What Amazon’s AI Leader Thinks About the Future of Your Job

Is AI coming for your job, or just the parts that you hate doing? Today on Digital Disruption, we’re joined by Rory Richardson, Director of Go-To-Market for Generative AI at Amazon Web Services (AWS). With over a decade of experience at AWS, she has been instrumental in launching and scaling breakthrough technologies, including non-relational databases and serverless computing. Rory has played a pivotal role in shaping the future of cloud development through her strategic leadership in go-to-market initiatives, empowering developers to expand the boundaries of what’s possible in the cloud. Driven by a passion for innovation, her current focus is on harnessing the potential of generative AI to transform developer tools, enabling more efficient, creative, and human-centered development processes. Rory sits down with Geoff to explain how generative AI can supercharge the developer experience, helping teams write better code, faster, and with fewer headaches. She shares her perspective that AI isn’t about replacing people, but about helping us do less of what we dislike, and why embracing AI now is one of the smartest moves leaders can make. Rory shares real-world examples from AWS, insights from developer culture, and a human-centered view on the rapidly changing AI landscape. In this episode: 00:00 Intro 01:04 Recap of Info-Tech LIVE 2024 in Las Vegas 01:35 What’s new in AI & developer tools 02:53 The rise of AI-powered command line tools 03:32 Why Amazon is paying people to play 06:44 “Vibe Coding” 08:10 Differential equations & development tools 12:00 Risks of rapid AI adoption 13:32 What does coding look like in 2028? 16:20 Creativity, context-switching & grit 18:07 How play drives innovation 19:34 Teaching AI in schools 23:10 Is age the biggest barrier to AI adoption? 28:20 What should junior Devs expect? 30:38 Better AI boundaries 35:01 Will AI replace Developers? 37:25 Accelerating innovation, not replacing humans 42:24 Why AI works best on repetitive tasks 44:11 Let AI do the dishes, not the art 46:04 From Paris, Texas to Amazon Web Services 48:19 Everything is learnable 49:09 The grit factor Connect with Rory: LinkedIn: https://www.linkedin.com/in/roryr/ Visit our website: https://www.infotech.com/Follow us on YouTube: https://www.youtube.com/@InfoTechRG

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Geoff Nielson?

  Hey everyone, I'm super excited to be sitting down with Rory Richardson. She's one of the most fun and exciting voices in AI right now. What's so awesome about Rory is that while on paper, her specialty is how AI is transforming the way we code, every time I speak with her, we end up having these fascinating conversations that can go absolutely anywhere. When I sat down with her last year at the Infotech Live conference in Vegas, we talked about everything from the perfect toaster to using AI for Dungeons and Dragons. I do want to hear about what's going on with AI and whether it's going to turbocharge or kill coding, but honestly, I'm excited to see what else she's going to throw at me while we're at it. Let's find out.

• 0:44Geoff Nielson?

  Hey Rory, just before we get too deep into it, I wanted to tell you, I don't know if you remember from Live and in Vegas, talking to Joel, who's my boss, but he was, I don't know if you remember him, he was a huge fan of yours. And completely, completely true story, when we started this podcast, he said, Jeff, it's two people I want on my podcast. I want Bill Gates and I want Rory back.

• 1:08Speaker 01

  Okay. That's a little freaky, just saying, like, that's top billing, like, I have to admit, y'all's production for your events is like, next level.

• 1:20Geoff Nielson?

  Oh, thank you. Yeah, we're excited. We've got actually the next one coming up in a few months in June, but yeah, no, we have fun and we were so glad that you could be a keynote there last time. So, you know, I wanted to ask you on that note Rory, it's been, you know, it's been a few months since we last spoke, I think, like, since September, which is like, I don't know, it feels like we're living in dog years right now, both in terms of like, technology and society. So, you know, I wanted to just start off by asking, you know, in that time, like,

• 1:49Geoff Nielson?

  what's, what's new in your world? As you look at the world, what are you most excited about? What are you most worried about?

• 1:56Speaker 01

  Okay, that's a, first of all, I agree with the dog years, I feel like, and we used to joke, like, working at Amazon, you have to measure it in dog years because it's such an intense place to work. I've gone from dog years to like, fruit fly, alright?

• 2:10Speaker 01

  There's been a progression because you're right. Everything changes with generative AI every, every week. And especially in the developer experience, which is, it is wonderful. The progress that the technology is making is like every week is leaps and bounds. Different and more advanced. I would say the thing that, there's two things that have happened in my world that have made me super happy. One, we launched the Q command line interface that's using Sonnet 3.7.

