Discernment & Discipline
The two capacities our children will need in the AI economy — and why the conditions that used to develop them are disappearing.
My son turned five this year. Which means I’ve spent the year school shopping — navigating the anxiety of not wanting to make a mistake with his education and the worry of whether I’ll actually be able to afford the one I want for him.
The other day, my wife and I visited another school on our shortlist. While we were waiting at reception to meet the admissions officer, she gestured for me to come look at the notice board. It had a large poster of the school’s top academic performers. “They’re very focused on academics here”, she said. We spent the next hour touring the school’s impressive facilities while the admissions officer told us about the school’s student-teacher ratios, experienced faculty and extracurricular activities.
We left with a brochure of term dates and the feeling that this school, like the others we had visited, was designed for an old economy — not for the one our son will be stepping into.
For decades, the knowledge economy rewarded the ability to acquire and apply knowledge. Doctors, engineers, consultants — anyone whose work depended on knowing things — did well. Schools were built to feed this system. The path was clear. Study hard to acquire knowledge, ace examinations to prove you’ve retained it and you’d be rewarded.
Earlier this year, Anthropic, the company behind Claude, launched Agents for Financial Services. Built specifically for the industry, these agents draft pitch books, prepare briefs, build financial models, review valuations, audit financial statements, and more. These tasks are essentially what a fresh graduate at an investment bank does — the entry point into one of the most prestigious careers in the world. The same shift is unfolding elsewhere. In law, AI agents are handling whole workflows that used to define junior associate work — reviewing contracts, surfacing precedents, drafting memos. In healthcare, diagnostic models are reading scans and suggesting diagnoses faster and more accurately than residents. In software, agents are writing and testing routine code with less and less human direction.
Most of us still picture AI as a chatbot — something you ask questions and get answers from. That picture is already out of date. AI is increasingly autonomous: not answering questions but doing the work. And it is rapidly disrupting the once predictable knowledge economy. Answers are becoming cheaper to produce and routine tasks easier to automate. In this new world, the routine acquisition and application of knowledge will no longer hold the premium it once did. Instead, two human capacities that used to form naturally — without our attention — will become far more valuable. The first is discernment and the second is discipline.
Discernment is the work of asking whether the answer you’ve been handed is the right one. It used to be a quiet capacity, exercised occasionally. It is about to become a daily one. When answers were scarce, the discernment required was modest — you had a small set of options and chose between them. AI has inverted that. The answers are no longer scarce. They are abundant, instant, and increasingly confident — whether or not they are true. AI will produce more of the analysis, the recommendation, and increasingly the judgement itself. What remains is the harder meta-judgement: is this answer complete, could it be framed differently, is it biased towards someone’s interests? Most of what AI produces will sound reasonable. Knowing what to trust, what to question, and what to ignore is the job.
And discernment runs deeper than evaluating outputs. It is also the capacity to know who you are. Millions of people every month run a Google search for some version of “How can I make more money?” Soon, AI will not just answer that question. It will give them a dozen plausible plans, in detail, with execution steps. The question is no longer can someone tell me how to make more money. The question is: which of these am I willing to stay committed to when it gets hard? That question cannot be outsourced. It requires knowing what you value and what kind of work would feel like a gift rather than a grind.
Young people used to wrestle with these questions naturally, in the long encounters of childhood — the stretches of boredom, thick books, and meandering conversations with best friends. Those conditions are disappearing.
Discipline, on the other hand, is the willingness to persist with challenging work. To be clear, discipline is not blind compliance. It is the willingness to choose a longer arc over an easier path, to fail and try again, to stay in difficulty long enough for something to form. The new economy is likely to demand more of this, not less. As AI takes over the work that is routine and known, what’s left is the work that is ambiguous and hard to know without long periods of immersion.
The old economy built discipline into people whether they wanted it or not. The 23-year-old analyst who spent her first two years grinding through spreadsheets at 2am wasn’t just producing reports. She was developing the capacity to stick with hard things. AI does much of that early-stage procedural work now. The grinding is gone. The discipline that the grinding used to develop now has to come from somewhere else.
Discernment without discipline is the perceptive critic — the one who diagnoses everything and builds nothing. They see clearly what is worth doing, name it well, and never do it. Discipline without discernment is the striver — the one who commits, heroically, to the wrong thing. The capacity to persevere becomes the engine that delivers them somewhere they never chose.
Both are recognisable failures of modern professional life. The pair matters because each capacity, on its own, curdles into something the other was meant to prevent. Young people need both.
The question hanging over all of this is what it means for jobs. There is a loud, ongoing debate. One camp argues we are on the edge of mass displacement. The other argues every technology has eventually created more jobs than it destroyed. Whatever happens to the jobs themselves, the future will not resemble the knowledge economy we’ve become accustomed to. In either scenario — job apocalypse or job evolution — the economy will reward discernment and discipline more than it ever rewarded knowing.
This is a hard diagnosis to absorb. I’ve heard three common counter-arguments to this thesis.
The first is the voice that says this is overblown. Every generation has its end-of-the-world technology. The printing press, radio, television, internet — none of them turned out to be the apocalypse their loudest critics promised. AI, this voice says, is the latest in that line. The economy will adjust. New jobs will emerge. Kids who do well in school will continue to do well.
