Hi everyone. My name is Ryan Powell, and I lead design at Waymo. Years ago, I was riding my bike through this intersection in Chicago when a car ran a red light and hit me.

So, I survived, obviously, but that moment stayed with me, and it redefined how I think about our streets.

Every 26 seconds, someone else's story doesn't end as lucky as mine. So, every 26 seconds, that's how often a life is lost on our roads.

When we talk about building autonomous vehicles at Waymo, we aren't designing a cool tech novelty. So, recently, the medical community published an open letter calling out America's road deaths for what they truly are, an ongoing public health crisis.

And for us at Waymo, bringing this technology to as many people and places as possible is a public health imperative.

So, back in 2009, uh Waymo started out as Google's self-driving car project, and for the longest time, it was viewed by the public as an eternal experiment.

But today, I think we should officially retire that notion. By a show of hands, how many of you have taken a ride in a Waymo?

That's incredible. So, I think we can all agree that autonomous vehicles are becoming a daily reality.

>> [applause] >> So, Waymo has driven over 200 million rider-only miles. We've completed more than 20 million passenger trips, and every week, we are serving more than half a million rides across 11 cities in the US.

And at the moment, we're currently preparing for our our international expansion into the streets of London and Tokyo.

So, behind every car in every city, there's a massive operational backbone that you don't see. So, in addition to the rider facing work that I'm going to share here today, we also design the data-rich interfaces that empower our operations team.

So, from our mission control dashboards that monitor fleet health um in real-time to the apps that our depot techs use to charge or maintain our fleet.

So, we're also designing the internal tools needed to bring this technology to more people and places.

So, that's all to say, this is no longer a pilot.

People are depending on us to get from point A to point B, and our safety data is very convincing. So, compared to human drivers, our vehicles have seen a 92% reduction in serious injury crashes.

But, here's the design problem. Safety is statistical, but trust is emotional.

Our riders don't think in terms of statistics. They want to get from point A to point B in a very convenient and comfortable fashion.

And so, if we don't understand their needs in an autonomous context, the underlying technology, something doesn't matter.

So, let's rewind just for a second because getting here required a massive product pit uh pivots, and that was largely driven by user research.

Back in 2012, we were testing a Lexus fleet, and it was equipped with early driver assist features. It's very much like what you see today um on vehicles that we have with where there are driver assist features.

And it was a driving system that still required the human to take over at a moment's notice.

So, our researchers uncovered something unsettling. The moment the technology worked reasonably well, users almost immediately over-trusted it.

So, testers were supposed to be monitoring the system, and again, they were supposed to be ready to take over at a moment's notice.

But, as you can see in this video, they were doing all sorts of things at highway speeds.

They're texting, they're putting on makeup here, and we even had some people completely nod off as again, as they were on the freeway and speeding along.

So, we've completely abandoned at that at that moment in time, we did we drew a line in the sand, and we said, "Okay, we're going to not pursue this idea of going with driver assist as a stepping stone to full autonomy.

We're going to be go all in on level four autonomy, which is where the passenger never has to intervene." So, we went fully autonomous in late 2019 in Phoenix, and we spent a lot of time with our early riders back then.

So, they were tech enthusiasts. They loved the novelty. If the car took a slightly weird route, or it paused awkwardly, they forgave the friction because they wanted to be part of the future.

But, as we expand uh from our early adopter base, our new mainstream riders, they care less about the novelty of self-driving.

They have a low tolerance for self-driving quirks.

And so, I know a lot of you are out there and dealing with the same friction right now. You've built this incredible product.

Your early adopters love it, but now you have to scale past them to the pragmatists, so the mainstream.

These are users who don't care how cool the underlying tech is, and they just want to it's a work flawlessly.

And so, to to design for them, you have to stop designing to educate or to impress, and you have to start designing for trust.

So, how did we do that? Well, we spent a lot of time in the passenger seat observing the nuanced dialogue that's often unspoken that defines a human-driven ride.

So, much of that communication is direct. You might ask the driver, "Hey, you know, which route are you taking?

" Or, "The light's green, you know, why aren't you moving?" Um but, the real sophistication lies in the non-verbal cues.

So, consider the scenario that you see here in this photograph. You're in the backseat, the car begins to slow down.

Instinctively, you look up and you see two things. You see the jogger ahead outside of the windshield there, but you also notice the driver's head turn slightly toward them.

So, in that split second, you haven't just received information, you've built a shared understanding.

You know why the car slowed down, and you feel secure because of it. And that insights became a cornerstone for our design philosophy.

So, when you remove the human driver, you don't just remove an operator, you create a massive communication vacuum.

And so, our jobs as designers has been to design a proxy for that missing dialogue. So, we are essentially building a new language of trust.

So, early on, a lot of our riders asked, "What can our cars see?" Well, there's a ton that our car, you know, that our cars see and that we could communicate.

Some of you have probably seen a video like this.

It's what our engineers use for development. But, this raw output is just too dense and demands too much attention for people that are just trying to get from point A to point B.

So, as designers, we spend a lot of our time curating exactly what we show in our UI.

And we want our riders to have a sense for just how much the vehicle can see and understand without overwhelming them.

So, we've learned that building trust requires sort of a ruthless uh curation over complexity.

So, let's take a closer look at the passenger UI that you see during your ride. We want to Again, we want to show you as as everything that the car can see and how sophisticated this technology is, and we want to be able to explain its behavior when that makes sense.

So, the in-car UI has to also assure um people and and not overwhelm them at the same time, and it has to do so whether if it's their first trip or their 100th trip.