• 2:49Speaker 01

  And it is fabulous. Like it's the week over week fastest, most adopted piece of the umbrella of Q developer. And the things that I see people that are able to do with it, it is absolutely astounding. I mean, anecdotally, my own team

• 3:11Speaker 01

  are building, you know, mini apps just with the command line interface. So no code, just using command line interface and they've decided to try to out app each other. Here, here's some examples of the apps they are building with the command line interface. Number one, a Rory decoder. They built an app to translate me.

• 3:33Geoff Nielson?

  Wow, wow.

• 3:35Speaker 01

  Another person on my team

• 3:39Speaker 01

  built a natural language interface for creating Dungeons and Dragons maps for his campaigns.

• 3:46Speaker 01

  And they're really cool.

• 3:48Geoff Nielson?

  So we're out with two examples, we already have the full spectrum.

• 3:52Speaker 01

  Yes, we're really improving workplace productivity over here. I mean, so the, and I think I, last time we talked, I talked about this, but I'm, I'm going to double down on it. We are paying people to play now because to play is the, is the highest cognitive load and the most creative and humanistic aspect of our work. And it's a shift, like culturally, because you know, Amazon, we pay for results.

• 4:23Speaker 01

  But our results are experiments, which means play. And that's a, it's, it's, it's a mind shift because the most valuable thing that humans do is something new.

• 4:36Geoff Nielson?

  Right.

• 4:37Speaker 01

  Like, nothing else can replicate that, not generative AI, not anything. So that's number one. Number two

• 4:45Speaker 01

  is there has been a shift in the market since we last talked that I think is super interesting, and that's, um, the rise of the IDE. So the IDE, you know, your developer environment has not changed in, I don't know, 10, 15 years. Like VS Code and JetBrains have been fairly static. The plugins haven't, but just the IDE has. And what we have seen is the rise of new IDEs that are optimized for AI. And that has been fascinating because one of the things that these new IDEs are really great at

• 5:22Speaker 01

  is collaboration. Now, collaborative is not the first 20 adjectives I associate with a developer.

• 5:29Geoff Nielson?

  Right.

• 5:31Speaker 01

  I mean, I think we

• 5:32Geoff Nielson?

  Maybe 20.

• 5:34Speaker 01

  I think we basically invented object-oriented programming and languages in order to not talk to each other. Like, you know, you have this thing to do and I have this thing to do. And eventually, if we, we'll, we'll give each other the right commands so that they will talk to each other. Well,

• 5:52Speaker 01

  with these AI-powered IDEs, you are collaborating with agents.

• 5:59Speaker 01

  Right? So then, um, it's like having junior developers all around you helping on one project, which is a very different style of, um, coding than I learned how to code, which was, you know, very focused, but kind of isolating, you know, Mindscape thing. But now, instead of like composing music for one violin, you're composing music for the entire orchestra, all at the same time, all at once. It's a different, it's a different flow of eh, of getting stuff done. So, you know, people have been calling it vibe coding. Uh, I don't know if you've heard that.

• 6:41Geoff Nielson?

  I've heard that.

• 6:42Speaker 01

  Yeah, I'm of mixed minds about it because

• 6:44Geoff Nielson?

  Yeah, what do I have to come up with vibe coding?

• 6:47Speaker 01

  Again, I'm old. I feel old because I'm like,

• 6:53Speaker 01

  I don't get it. So being able to get into the zone and orchestrate instead of author

• 7:03Speaker 01

  is sort of the quintessential part of vibe coding. Now, I was talking to one of our leaders here at Amazon and he was like vibe coding, but isn't it nice to know what you're doing? Like, isn't that, isn't that nice?

• 7:19Speaker 01

  Which cracked me up because I don't, I don't think he's wrong, but I don't think he's completely right either because you still have to have that artisan mindset of creating something out of the ether that, that, that's your developer mindset. See, I think you do sort of know what you're doing, but it is a, but it was, it's enough of a cognitive shift when you're in that mode that I, I think that is representing a major shift in technology that's happened in the last six months.

• 7:52Geoff Nielson?

  And I'm, I love the, I love the orchestra analogy. That, that one really resonates with me and no, no, no pun intended there with music. I'm curious, I'm curious too, like, when I think about this and I think about the way people treat it, like, is this similar to like when photography started versus like painting? Like, I feel like people, oh, it's not, it's not art. You just took a picture and it's like, okay, like we, yes, there's technology involved, but you know, there's still artistry involved too. Or is that like way out of left field or not the right metaphor?

