Host Intro (Gene Kim)

All right. I first met Michael Winslow in 2018 when he signed up to give a last-minute, five-minute lightning talk here at our conference. It was an amazing talk called "There Are No Side Projects," and it was one of my favorite talks of all time.

He was a senior director at Comcast and chose to become an individual contributor as a distinguished engineer. And I was so incredibly happy for my friend when he took a role at Amazon Music as their director of engineering, because it represented the realization of a lifelong dream to combine his two loves: technology and music.

So here to share his adventures is Mike.

Michael Winslow

ETLS, make some noise. Let's go. Let's get the energy up here, guys.

All right. That's me, Michael Scott Winslow. If you want to stay in touch with me, don't look up Michael Winslow. I'm one of the least popular Michael Winslows in the world. Michael Winslow is the guy from Police Academy that makes all the noises. So look up Michael Scott Winslow, and you'll find me.

I'm the head of client experiences engineering for Amazon Music, and like Gene said, it's kind of a dream come true.

All right, so the last time I was on stage was in 2022 at the DevOps Enterprise Summit at the time. I was lucky enough to be able to talk about my journey in tech leadership. Part of that journey I had explained at the time was, in the 2000s, I was a DJ in Philadelphia for about 15 years, but I kept that secret from my day job. In my day job, I was a rising tech leader.

Interestingly enough, around that time when I started talking about my past as a DJ and my love for technology, people started asking me the question, "Well, why don't you try to do both at the same time?" Somewhere around that time was when Amazon Music came along and had an opening, and I said, "I'm going to go for it."

In that talk in 2022, I alluded to a big change that was coming for me personally, but I could not talk about it yet. A month later, I became the director of client experiences engineering at Amazon Music.

Thank you. Thank you.

So I ask now, in the words of Jay-Z, please allow me to reintroduce myself as that director of Amazon Music.

All right. This is a lot of my team right here. Not only do I live the dream of being able to work in music, but I have teams all over the world: Mexico City, Bangalore, Los Angeles, our headquarters in San Francisco, New York, and of course, this is Amazon Seattle.

What I love about it: that picture there with the music, that was my first day in our offices in San Francisco. I was super proud of it. But even though I went through probably about six months of interviewing to become an Amazonian, I thought that I would know a lot of what I needed to know about Amazon Music, the product, and the company. What I found out when I got inside was I just didn't know that much about it at all.

And I assume, if I didn't know that much about it after six months of interviewing, that a lot of people in this audience don't know that much about Amazon Music. So, with the help of our audio visual department, both back at Amazon Music and the wonderful, excited people backstage, there's a video I'm going to show you that might give you a better idea than I had about what Amazon Music is really all about.

But it's a very energetic one. So if you don't mind, I would ask all of you to please stand up, because it is time to really get excited on this one. This is music, y'all. Music and technology.

All right, in three, two, one. Make some noise, y'all. Let's go.

[Amazon Music video plays.]

All right. Anybody surprised a little bit? Is that a little bit more than you expected from Amazon Music, the player that always kind of bothers you and tells you you're a Prime member, so you should be joining, right? I'm not responsible for those emails that go out, but I know that they do go out.

All right, so let me just talk about a few things and recap some things that were in that video.

Amazon Music is more than just the player that you have on your mobile device or the web or even the TV. Amazon Music Live: you saw a lot of examples on there of the live concerts that we do. I'm excited because now Prime Video does Thursday Night Football. Right after Thursday Night Football, starting in October, there's always a live event put on by Amazon Music that you can watch right after the football game. They always tape it down in the Los Angeles area. I try to get down there to see some shows live. It's a fun perk of being there.

There's been three seasons so far. We're in season four. Concerts: Outside Lands in San Francisco, Fuji Rock in Japan. We broadcast live from a lot of these concerts as part of a way to create content on both Twitch and in our product.

And then I want to talk about the artist relationships, because when I say that this is my dream job because it mixes music and technology, part of that is the idea: if I worked with another company, I probably wouldn't see music stars around the hallways of the office and things like that.

So I have a story that I'm pretty proud of. January 18th, 2024, I remember the date very specifically. We had an all-hands in Los Angeles, and this young man came, an unknown artist came and performed for us. He basically said, "I'm going to sing this song for you, and the moment I sing this song, I'm going to turn around to my laptop and I'm going to release it on Amazon Music." That young man: Benson Boone. And that was the day he released "Beautiful Things," which ended up being the number two downloaded track in the world.

So I walked outside after that and I asked this kid, "Can I take a selfie with you, just in case you make it big one day?"

And then for all those folks in India, I have a special one. Maybe you'll see this next artist and know who it is. Does anybody know who that is? Is it Udit? That's right, yes. In 2023, he was the number one streaming artist on Amazon Music's platform. I met him in Singapore. It was fun. These are the perks. This is why it's a dream job.

