First, we wanted it to be smarter than us.

For a while, understanding us didn’t matter as much.

Asking AI to do a thing was like grabbing coffee from a barista you’ll never see again, you just want the order right, not a relationship. The exchange starts and ends at the counter. Whether it gets you was beside the point.

That era is closing. AI is moving into the cubicle next to yours. It’s going to help you write, decide, plan, sell, and think, all day, every day. And the moment a tool becomes a teammate, a new question starts to matter more than raw horsepower:

Does it actually understand the person it’s working with?

The industry has spent years in a single race: make the model smarter. Bigger benchmarks, harder exams, more PhD-grade reasoning. It worked. The intelligence is here.

But intelligence was never the thing that made a colleague worth having. We’ve all worked with someone brilliant and impossible, right on the facts, wrong on the room. It turns out being the smartest one in the building and being a damn good teammate are two different skills.

So there’s a distinction worth getting straight, because the whole future of this technology sits on it.

Most “personalization” today is mimicry (i.e., the AI changes how it sounds). Pick a tone, give it a persona, and it does an impression of someone who knows you.

What actually matters is understanding, whether the AI builds an accurate read of who you are: how you decide, what you can handle, what you mean when you don’t spell it out.

This might sound like a semantic difference. we assure you it is not. It’s the difference between an actor doing an impression of a person and a psychologist assessing one. One is a voice. The other is a model of a mind.

We call the second thing "Human fluency."

Get fluency wrong and the smartest model on earth is still a bad fit.

Every generation of workers brings their own unique spin to the workplace (Gen Y, Z, X, etc.). There's a new generation showing up in the office. We call it Gen S (for Synthetic). It's as brilliant as Spok, with the same lack of emotional awareness. And, a system that misreads your preference for action will over-explain what you already know, or bury the part you needed in piles of minutia you don't need. It will misread your risk tolerance and give advice calibrated to the wrong person. It knows you're a unique person, without any experiential understanding of your person-ality. It’s just running a generic script with your name pasted on top.

Adoption, trust, safety, alignment, sycophancy, plain day-to-day usefulness, they all run through the same bottleneck. A teammate you have to constantly correct isn’t a teammate. It’s a stranger, a potential liability, disruptive force and overhead.

Here’s the strange part. We can measure almost everything AI does. Its ability for complex mathematics, its code, its factual recall, how well it reads a legal brief, speed, endurance, research taste...all have a public benchmark that vendors compete on.

The one capability nobody measures is the one that decides whether AI works next to a human being: how accurately it models that human. Every major platform now ships “personalization.” Not one of them can tell you how well it actually works.

You can’t improve what you can't measure.

So the field keeps shipping the easy thing (i.e., a better "sounding" voice) because the hard thing has never been scored.

Resonant is a research lab built on one conviction: the next leap in AI won’t come from a SMARTER machine. It will come from one that better UNDERSTANDS the people it serves.

We study human fluency, we measure it, and we build the applications that put it to work, to test our hypothesis live, to check real validity and reliability, not vibes. Honest answer on where this stands: it’s early, and we’d rather show the science than pound the table. What’s below is what we’ve built so far.