So, let's zoom in then to the 3D uh part of our UI, and here you can uh see that we show a road surface.

It's flattened and it's grounded. The features of the road like crosswalks and street markings are things that we highlight here.

The car's trajectory is a predictable green path, and then nearby vehicles are rendered as simple, clean, sort of stable 3D geometry.

When it comes to pedestrians, they have a crisp sort of light blue glow underneath them, and cyclists have a green underlay.

And when with pedestrians and cyclists, we got feedback early on that abstract icons felt too detached.

So, instead, we take a hybrid approach here. We layer the real-time laser points over the figures, so you can actually see the biological motion of their arms and legs moving, and this subtle design language communicates an incredibly complex message.

It tells the rider, the car doesn't just see that object, it knows what that object is, and it knows how it behaves.

So, occasionally our riders might glance up, they might wonder where they're at. So, in addition to showing what the car can see, we also pull in buildings to provide some geographical context, but because they're less important to understanding what the car can see, we sort of dim those down so that they don't compete with the more important elements in the scene, namely the the nearby cars and pedestrians.

So, when it comes to building trust, um sometimes the most valuable pixels are on the screens are the most ordinary ones. One of the highest yield design decisions that we ever made was rendering orange traffic cones on the screen.

I'm not kidding. >> [laughter] >> Construction zones are sometimes chaotic and stressful.

And when a rider looks out the window and they see a row of orange cones and they see those exact cones rendered in real time on their screen, they instantly relax.

We hear this time and time again in the user research that we do. So, that simple detail in our UI says, "I see that barrier.

I understand the detour, and I've got this." So, in addition to the 3D scene where we show what the car sees, we also have the status layer that sits on top.

So, this is where we can be more direct when we need to be.

So, it's essentially the voice of the driver and it communicates things like the trip progress and what the car is doing.

We might be slowing down in a school zone, and so again, you're going to feel that.

You're going to look up, kind of wonder what's going on, and you'll see that on the screen in the status layer.

Um in this case, we're letting you know that we're waiting for the intersection to clear before we we make this right-hand turn.

So, we've also invested a lot in sound design. So, when you design for trust in a deeply personal space like a vehicle, you quickly realize how important it is to respect that space.

So, independently, these are nice sounds.

But when we overuse sound, it's easy to hear how it can quickly become annoying. So, remember, in cars, alerts are mostly around for drivers who need to react.

But our cars are filled exclusively with passengers, and passengers don't need to manage the vehicle. We just want them to sit back and chill.

So, to achieve sound harmony, we lean into psychoacoustics. We've anchored our entire vehicle soundscape around the musical key of E.

So, in psychoacoustics, the key of E is uniquely associated with feelings of joy and clarity.

It feels bright and and inherently stable. And we use sound design as as a systemic hierarchy. So, for instance, when you open the door for the first time, you're greeted by a soft ambient harmonic soundtrack in the key of E.

Our team affectionately calls this spa music.

Early on, one of the insights that we learned about was that it was kind of eerie when you got into a a self-driving car and it was completely silent.

>> [music] >> And so, this soundtrack is meant to transition you from the outside world into one of our vehicles.

So, from there, every functional notification is musically paired. So, the chime to start your ride is a rising scale.

>> [music] >> Heading to Dolores Park. >> So, that rising scale denotes the onset of an adventure or the start of your trip.

And then when you get to your destination, the arrival chime is a falling scale. >> You're here. Please make sure it's clear before exiting.

>> So, that signals a gentle return or the fact that you've arrived at your destination. So, if a situation requires slightly more attention, like closing the door, the notification shifts deliberately away from the base key to create a mild polite tension that draws your eye to the screen.

>> [music] >> So, if you left the door open and we were trying to get you to close it, you'd hear that sound, you'd look up at the screen, and you'd see a notification there asking you to close the door.

The rest of the time, you know, we want you to lean back and enjoy the rides.

We're very judicious about how we use those notifications.

So, so far I've talked a lot about the rider experience inside our cars, but our design challenge doesn't end at the car window.

It extends into the living, breathing machinery of a city, where we design for the safety of everyone sharing the street.

In 2021, we graduated from the wide, sort of predictable roads of Phoenix into the dense urban environment of San Francisco.

And that's when we started dealing with 15 times more pedestrians in the neighborhoods and 50 times more pedestrians in the financial district.

So, we had to learn how to talk to the world outside the car, what our industry calls other road users.

So, every day on a city street, humans engage in these micro interactions. Who goes first? Do I cross now? Do they see me?

In a traditional car, you resolve that ambiguity through body language. So, if a driver's wheels are turned to the right, but their head is turned to to the left to check traffic, a pedestrian instantly reads that posture and thinks, "Okay, they're not looking at me.

I better wait." But when you remove the human driver, you erase the eye contact and the body language, and so we had to invent a design proxy for those human cues.

So, in robotics and animation, this is often achieved by motion. So, movement enables the inference of intentions.

So, we do this at four-way stops. It's a good example. So, the vehicle deliberate deliberately nudges forward to signal that, you know, it plans to go next.

It plans to take our turn. So, this little nudge is a way that we can kind of communicate using motion about in terms of what we're about ready to do.

But, in a dense urban environment like San Francisco, there are a lot of times where our cars are completely stationary.

So, we might be yielding to a lot of pedestrians or something that's happening on the street, and so we stop.

And how do you tell if the Waymo sees you and it's okay to cross when the vehicle is not in motion?

Legibility by design ft. Ryan Powell (Waymo) | Config 2026

Scripts