verse, is precise about what each word means. He is not offering a general sentiment about being a good student. He is describing a method.
The three-part method
Pranipata — prostration. Not the performance of humility, but the actual thing. Approaching someone who knows more than you without the armor of your credentials, your opinions, or your need to appear capable. This is harder than it sounds, especially for people who are technically skilled and used to being the smartest person in the room.
Pariprasna — inquiry. Not asking surface questions to seem curious. Asking the real ones: Why did you make that call? What were you wrong about? What does this look like when it goes badly? These are the questions that extract genuine knowledge rather than polished answers.
Seva — service. Making yourself genuinely useful to the person you are learning from. Not networking. Not managing up. Actually doing work that helps them, so that the relationship is built on something real.
Those three words describe the entire apprenticeship model. And it is precisely this model that is being dismantled by the current AI adoption cycle.
Jnani versus Tattva-Darshi
The sharpest line in Shankara’s commentary is a distinction he draws between two kinds of knowers.
The jnani is the person who is learned, credentialled, fluent, and well-read. In today’s terms: someone who can produce polished output on any topic, speak confidently in meetings, and appear competent across every domain.
The tattva-darshi is different. Shankara says the word means one who has seen the truth. Not read about it. Not synthesized it from other sources. Seen it — through direct experience, through having done the work long enough to understand where it actually breaks.
His point is direct: knowledge imparted by those who have seen the truth takes effect. Knowledge from the merely learned does not, or not in the same way.
This is the whole game now. AI will make everyone look like a jnani. Fluent, articulate, able to produce output on anything within seconds. What AI cannot manufacture is the tattva-darshi: the person who has done the work long enough to know when the confident answer is wrong, to make a sound call on incomplete information, to see the thing beneath the surface that the tool cannot access.
The practical implication
For young people entering the workforce, the advice follows directly from the analysis.
Do not optimize your first job for title or brand name. Optimize it for how fast you get good — which means: how close you are to people who have actually seen the truth in your field.
A well-known firm where you spend three years producing AI-assisted output with minimal senior exposure will leave you fluent and shallow. A less prestigious role where you sit next to someone who has been doing this for twenty years, who makes real decisions and lets you watch — that will make you rare.
Approach those people through pranipata, pariprasna, and seva. Stay low. Ask the real questions. Earn your place by being useful. This is not advice about networking or impression management. It is a description of how knowledge actually transfers between people.
The market is quietly eliminating the apprenticeship. Your job is to find one anyway.
(This piece is mostly based on a post written by a dear friend, who is a great exponent of Shrimad Bhagwat Gita, and regularly delivers talks on Gita.)