Why We Need Good Friction in the Age of AI

Recently, I was at an alumni meet at a popular night spot in Mumbai, catching up with about a hundred alumni—perhaps a third of whom I have had the pleasure of congratulating at convocation during my tenure as Dean of SPJIMR. Naturally, the conversation turned to what's on everyone's mind: artificial intelligence.

At one point, a recent alum grinned and teased me: "Dean, honestly, what is your relevance or the relevance of any faculty member in the age of AI?"

Touché.

Playfully, I shot back: "Enough about me, what about you: what is your relevance to your company if a machine can do your job?"

We laughed, but that exchange stuck with me. It raised a question we all need to ask. And I don't mean a sci-fi scenario where a self-learning and autonomous rogue AI takes over the world. We have to face a much more immediate, everyday question: What does it mean to be human when we outsource the effort of daily living to a machine?

Think about what happens when you use GPS to navigate through Mumbai's notoriously chaotic streets. You reach your destination, but you learn absolutely nothing about the city. If your phone dies, you are completely lost.

This is the paradox of modern convenience: by eliminating the friction of finding our way, we lose our capacity to navigate. It is what could happen to our minds as we adopt AI.

As an innovation professor, I look at how technology evolves. One of the big debates around AI is whether it will "augment" us or "automate" us. It is a crucial distinction. Augmentation expands human possibilities; it gives us superpowers to know and do. Automation, by contrast, reduces or removes human effort to make things simpler and faster.

Of course, not all automation is bad. Good automation removes bad friction—the mindless slog of sorting databases or managing basic scheduling—freeing us up to do more meaningful things.

But bad automation removes good friction—the productive struggles that actually shape who we are.

The danger today is that while trying to augment our work, we may actually be using AI to automate away much of this good friction. By promising a frictionless life, AI could eliminate the very struggle that builds human capability.

In the gym or the classroom, the old rule still holds: no pain, no gain.

For leaders, this creates a difficult choice. Where should we use AI to remove effort, and where do we need to preserve some of that effort because it is essential to how people learn, exercise judgement and work with others?

From my own experience as a manager, educator and institutional leader, I find that question particularly relevant across three sets of human capacities: knowing and doing, judging and deciding, and mobilising and executing.

Knowing and doing

To navigate the world effectively, we must first understand it, and then act upon that understanding. Knowing is the process of building a coherent mental model of reality, while doing is the translation of that model into execution.

In human development, these two processes are not separate steps. They form a continuous, reinforcing learning loop. We act on our knowledge and, in doing so, further build knowledge.

Historically, major technologies have stepped in to augment specific halves of this loop.

In the seventeenth century, the microscope extended our sensory perception to the micro-scale. During the Industrial Revolution, the slide-rest lathe did the same for our "doing", allowing human operators to carve metal with unprecedented precision.

But these historical tools did not run themselves. A scientist still had to squint through the lens, adjusting focus to make sense of what they saw. A craftsman still had to turn the lead screws, watch the iron shavings and adjust cutting speeds based on the physical feedback of the metal.

These technologies expanded human capability, but mastering them required active participation, repetition and hands-on practice.

Now think of AI.

It acts as both a microscope and a lathe for the mind. AI can automate both halves of this knowing-doing loop simultaneously: it can instantly synthesise vast fields of knowledge and immediately apply that knowledge to create a finished product—from generating software to drafting a business proposal.

But this complete outsourcing is a seductive shortcut.

When we let AI automate both the "knowing" and the "doing", we bypass the productive friction of turning information into action. When we rely on chatbot summaries to avoid the friction of deep reading, or accept AI drafts on faith, bypassing the struggle of trial and error, we compromise the learning.

The hours spent wrestling with a difficult concept, hunting down a broken variable in a line of code, or critically redirecting AI are not wasted time. They are the gym where critical thinking and struggle turn raw information into human intuition, heuristics and creative taste.

Without that hands-on struggle, we risk training—not educating—a generation of high-level orchestrators who have no calluses, leaving them unable to recognise when the machine has designed a seemingly beautiful but fundamentally flawed house of cards.

We risk training—not educating—a generation of high-level orchestrators who have no calluses.

Judging and deciding

In leadership and daily life, choice is never just a cold calculation of mathematical probabilities.

Judging is the deeply human, contextual process of evaluating ambiguous, value-laden scenarios, while deciding is the act of choosing a path and taking responsibility for its consequences.

This is what philosophers call phronesis, or practical wisdom. Making wise decisions demands much more than raw data processing. It requires intellectual humility, empathy and perspective-taking. These are deeply relational muscles as much as cognitive ones.

Centuries ago, maritime navigation changed how humans crossed the globe. The compass gave sailors basic direction. The sextant helped them establish where they actually stood. Later, the gyroscope helped keep navigational instruments steady even as a ship pitched and rolled.

