Innovating Leadership:
Co-Creating Our Future

Resources:

Learn more about the new delivery technologies Beryl mentioned at https://bit.ly/DeliveryTech. To discover more about Amazon’s robotics, check out https://bit.ly/RobotsAtAmazon.

Our host Maureen Metcalf posts a newsletter every week on LinkedIn. You can subscribe here.

Maureen’s latest book is Innovative Leadership & Followership in the Age of AI. You’ll find details about it at https://bit.ly/LeaderInAI, or check out the Kindle version at https://amzn.to/44buVz8. The audiobook version is now available at  https://amzn.to/4dTCleZ.

Her other 10 books are available on Amazon here.

Related episodes you’ll enjoy:

• Amazon’s Innovation Secret: Look to Failure for Success with Beryl Tomay
• Delivering the Future: Why Amazon Execs Lead Beyond Retail with Kara Hurst & Ryan Redington
• Four Key Lessons from Amazon’s Head of Family Trust with Catherine Teitelbaum
• How Moonshots & Robots Put Packages on Your Porch with Steve Armato

Guest(s):

Guest(s) Bio:

Aaron Parness works as a Director of Applied Science in Robotics and Artificial Intelligence at Amazon. His teams in Seattle and Berlin build robotic work cells to increase delivery speed and reduce the cost of order fulfillment for Amazon customers. Specializing in high contact and high clutter applications, he has led advances in giving robots a sense of touch by incorporating force and torque sensors into the robots’ motion plans and control loops. From 2010 to 2019, Aaron worked at NASA’s Jet Propulsion Laboratory, where he founded and led the Robotic Rapid Prototyping Laboratory specializing in grippers and wall climbing robots. He received his PhD from Stanford in 2009 advised by Mark Cutkosky; and earned a BS in Mechanical Engineering and BS in Creative Writing from MIT in 2004.

Beryl Tomay has been at Amazon for 20 years having joined in 2005 as a Software Development Engineer. She was part of the small team that launched the original Kindle device and remained in the Devices organization for the subsequent 8 years. She joined the nascent Last Mile organization, the logistics business that gets packages through the final steps on their way to customers’ doorsteps, in early 2014 as one of its first employees. Today, she is responsible for Amazon’s Last Mile delivery technology and businesses as well as Amazon’s customer delivery and returns experiences. Prior to Amazon, Beryl received her undergraduate degree in Mathematics and Computer Science from the University of Waterloo in Canada. Beryl and her husband love going to Kraken hockey games, walking their dog Luna, and visiting new and diverse restaurants.

Transcript
(auto-generated)

Maureen: [00:00:00] Welcome to Innovating Leadership Co-Creating our Future, where we explore the evolving landscape of leadership and the bold ideas shaping tomorrow’s world. Your hosts today are Maureen Metcalf, myself, the founder, and CEO of the Innovative Leadership Institute, and George Limbert, our new special events co-host who’s held multiple C-level roles in construction and hospitality.

In this episode recorded as part of Amazon’s Delivering the Future event, we’re joined by two transformational leaders who are shaping the future on how we move both goods and ideas. First we speak with Aaron Parness, director of Applied Science at Amazon Robotics. Aaron’s work is at the forefront of intelligent automation, including the development of Vulcan, a robot with a sense of touch.

Then we’re joined by Beryl Tome, vice President of Transportation at Amazon, whose team is re-imagining the delivery experience through innovations in logistics, smart glasses, [00:01:00] solutions, and human-centric systems. Together these conversations reveal a powerful truth. Technology is only as transformative as the leadership behind it.

Welcome. Aaron Parness, director of Applied Science at Amazon Robotics.

George: Aaron, thank you so much for being here today. Thank you so much for talking to us and taking some time off to tell us a little bit about the new Vulcan project. Why don’t you tell us how it interplays and is gonna make life better for the experience of the Amazon employee?

Aaron: Yeah, thanks for having me. Vulcan is Amazon’s first robot with a sense of touch. What that enables the robot to do is interact with. More cluttered environments and understand physical contact. So the first two applications we’re using Vulcan for are stowing and picking items off of our bookshelves.

They’re big yellow kind of canvas fabric bookshelves, and they hold all the items that are available for sale on amazon.com. Think about adding a book to a [00:02:00] bookshelf that already has a bunch of books on it. You have to push the books and you have to kind of interact with that clutter to make space and to add that book you wanna add.

