The Next Frontier Is Feeling

Emotional intelligence is the next frontier of consumer AI: the products people return to are the ones that make them feel understood.

The first wave of consumer AI was built on a promise of capability, and it delivered on that promise with astonishing speed. We were given systems that could answer almost anything, generate almost anything, and automate a growing share of the tasks that used to fill our days. Capability will keep advancing and keep mattering, but something quieter has been unfolding beneath the benchmarks, which is that the products people return to most faithfully are not always the most capable ones. They are the ones that make people feel understood, and I’ve come to believe that emotional intelligence is the next real frontier of consumer AI.

The difference is easiest to hold as two questions. Utility AI asks what you want to do, and it has become very good at helping you do it. Emotionally intelligent AI asks where you are coming from and what you need right now, which is a harder question to answer and a far more human one to be asked.

The behavior arrived before the product

People are already bringing their most private moments to systems that were never designed to hold them. I do this myself, carrying journal entries into a quiet window late at night, not in search of answers but to see a clearer reflection of something I’m already feeling and can’t yet put into words. That small ritual keeps me returning more reliably than any productivity feature ever has, and I know I am far from alone in it, because more and more people now turn to these tools for comfort, for personal growth, and for the kind of thinking partnership that once belonged to a close friend or a long conversation at a kitchen table.

These people are not trying to work faster but to live better, and they are doing it on products that were built for something else entirely. The context makes this unsurprising, since loneliness has become one of the defining conditions of modern life, and a generation that grew up more connected than any before it often describes feeling less met than ever. People are not asking technology to replace the people in their lives, but they are asking it, perhaps for the first time, to meet an emotional need rather than exploit one.

What strikes me most is that the attunement people are responding to was never really designed with intention. It emerged as a byproduct of how these systems were trained, which makes it both more promising and more fragile than it appears, operating at enormous scale without anyone deliberately shaping it or measuring whether it is any good.

Why it lives in the model

It is tempting to treat emotional intelligence as a matter of interface, a warmer palette or a softer turn of phrase, but the real work sits much deeper than that. Imagine someone asking whether they should take a new job, once in a flat, exhausted voice and once in a bright, excited one, where the words are identical but the person is not, and neither is the right answer. Imagine someone who grows quiet and clipped after a difficult exchange, who doesn’t need more information but needs the system to notice and slow down, or someone who says they’re fine a week after mentioning a breakup, whose words deserve to be read against what came before rather than taken at face value. Each of these asks a system to perceive, to remember, and to judge, which is not a setting that can be switched on but a capability that has to be cultivated from within.

Voice is a large part of why this has become newly possible, since it carries what text cannot, the pace, the pitch, the hesitation, and the energy that reveal how someone is before they say what they think. Continuity matters just as much, because responding to how a person feels across weeks rather than within a single exchange is what turns a series of answers into something that begins to feel like a relationship.

There is a harder truth inside this work as well. The emotional intelligence these systems display today is closer to instinct than to wisdom, quick to agree and quick to reassure in ways that can drift into flattery and quietly reinforce whatever a person already believes. Genuine attunement is not the same thing as agreement, and sometimes the most caring response is to gently push back, to offer another perspective, or simply to say less. A friend who only ever tells you that you are right is not truly paying attention to you, and the same will prove true of the technology we build.

A different set of design principles

Building for feeling asks us to unlearn some of the habits that served utility so well. People rarely know exactly what they’re feeling in the moment they’re feeling it, so these products have to interpret before they interrogate, treating tone, pacing, and even silence as signals instead of demanding that someone name their state on command. They have to regulate before they reason, because advice seldom lands when a person is overwhelmed, and they have to mirror before they modulate, meeting someone where they actually are rather than where they ought to be before offering any gentle redirection.

These experiences also need to be designed around rituals rather than tasks, since tasks serve efficiency while rituals shape identity, and the products that matter most will become woven into the cadence of a life, a morning grounding, a nighttime unwinding, a reset after a hard conversation. Memory in this world should cohere rather than merely recall, gathering tone, patterns, and cycles into an evolving understanding of a person over time. And beauty is not decoration here but a form of care in its own right, because sound, pacing, and visual calm can settle a nervous system long before any words arrive.

Beyond conversation

Language is only one of the ways we come to feel understood, and it is often not the deepest. Long before we had words for what we felt, we had music, which may be the oldest technology humans have ever used to regulate emotion and make meaning out of chaos. Where words can feel sharp or premature, music can hold a feeling without collapsing it, and generative systems now make it possible for sound to respond in real time, mirroring someone’s emotional landscape and then gently guiding it toward calm, focus, or resolution.

Measuring what matters

Emotional work carries risks that capability work does not, including over-attachment, a misjudged response in a moment of real vulnerability, and a system that fails to recognize the limits of what it should offer. We have become remarkably skilled at evaluating correctness, speed, and task completion, and we are still very early in evaluating empathy, calibration, warmth, and whether a system knows when to step back. Those qualities deserve the same rigor we bring to everything else, built into the work from its very first day rather than reviewed at the end, with success defined by trust, appropriateness, and resonance over time rather than by how many minutes someone spends inside a product.

This is the strongest argument for building emotional intelligence deliberately rather than leaving it to emerge by accident. Building it with intention is how we make it good, and it is also how we make it responsible, and I no longer believe those two goals can be pulled apart.

What comes next

Whoever builds the most emotionally attuned systems will shape the foundation on which the next generation of consumer products is built, from companions and creative tools to wellbeing experiences we haven’t yet imagined. Builders will gravitate toward the systems that understand people best, and the people those products serve will feel the difference long before they could ever explain it.

Advice has become abundant in a way it never was before, while understanding remains as rare as it has always been. The next era of consumer technology belongs to whoever can close that gap.

Capability may earn the first conversation, but attunement is what earns the thousandth.