And for me, on top of that, we tackle really tough technical problems. So I'll talk a little bit more about that.

Three of the things that were in that video that I want to go over: I'll show them high level, and then I'll go on. We are an all-audio experience now, meaning music, podcasts, and audiobooks. So we have Audible in the app; I'll talk about that in a minute. Another thing we delivered was something called Amazon Music Delivered, and I'll go over that. And then Maestro is our AI-powered playlist generator. Let's talk about those things.

Audible: in case you didn't know, right now, if you are an Amazon Music Unlimited customer, you get one free checkout of an Audible book per month to be able to use within the app. You don't have to jump outside of the app. You just do it. I think I was somewhere and somebody was mentioning "Atomic Habits." That was my last listen, and that's a great book if anybody wants to see it.

All of the streaming platforms now do podcasts and audiobooks, but how many of them actually work in the same company as the most recognized audiobook company in the world, and can say that you have Audible within your application? One of the perks of being at a large company and delivering music and spoken word.

Amazon Music Delivered: this was something that was far overdue that we delivered at the end of 2024. Other streaming platforms basically told you what your listening for the year was like, and people look forward to that at the end of the year. Our teams came together and did an amazing job of delivering this.

Now, this was my actual 2024 Delivered. I turned down the sound so I can talk over it, but you can get an idea of what my 2024 looked like. It was polluted a little bit with my teenage son, because we listen to music a lot together.

But that's a good point about it, because data pipelining was a huge piece of being able to figure out what people listen to. In 2024, we have profiles on our application that separate out children from adults, because we all know that if you were on the same profile as a five-year-old, everyone's number one song would be "Baby Shark."

All right. So we launched it for the first time in 2024. It's all about having that clean data. We found out what your top genres are, which tracks you listen to by different dates throughout the year. And if you want to get your 2025, you have to start building up your listening habits now. So please download the app and start listening for your 2025 recap.

All right. And then, you know, we cannot escape AI at this conference, right? This is our first venture into AI in the app. We launched the beta in April 2024, and then have kept it going and made small improvements while we're working on other AI-powered things.

Now, in order to show how this works, I didn't want to just do it myself. I actually brought in the help of someone. So put your hands together for our special guest who's going to create a playlist. Marguerite.

I'm looking for music suggestions to use at Enterprise Technology and Leadership Summit. One of my favorite songs to use is "High Hopes." Can you suggest a few more like that?

Okay, so now we did this yesterday, and if anybody wants the playlist that came out, please go ahead and grab this. This is the playlist that Marguerite created yesterday. You'll see that it's going out, it's using personalization plus the prompt that came in to decide what the right things to come out were. I'm sure you heard several of these songs throughout this conference. This is typical high-energy stuff. And so anything you want to do, you're in a bad breakup and you want a soundtrack to that, please go ahead and use this to create your playlist.

All right, so that's how we deliver AI to the customer. What about in-house? What about our engineers? Clearly we have to be using AI to make ourselves better as well, right?

Well, the first thing we had to do was find out which one of our engineers are actually using AI tools. We are inside of Amazon. We have access to all the cloud models, the Claude models, everything, but we want to know exactly: are they actually adopting it? So how do we find out who's adopting, who's rejecting, things of that nature?

Since we are in-house, we are lucky enough to have the Q Developer team create an entire dashboard for the entire company where you can break down by manager how much adoption you're getting in there. It breaks down what's your total addressable market of engineers, what's your adoption rate, what's your churn rate, and how much do they use it on a daily basis.

One of the great things that this caused, and you can see that this was a huge jump once we realized we were being measured, that we jumped up and we could do things like that. Now, one other thing that ends up happening is you create a sense of competition with other managers. So that's mine on the bottom, and then that's my peer on the top. I was a little jealous, I've got to be honest, when I saw that 92%, and I came to the team and I said, "You know, we can do better." But then I was explained that he may have a higher percentage, but our people actually rely on it on a day-to-day basis more. There's more usage per engineer in ours. So I feel like I have him on that one.

All right. So you might think, if all you're worried about is measuring adoption, you could fall into the classic Goodhart's Law, which states, "When a measure becomes its target, it ceases to become a good measure." I like to look at it on the flip side in some cases because there's also an effect called the Hawthorne Effect, right? That says that performance increases in individuals who are noticed, watched, or paid attention to. If you think about it, it's two sides of the same coin. One can create enhanced engagement and participation, and the other one is about gaming the system. It's about watching and monitoring those things very closely.

All right. So I want to give a quick example of a way that we actually benefited. Once we knew we had the adoption, now we had to find anecdotes of who's using it in a good way.