It gets harder, of course. ’cause they’re not all books, it’s everything Amazon sells from dog toys to bowling balls. And so we are giving the robot this sense of touch so it can interact with that cluttered and, and high contact environment. The way it makes life of our employees better is we are targeting the top rows and the bottom rows of those bookshelves.

Currently, employees have to use a ladder to get up to those top rows or crouch down to get to those less ergonomically friendly rows of inventory. And if we can leverage the robots for those rows, it allows the employees to work in their power zone. So the employees are both safer and more productive.

They’re able to do more work. And so it’s a win-win.

George: Wow. Safety first.

Maureen: So for sure, people like me, this really aim. Increases their probability of success.

Aaron: It

Maureen: does, yes. So you’ve led teams through complex innovation cycles from [00:03:00] NASA to Amazon. So you’re a NASA engineer.

Aaron: I was, yeah. For 10 years.

Maureen: Yep. So what leadership practices have you found most effective in fostering a culture that embraces experimentation?

Aaron: Mm-hmm.

Maureen: Failure and rapid iteration at scale.

Aaron: I am a huge believer in iterative design. You definitely have to coach, especially junior employees that have have joined. I hire so many overachievers and they wanna get the problem right. Because that’s what you learn in school. You gotta get the problem.

Right. And with iterative design, the idea is not getting version one. Right? It’s getting version seven. Right. And the way you do that is by getting version one out into the field to figure out all this stuff. You weren’t even. Thinking about all the questions you had that you didn’t even know were important.

So one of the things I do is push us to go fast with the first prototypes. I would say the second thing that’s been important to me as a leader is making sure everyone knows where the end zone is. Like everyone knows and has a common understanding of what winning [00:04:00] looks like. Um, sometimes you get, you know, six or 10 or 11 players on the field and they’re all going different directions.

And as the leader, you can remind everybody with the consistent message that. Here’s our milestone or here’s why we’re doing this. And you see this boost in like collaboration after you do that. ’cause you align everybody’s incentives and that’s been really important.

Maureen: In a prior interview, we talked to somebody else, not you, about one way doors and two way doors.

Oh

Aaron: yeah.

Maureen: Share with our listeners what that concept is.

Aaron: Yeah. This is one of my favorite Amazon phrases. A one-way door is a decision that is really hard to unwind. It’s something like you’re purchasing, you know, a big piece of equipment. It’s not easy to return. That piece of equipment after you purchase it.

A two-way door is something that if you realize you made the wrong choice, it’s easy to go backwards and try another option. And so software can be this way at times. Like you can roll something out and we do what we call an AB test. So you test version one and [00:05:00] you test the control and if it doesn’t work, you roll back.

You wanna go really fast through two-way doors. It’s low stakes and you wanna stop and you wanna kind of make sure everybody’s aligned when you have a one-way door decision because it’s much more difficult to unwind.

George: So a lot on your leadership journey, I think the culmination it seems like is to create a robot that can have a sense of touch.

And I understand this may have been an idea of yours for a very long time. So tell our, our leaders listening today, how that journey went.

Aaron: We have been working on it for a while. It started with three or four of us in the back corner of a lab. And in fact there was a recent PhD graduate who was the first one that said, oh, we wanna try and stow in these densely cluttered bins.

And my initial reaction was, ah, how naive. This junior roboticist doesn’t know how hard the real world is. But it was interesting problem. I like hard problems. I, I wasn’t as motivated then about like, oh, we have to go solve this, you know, enormous business critical thing. And so I got [00:06:00]involved and what we realized is we made a couple of innovations, we did some experiments, and we realized rather than putting the whole robot hand into the shelf, you could put these paddles with conveyor belt and feed the item into the shelf or pull it back, and you kept control of the item as you were doing that.

It’s like, oh wow. Light bulb moment. We got through one or two of those. And I flipped completely. I went from, we’re gonna learn the reality. Robots are not good enough to do this, to, oh my gosh, this is robotics 2.0, this is the future. And we tried to go quickly and iterate and go through those prototype cycles.

And so the team grew from three or four in 20, 20 to 10 or 12 and 20, 21, and I think we’re between pick and sto. My team’s in Seattle and Germany we’re approaching 300. Today we’ve got robots out in real fulfillment centers in both Germany and the United States. And it has been wild and super fun. I tell people I’m living my robot dreams.