• 8:26Speaker 01

  I don't, I'll have to think about it more if that, if that one works for me. I'll just have to think about it a little more because the metaphor that I use the most is differential equations, which by the way, this joke, yeah, you know, this joke actually plays with our core audience, right? Because I swear every engineer took Diffy Q, right? And especially old people like me, when I took Diffy Q, you weren't allowed to, you know, you had to do everything manually. And one Diffy Q equation, when you, when you were taking the test would be like four pages, right? As you're solving through differential equations. And then, and then eventually there was that day, that magic day that the professor let you have a T65 on the test. And you were like,

• 9:13Speaker 01

  What? What? Now, did the T65 kill differential equations? Nope, people still take it. And knowing how to apply differential equations, that's still there. The creativity of understanding how to apply different concepts, that never went away. What changed was that the results are way more reliable, right? I mean, I always felt sorry for the TAs because, look, I mean, I was really good at math, but I was terrible at arithmetic. And so, within those four pages, you know, one arithmetic error would just mess up your entire thing.

• 9:58Speaker 01

  And so, the TAs would have to go through and give you partial credit. I felt really sorry for them. But, once we applied a T65 to that, the cognitive load of mathematics shifted. And I think that's what we are witnessing right now, is we're basically giving developers a development calculator that accelerates and improves the quality and accuracy of the code, but it doesn't actually change computer science. It just, it changes how we are relating to the task. We're really attacking MUC, right?

• 10:41Speaker 01

  Not a hard stuff, not the creative stuff, not the risky stuff, or the sexy stuff of development. Yes, I said development sexy. But, the boring parts? Like, the syntax of a language? No one cares anymore. Like, syntax is boring, in my opinion. But I think that's shifted. And I think how we are going to relate to all technology has fundamentally shifted. And while we see the most impact within developer tools and access to technology, sort of thing, I think what we're going to see over the next two years

• 11:26Speaker 01

  is it's going to affect every persona within your software development lifecycle.

• 11:34Speaker 01

  Forevermore. And that means the things that we are already able to do with Q affecting DevOps, operations, cloud architects, and eventually DBAs, like, there's something for everyone, which means that it's going to accelerate and democratize how we relate to technology for all time and space. And I've got a 15-year-old kid, and I think about these innovations through his eyes, that this is the normal, right? This is just normal. And I'm like, I don't know what's gonna happen. So let's get to the part,

• 12:17Speaker 01

  the second part of your question, what scares the crap out of me? This scares the crap out of me. Because if you, at every inflection point where you are accelerating and democratizing, you have to build the plane as you're flying it, and you have to build the guardrails at the same time. I mean, think about, think about cars, right? What's cars started hitting the road? We didn't have speed limits. We didn't have traffic laws. We had cars. That's, that's what it feels like right now, is we can go super fast. Wouldn't it be great if we had some traffic laws?

• 13:00Geoff Nielson?

  So, just reflecting on that, and, and, you know, the, the traffic law, like, yeah, that's, it's, it's fair, right? Like, it's always, the, the regulations are always, like, you know, one to two steps behind, like, the, the technology. But I want to come back to what you were saying about, like, kind of what it means to be a developer, and how much is this is changing? And, and, you know, how this is going to change again, like, through the next two years. In, in your mind, like, what does development look like, you know, in, in 2027, 2028? Like, what does it mean to be a developer, once we kind of go through this shift?

• 13:38Speaker 01

  Um, I think having a Babbel fish for all languages.

• 13:46Speaker 01

  Well, we'll change the languages themselves. And for those of you who didn't read Douglas Adams, Babbel Fish was a universal translator you stuck in your ear. If you are able to, I mean, and most developers know like two or three languages. But the thing about those two or three languages is they're usually in the same language families. It's like, yeah, I learned Spanish and French. Whereas what we, what this is going to be able to fills facilitate is new development architectures. Um, and one of the, in order to be able to explain that concept, think about like COBOL. COBOL is a function based language. It's not object oriented at all.