I remember a post, and I can't even say who the person was. I don't want to; I didn't get their permission. But they put a post out on LinkedIn and they said, "Hey, using GenAI for code reviews has been a game changer." And I went to another meeting with one of my engineers who lamented to me that he couldn't get his code through because a senior dev was on PTO.

I scheduled a meeting with one of our build managers, a build engineer, and I said, "Hey, are we using GenAI code reviewer?" And he started using it. I was happy that he finally did it when he was prompted to, but maybe he could have done it sooner. Within a week, we started seeing value.

This is a blow-up of one of his things: "I've absolutely been blown away by the positive improvement since adding Auto SDE" -- that's internal; you don't have access to it. "I've had multiple people reach out personally and several more on a previous thread about how it provided good actionable feedback."

Now here's this next quote you'll see is a way that you as leaders can keep having conversations with your engineers and can have an effect on them. He said later, "I naively thought I would be telling you things in our one-on-one, but I'm glad you asked me this question to trigger me." This was something I brought in from the community, asked our internal people about, and caused some change. So I'm pretty proud of that and excited.

All right, so now our engineers are using it. We know how to measure it. What about the leaders who are pushing for all this AI movement? Are they actually using it? As I speak, that is real context that I have in the background. I'm going to move on from it and explain what it is, but it's a surprise for Gene.

In order to get our leaders to be excited about it, I had to show them a simple tool that we had that could be effective. We have a tool internally, and it's used externally, called PartyRock. With that, I'll tell you about what PartyRock can do.

I gave it this context because I was so excited about my days reading The Phoenix Project and The Unicorn Project. I said, "I wonder if I could turn my day-to-day activities into small chapters of The Unicorn Project." And so I told it: create a story for complex documents featuring Brent or Maxine. These are for doc reads that we have within Amazon all the time. I just want to upload the document and I want you to create these characters. The expert character should guide the frustrated character through a technical solution. The story must have three aha moments. It needs to be set in the Phoenix Project universe. It has to include real data and calculations, and add some coworkers that come in every now and then as a subplot. And then always end it with, "I learned all this from reading this doc."

It ended up being a precursor to my doc reads all the time. It got me excited about the doc, and it told me what I should be expecting when I actually open up this doc and start reading.

Since I can't use internal docs, this was a public DNS science paper that I had in 2019, I believe, or something like that. I threw that into the engine. If you pay attention to this line right here, "Our primary objective is to connect people with knowledge resources, like frequently asked questions and how-to documentation." This was the primary purpose of the doc that we created back then.

So it created this story, and as you can see, it starts, "The Tech Adventures of Brent and Maxine: The Bot That Saves Support." If you look at this as it's set up, Maxine's frustrated about something and she says, "I swear if one more person asks me to walk them through something that's literally documented on our FAQs, I'm going to lose it."

So it basically goes in and creates a whole story and makes me feel like I'm in The Phoenix Project or The Unicorn Project, and I get pretty excited about it. Then at the very end, like it said to do, Brent smiled and pulled up the document on his screen and said, "I learned all this by reading this doc." And then I go into reading the doc.

Our leaders thought this was so cool. What a cool, quaint way to use AI, right? But really, I was just using this as a lure to something that's even more productive than this, and saying, "You could be using this AI, and you can use it for more than just silly games like this." Which I don't think it's silly, Gene. I think it's very important.

Once they get in, I show them my library of all the things created in PartyRock. You can go to PartyRock.aws and make your own. Sure, there's the tech stories one, but wait until I show you this stack flow mapper. I'll show this to you quickly.

In this one, I take all of the chatter in a Slack conversation for a team for the past three months, and I said, "Create me a value stream map of the conversations that are happening in there. So if you see something that looks like it's non-value-add, throw it in the list as a non-value-add. If you see something that looks like it's productive, throw it in there as value-added or blocked."

This is an example, and we're real close to the end. I took out everything that needed to be taken out. This is an idea over one week during that time where you can see the lightning bolt meant it's value-add. The sand timer is waste, and the wrench means it's non-value-add, but necessary. This allows leaders to go in and take the actual chatter that's going back between engineers and get an idea on a week-to-week basis: was this a good one? You can see at the top the impact score it gave it, and then it gives it a summary of why it gave it that impact score.

All right. So I want to leave you with this quote: "The world is changed by your example, not by your opinion." That's why I was saying that it's important to really get in there with your engineers, be a leader by example, use the AI, and then encourage them to use it as well.

And what I need from you: go create a playlist. Please create a playlist. Share it with me directly. You can DM me on LinkedIn and share it with me. Any feedback that you have about the app, I'm your inside guy. Please let me know about it. Thank you, ETLS.