Um, but it keeps being challenging too in new ways. And you know, scaling a robot is [00:07:00] just as hard as building the first proof of concept.

George: I have to ask the Vulcan name.

Aaron: Yeah,

George: the origins. There’s some people that. May think it came from the Star Trek world.

Aaron: Yeah, we get that a lot. It is not though Vulcan is the Roman god of the forge.

He’s a builder and we think of ourselves as builders and so it is more about making stuff. Yeah. Awesome.

George: Very cool.

Maureen: Tell us a little bit more about the complexities of scaling.

Aaron: An Amazon building a fulfillment center can process 1 million orders in a day, and there are more than a hundred of those buildings across the United States.

And so you think of this long tail of things that can go wrong if something happens. One in a million happens every day in every Amazon building, and so the challenge of scaling, it’s not to solve all of those, but how do you make the system robust so that. It can automatically recover if it has a problem or it can expose the right information when it has a problem so that your maintenance engineer or your [00:08:00] technician is able to quickly get it back online.

Those are some of the problems of scaling, just getting repeatability, getting a robot that’s dependable for our operators. When we in Amazon Robotics, think of our customer, we usually think of those frontline workers and the people that are operating the fulfillment centers. And if we can make their jobs better, they make the Amazon customer better, who’s buying things from our storefront and we get a lot of feedback from them.

And, and part of scaling is listening to that feedback and adding and adapting and making sure your product is lovable by your end user. Yeah, it is a lot.

George: As you were thinking through the lovability mm-hmm. Of Vulcan in particular to enhance the employee experience, what are some of the things that may be.

The typical person out there, the typical leader that listens, wouldn’t necessarily think that you had to do to make it easier to utilize in a workplace.

Aaron: There are some big ones like using the top rows and the bottom rows that we talked about, and that feedback came super strong. It’s by far the most lovable feature.

[00:09:00] And then there’s some other ones that you think like, oh, that’s obvious, and yet. And yet, so as we were building the robot, there are actually three robots connected and we gave each of them their own touchscreen interface. ’cause as a builder, you’re like, oh, this is modular and this makes it easy. And then you get out in the field and they’re like, I hate having to go between the different monitors to figure out like what I’m doing with this system.

We want it in one monitor, one user interface. And you’re like, of course you do. Um. We’re gonna fix that. And so that’s one of the things that user feedback you get also, there’s just, people are so different and so how people interact with the robot, you’re always learning new things, sometimes conflicting things, and you have to try and balance, like someone who wants the logout button in the upper left versus the lower right.

And you know. One thing that’s really important there is consistency. So Amazon Robotics has a fleet of robots and so we’re always talking with the partner teams to make sure that user experience is consistent and folks don’t have to learn, you know, seven different user interfaces. Those are some [00:10:00] of the things we work with them on.

George: Always listening to the team members in their experiences.

Aaron: Yeah, we have a whole section in our. The business reviews called Voice of the Customer and we’re getting quotes from those end users and diving deep on what went wrong or what they really liked. Uh, you get both the positive and the negative signals from that information.

Maureen: Great. So that consistency feels like street signs around the world. Mm-hmm. And. Rental cars. Yeah. Gas is always in the same place.

Aaron: Yeah, that’s a good example. Yeah. You also want intuitive signage and intuitive user experience. So even if someone is new to Vulcan, but they’ve got experience with other robotic systems, they can learn it quickly or it feels familiar.

I’m writing my instructions. I expect everyone to read every last word of that instruction manual that I’ve put all that effort into writing and you know. We don’t, we don’t do that. And so you gotta make those cues and those experiences for the customer just. Super [00:11:00] streamlined, easy, intuitive. We partner with a team.

Engineers are not always the best at intuition and, and user experience, so we, we partner with product design and ui ux folks to, to get that right.

Maureen: Does Vulcan play well with the other robots?

Aaron: Yeah, the. Thing we think about more than having Vulcan play well with other robots is play well with the people that are using it.

And those aren’t just the people directly interacting with it. So if we had to get Vulcan to handle every kind of item that Amazon sells, we would be waiting 50 years. You can’t go for a hundred percent coverage. We don’t believe in a hundred percent. Automation or, or robotics. And so making sure that workload is shared in the right ways and collaboratively between the people that are working, not just with the system, but downstream of the system.