This voice is hard to dismiss. My wife and I often wonder if we’re overthinking it. Whether or not the diagnosis is overstated is not the right question. The right question is “What is the cost of being wrong?” If we are wrong — if we invest in developing discernment and discipline only to find the diagnosis was overblown — we will have raised children who can question what they are handed and stay with difficulty when the work gets hard. That is not a bad outcome in any economy. If we do not reimagine how we develop our young today, we run the risk of raising children who are optimised for an economy that no longer exists. The cost of taking this seriously and being wrong is small. The cost of dismissing it is large.
The second counter-argument is the voice that says schools will figure it out. Curricula will change. New subjects will be added. AI literacy will be taught alongside maths and English. Schools will eventually catch up to what the world requires.
We tend to think of schools as independent institutions that have agency. The reality is more sobering. Schools are at the mercy of their customers. The parent market has asked schools for one thing above all else for the past two generations: credentials that increase the probability of a stable, well-compensated career. Schools have responded by becoming extremely good at producing those credentials. They have optimised for measurable academic output and removed the slow, unmeasurable parts of education that did not feed the credential. This is not a failure of schools. It is evidence of their effectiveness — they are delivering exactly what their customers asked for. And their customers are us.
The third counter-argument is the voice that says kids will figure it out themselves. They have grown up with these tools. They will navigate AI the way our generation navigated the internet. We would like to believe that adolescence is a self-correcting system, and that a curious, capable young person will encounter the world and develop what they need. They will figure it out if the environment lets them. The problem is: increasingly, the environment doesn’t.
We know what phones and social media have done to teenagers. The collapse of attention, the erosion of solitude, the narrowing of disagreement. AI introduces something new. It takes away the work itself. A teenager who would once have sat with a hard question now asks a chatbot. The wrestling that used to do the developmental work is gone before it can begin.
If the environment that once developed these capacities is no longer doing that work, the obvious question is what should we do about it. There are two conditions we can cultivate to address this challenge.
The first is sustained engagement with difficult questions. A young person who is allowed — and expected — to sit with a hard question long enough for something honest to emerge is developing a different capacity than the young person who is rewarded for fast, correct, predictable answers. The wrestling is the work.
This is the developmental condition most directly at odds with how modern education is structured. Schools are optimised for the rapid production of answers. The faster and more accurate, the better. We need to do something slightly counter-cultural — to value the slow, uncertain, often uncomfortable work of wrestling with a question more than we value the speed and confidence of the answer.
This is not about finding purpose. The language of purpose, of self-discovery, of authenticity has become a kind of consumer good in the developmental space, and most of it is either too lofty for a teenager or too neatly resolved to be honest. This work is different. It is the practice of returning, over months, to questions that matter. Questions like: what do I actually want, what do I believe and what am I willing to struggle for? Young people who do this repeatedly develop the capacity to hold their own ground — to resist the easy answer that is always available.
The second is meaningful work with real stakes. A young person can become extremely good at meeting externally imposed expectations without ever developing the internal capacity to sustain difficult work they chose for themselves. The experience that develops discipline is different. It is having committed to something that matters, where the outcome is genuinely uncertain, and where staying with the work is the only way through.
The current environment offers young people very little of this. The work they are asked to do is mostly assigned, mostly graded, mostly designed to be completable within a known structure. The stakes are real but narrow — a mark, a placement, an admission. The young person does not get to discover, in the doing, that they have more capacity for difficulty than they thought. They only get to discover that they can or cannot complete the assignment.
In a heavily scheduled and heavily optimised teenage life, there is very little that meets this description. Protecting the conditions for discipline means more than making time in a child’s schedule for a personal project. It means letting a young person choose something that matters to them, and then not rescuing them when it gets harder than they expected.
Beyond cultivating these conditions for young people, the most consequential thing a parent can do is to role model these capacities. Discernment and discipline cannot be cultivated by adults who are pretending. If we hide our own uncertainty, we cannot develop our children’s capacity to sit with uncertainty. If we hide our own struggles, we cannot develop their capacity to stay in difficulty. We need to let the work be visible. Not theatrically, but honestly. Our children need to see us wrestling with real questions, keeping going when something is hard, admitting we do not know.
When I was a teenager, my mother started spending long stretches of time in north-eastern Sri Lanka. It was a period of peace during the country’s brutal civil war, and the Tamil community there was trying to rebuild something — including, for a while, the early structures of a government. She was a consultant and a lecturer in Kuala Lumpur. She had three children at home. She was not someone whose life had any obvious reason to bend toward that work. But she became curious about it, and then committed to it, and then was conflicted about it in a way she never tried to hide from us. I watched her wrestle, openly, with whether it was right to keep going. Whether the work mattered enough. Whether her children would understand. She kept going anyway. She started a small nonprofit. She made the trips. She came back tired and went again.
She did not sit me down and explain what she was doing. She did not narrate her courage or her uncertainty. She just lived it where I could see it. That is the developmental condition I am describing.
Not instruction. Example.
Discernment and discipline are not only about economic security. They are the capacities that have always been required for a human being to flourish — to know what they value and to persist through difficulty.
The need to develop these capacities in the young is not new. What is new is that the conditions that used to develop them automatically are disappearing.
We are the first generation of parents who have to think about this deliberately.