These tools didn't make decisions for the crew; they augmented their judgement. Sailors still had to read currents, weigh risks and chart courses.

Today, AI is the ultimate simulator. It can model complex feedback loops and downstream consequences across massive networks, helping us evaluate dilemmas that are far too large to hold in our heads at once.

But AI can truly augment our decision-making only if we deliberately inject the very friction points we are tempted to automate away.

Because AI's suggestions are so well structured and confident, they are incredibly seductive. If we don't bring our own friction to the table, we can succumb to automation bias, letting our decision-making muscles go soft.

We trade phronesis for mere optimisation.

Good judgement needs more.

We need our internal sextant—perspective-taking and empathy—to look outward, ensuring that we do not blindly accept optimised metrics while forgetting the human lives affected on the ground.

And we need our internal gyroscope—the intellectual humility to doubt the machine's "perfect" recommendations and the quiet reflection to resist the pull of convenience.

Without these friction points, we risk mistaking optimisation for wisdom. And that leaves us unprepared when we confront a novel situation that looks nothing like the algorithm's historical training data, or when we have to navigate intellectually and morally ambiguous waters.

Mobilising and executing

Finally, once a decision is made, we must turn it into reality.

Mobilising is the social work of aligning hearts and minds, earning trust and inspiring people to move together, while executing is the physical and logistical coordination required to deliver the outcome.

AI can spectacularly augment coordination. It moves us from rigid planning towards adaptive, real-time choreography at a massive scale.

Think of trying to organise an annual bicycling trip with your college buddies or a neighbourhood clean-up. Humans quickly hit a coordination ceiling. We fall back on rigid schedules because we can't process a hundred moving preferences at once.

AI-augmented coordination changes this. By managing thousands of real-time variables in the background, it can dynamically tailor schedules, tasks and logistics on the fly.

But while AI can spectacularly augment coordination, true cooperation requires human effort—and the attendant good friction involved.

We use AI to augment our execution when we offload the dizzying logistics of coordination to the machine, freeing up our scarce time and energy so we can actually show up for our people.

But we fail when we hide behind the machine to avoid the uncomfortable, messy friction of real human contact.

The real crisis of the AI age is not that the machines are coming for our jobs. It is that they are coming for our effort.

Cooperation is not a transaction to be optimised. It is the shared work of reaching alignment.

Getting friends to agree on a trip or mobilising a community requires active human presence, earning trust and navigating delicate emotional dynamics. It requires sitting across a table, practising empathic persuasion, actively listening and having the humility to co-create solutions rather than imposing optimised plans.

That dialogue is how shared ownership is built and collective action becomes possible.

If we treat cooperation as an algorithmic transaction, our relational muscles atrophy, leaving us with on-paper perfect plans that fail because we have lost the ability to move human minds and hearts.

The real choice

Ultimately, the real crisis of the AI age is not that the machines are coming for our jobs.

It is that they are coming for our effort.

By making life so convenient, AI invites us to trade the messy labour of self-creation for the passive consumption of optimised outcomes.

But human capability is a set of distinct cognitive and relational muscles—our curiosity, critical thinking, repetition, reflection, perspective-taking, empathy, intellectual humility and active human presence.

None of these can grow without resistance.

The task, then, is to become much more deliberate about which friction we remove and which we preserve.

We may need to choose the frustration of an unsolved problem, a clumsy first draft, the weight of moral doubt or the friction of a face-to-face disagreement.

That friction is not always the obstacle to growth.

Sometimes, it is the growth.

And if we want to keep our edge in a world of effortless answers, we may need to remember the old rule of the gym: no pain, no gain.

Behind the writing

In the spirit of practising what one preaches, a brief disclosure: I co-piloted this essay with an AI—Gemini in Canvas co-writing mode.

But if you think I simply let it generate a draft and then polished away the synthetic "AI-speak", you've missed the point of good friction.

I treated the system as an infinite, hyper-responsive whiteboard—or as a favourite graduate student—to stress-test my own theories. I threw tangents at it, directed it to build on my intuitions, laid out conceptual dots and used my own experience and personality to guide the flow.

Whenever the AI connected those dots clumsily, I sent it back to the drawing board.

This wasn't about outsourcing the writing. It was about using the machine to rapidly visualise, test and reject different logical combinations. The AI didn't do the thinking for me. It was a tireless sparring partner that forced me to sharpen my own arguments.

And I have no illusions that AI is going to help me actually convince anyone of these ideas.

To win over my colleagues in the academy, my industry clients, or my students and alumni, I will still have to do the real work in person—one messy, face-to-face discussion at a time.

That pain is a small price to pay for being human.