Upstream of the system. That’s where we spend a lot of time making sure it plays well. There are some things that matter. So we have a robotic system that like [00:12:00] applies the shipping labels to packages. If it does that poorly and then later on we grip that package and you rip the shipping label off. I say this ’cause that was a real problem.

That’s a. Hey you. You two teams need to talk and sort this out, but more often we’re trying to make sure we fit into this symphony of a warehouse where so many different things are happening. Many of them with human workers and some of it with automation. You’ve gotta get that whole system to play altogether.

George: Great. Thank you very much. We really appreciate the time today.

Aaron: Yeah, it was fun. Thanks for having me. Thank you, Aaron.

George: Our thanks to Aaron. Now let’s join our conversation with Beryl.

Maureen: Welcome to the conversation with Beryl Tome. She’s the Vice President of Transportation of Amazon Beryl. Let’s talk a little bit about your role and the driver Academy, the glasses and wellspring.

Beryl: Yes, thanks for having me. We’re in the Bay Area at our [00:13:00] Delivering the Future event, and I was able to announce big news, which is the glasses.

So what these are is our latest innovation for delivery drivers. They’re designed working back from enhancing driver safety, comfort, and experience. Effectively, what they are is they use. AI and computer vision to be a companion to drivers in their journey of delivering packages from the delivery van to the customer doorstep.

And they have a heads up display that will display information in their field of vision so they don’t have to use their phone or interact with their phone. And that allows ’em to be hands free, which enables them to really. Focus on their surroundings and be present in the moment, which is the part that enhances safety.

So just to walk you through and give you a sense of how they work. When the van is parked, the first thing the driver has to do is find the right set of packages for that delivery stop. Now, all they have to do is glance around to the packages in the vehicle and the glasses will tell them, okay, [00:14:00] here are the ones you need to grab for this particular location.

They grab those and it’ll then help them with navigating to the. Doorstep. This is actually pretty important for complex locations like apartment buildings where it might not be trivial to know where to go. So it shows building outlines. It shows the navigation path, what direction they’re facing with respect to where they’re trying to get to.

And once they arrive at that location, they drop off the package and just look at it, which scans it and takes the delivery photo. If you’ve ordered something, anything from Amazon, those delivery photos that you see in the your orders page that gets snapped and then they return to their vehicle. We’ve been trialing this with hundreds of drivers over the past few months and incorporating their feedback from anything around the hardware to the user interface, all aspects of it.

The feedback that we’ve gotten has been very positive with drivers saying it’s been. Game changing, enhancing their safety, making them feel more comfortable in the [00:15:00] environment and improving their experience, saving them interactions with the phone.

Maureen: One of the things that it mentioned this morning was alert you of dogs.

Beryl: Yeah.

Maureen: I have a friend who was bitten by a dog terrified of rogue dogs. You are able to log things like that to keep your drivers safer, right?

Beryl: Yeah. One part of this device that I’m really excited about is hazard. Detection and display capabilities that we’re gonna build into it with more and more hazards.

One of those is dogs, so it’s able to detect the presence of a dog, alert the driver, but just as importantly, it’ll be able to store that in our systems to alert and warn. Future deliveries to that location by other drivers. It’s even broader than that though. You know, you can envision use cases like uneven grounds where drivers might slip, trip, and fall, and that’s not specific to a delivery driver, any of us.

Mm-hmm. You’re walking around. If you don’t pay attention or you just miss [00:16:00] it, you might. Slip, trip and fall. It’s a safety incident. So I’m very excited about the potential for this technology to be able to detect these hazards and build up our hazard system to be able to show it to more drivers.

George: As you are exploring and training the drivers for the different hazards that they encounter, how do you go about providing that training so that they’re prepared in those moments?

Beryl: Yeah, we have a very comprehensive training program called Last Mile Driver Academy. We’ve been expanding it quite a bit. It has a mix of traditional classroom training, and what we find really effective is the virtual reality simulations and the slip trip and fall. Course, it’s an actual obstacle course that I’ve taken myself.

It was difficult, but I passed it. That really prepares ’em for the real world without any kind of risk. Even the slip, trip and fall course you’re attached to a harness with that’s connected to a beam up top. So if you actually fall, it catches you and you’re safe. But it prepares [00:17:00] you for slippery surfaces, water, uneven ground obstacles in your way.

So we’ve helped train over 300,000 drivers with this training. One data point that I’m really proud of is we’ve seen a 15% reduction in slip, trip, and fall incidents, and 93% of drivers have reported feeling more confident before they go out on the road to deliver.

Maureen: Wow,

George: great statistics. Pretty amazing.

Maureen: Yeah. Amazon’s transportation network clearly evolving rapidly from smarter routing to AI powered logistics. All the stuff that you oversee with your colleagues. So for senior leaders, what lessons can they draw from Amazon’s approach to scaling innovation? And concurrently maintaining reliability.

Beryl: I’ll say a couple of things.

We do take a lot of big bets and we look at moonshot ideas, which was, you know, part of the genesis of the smart glasses. It’s something we try to do continually, where we [00:18:00] force ourselves to think, how do we think outside of the box here? What crazy ideas do we have while recognizing that there is inherent risk here and.

Not a lot of them, and even sometimes most of them are not going to work out. But if you’re not willing to take those risks, you’re not gonna invent the really important critical products and we’re okay with failure. We’re okay with some of these things and expect some of them to not work out. I think that’s one.

And not giving up too early when we actually take those bets. The second thing that we do quite a bit in our operation, and I actually, I’m out on the road a lot, is I’m. Traveling to different delivery stations, doing ride-alongs, doing deliveries. Myself, I was in Finley, Ohio earlier this year. I went deliberately when it was very, very cold, just to be able to experience what dsbs and drivers experience in those conditions, because it’s vastly different from what I experienced in my home in Seattle.

That is [00:19:00] also an idea generator, but getting close to our partners and understanding their experiences is a great. Piece of input into being able to innovate for the right things. So I think the combination of those two things is kind of a necessary part of it to be able to do all that.

George: Clearly, safety and the drivers, giving them the best chance of success is very top of mind and you can see or hear the passion in your voice as it relates to that.

Where do you get that inspiration from as a leader?

Beryl: From a personal motivation perspective, I don’t think there’s anything more gratifying than knowing. People are safe. We all have jobs that we do. We wanna keep communities safe. We wanna put the weight of Amazon from an innovation perspective behind these delivery service partners to enable them to really focus on running their businesses, building amazing teams, and delivering for customers and doing all that.

Being as safe as possible is just super, super heartwarming. And that’s probably the intrinsic [00:20:00]motivation. Safety for safety’s sake, right.

Maureen: Yeah, we were talking yesterday about changing economic landscape and how some manufacturing jobs are going away. Some of those jobs were dangerous. Amazon is filling the gap and creating another path to middle class.

And those jobs are a lot safer listening to the commitment to continue to make them evolutionarily more safe. So as people age, it’s not just 20 year olds delivering packages, but people can continue in a role as their life evolves.

Beryl: Definitely. We’ve introduced robotics into our operation over 10 years ago.

And robotics and automation really helps with repetitive tasks, manual labor, like lifting heavy things. And when we’re able to move that focus to the robot, which acts as a companion to the associate, they can then focus on. Other tasks that are [00:21:00] just better for their safety, better for their experience.

So that’s also very compelling. The thing you said really struck me about aging populations. I was in Tokyo, Japan last year, and every time I do these visits to our delivery stations, I meet with a local group of delivery service partners, and we have the DSB program in 21 countries. Japan is one of them.

And the first question from the group was something I’d never heard before, which was. We have drivers that are older. What can you do for them to make their experience better and easier? And I never gotten that question before. I’ve talked to dozens of delivery service, over a hundred delivery service partners, and that was really heartwarming because I thought just the fact that they’re asking that and they care about that is incredible from a cultural perspective and how they’re thinking about their teams and employees’ perspective.

That was something that stayed with me. It was over a year ago, but I think it’s a, it’s a very noble thing and something we should all strive to look at.

Maureen: As we look at changing [00:22:00] demographics, it seems like it will become more important.

Beryl: Yeah, especially in some locations too. I mean, it looks totally different around the world.

There’s different considerations everywhere, but the more we can do for safety, comfort, experience, I mean, there are things we’re gonna keep investing in.

Maureen: Beryl, thank you. It is lovely to, again, get to talk to you and hear how Amazon is evolving and driving toward safety and being an important part of our business ecosystem.

Beryl: Thank you so much, Maureen. Love to chat.

George: Thank you.

Beryl: Thank you.

George: Thank you all for listening today as we took a look at delivering the future and all of the day-to-day that occurs. If you enjoyed this podcast, please be sure to like, subscribe, and recommend the podcast to others. Thanks again.