• 14:31Speaker 01

  Like you ascribe the B CPU at the function level and then the messaging broker is the neural network of COBOL. So the very architecture of the language is unique compared to object oriented. So it's not about the language translation between COBOL and Python, right? It's a, it's the architecture of the language itself that that's really challenging to move between. Gerniv AI smashes abstraction layers, right? It compresses them. So the diversification that we've seen of computer languages, since there was a computer

• 15:13Speaker 01

  language, you know, from assembler on up, I really think that's going to compress because those abstraction layers, you know, are performant against different pillars or matrix. But I don't, I don't know if that's entirely necessary if you are smash, if you, if you compress the abstraction layers. The other thing that's going to change for developers.

• 15:36Speaker 01

  And this really speaks to the democratization piece is a lot of studies that we're seeing as far as the impact of generative AI on this particular population is about it getting your most junior noobs more perform it. That's where we see like the biggest impact change. Not really in your master veteran developers, but really in the new population. So that democratization piece is really fascinating to me because it's opening the aperture of who's going to be able to relate to the technology. I mean, I gave the example of the command line interface and those are all no code, right?

• 16:23Speaker 01

  So when my kid is applying for a job in 27,

• 16:28Speaker 01

  is it going to relate to language in the same, the computer language isn't the same way? Probably not. But the role of invention, context switching, synthesis, those higher order ideas that that make humans super valuable are going to be even more valuable. There's going to be a stronger delineation because you can't just show up and do a nine to five because a lot of that, that mucky stuff that is just, you know, clocking in, clocking out is going to be automated. So that means that we have to shift what we value most in each other and in our work because of that democratization piece and

• 17:11Speaker 01

  that automation piece is going to be hugely impactful across all the job families.

• 17:18Geoff Nielson?

  Right. So I'm very much of the same mind. And there's like, there's a lot of things I want to talk to you about right now. There's a lot of directions I want to take this in. But one of them is I want to come back to this notion of play that you mentioned before, because to me like this ties back in, right? Like as we think about people free from the muck of development, the play becomes more important, the innovating becomes more important, the, you know, the mindset of like, what, you know, what is the art of the possible? How could we do this better? Kind of takes center stage. And so I'm curious, you know, and maybe at Amazon specifically, or if you have any, you know, client examples.

• 17:58Geoff Nielson?

  This is sometimes a tough mindset to get like business folks or non-developers, maybe call it management into like you say like, hey, we want to play and they say that's cute. But like, we've got a quarterly number to head and you've got a release schedule. You know, get back to work. And so my question is like, is that a challenge? Is that tough to sell? And how do you get the, you know, the ecosystem of leaders who, you know, you know, God bless them, are doing their own jobs and want to see results? How do you get them to give you permission to get out of the grind and be able to think longer term and bigger picture?

• 18:39Speaker 01

  Yeah, in 2024, I think every CFO thought they could lay off 40% of their developers.

• 18:44Geoff Nielson?

  Right, right.

• 18:44Speaker 01

  They were like, we had developer productivity. Turns out, we haven't fired anybody yet.

• 18:50Geoff Nielson?

  Yeah, and turns out that they actually never, they never really understood the value of a developer, right? If they thought they could.

• 18:55Speaker 01

  They had no idea what their developers were doing. Like, no, they thought.

• 19:00Geoff Nielson?

  You have right lines of code, right? You just type on keyboards all day?

• 19:04Speaker 01

  Yeah, 40 hours a week. Turns out, it's more like four to five hours a week. Like, that's just a component of what they do. I think, I think the revolution is happening to management as well. And they're feeling it just as much. So I have two examples. One, back in, I don't know, February, January of 2023, I had asked my son to start doing his homework with chat. And then very quickly, the school was like, you can't do chat. You can't use chat for your homework. That is cheating. The schools were telling my kids this was cheating.

• 19:49Speaker 01

  So I had a parent to you conference. I'm very popular at school. I'm not.

• 19:54Speaker 01

  Or I was like.

• 19:57Speaker 01

  How many hours per week are you grading papers? Because I can show you how to do that a lot faster. Wouldn't it be cool if you were able to create a lesson plan specifically for the behavior characteristics you see in the class? Like, what if you just research and analyze the characteristics that the kids were exhibiting? And then curtail or customize your lesson plan according to what your behavior characteristics? I can show you how to do analysis much faster. And what they were hearing was they can get to happy hour sooner. And all of a sudden, their opinion about charity AI changed because it was real to them.

• 20:42Speaker 01

  Right? So in other words, it's not about control. It's about just finding the value in the right lever. The same thing, by the way, happened to me personally about the same time. Because we had these mandatory docs that are that we do in Amazon in Q1 across the entire organization.

• 21:05Speaker 01

  And they're confidential. And there was this one team that I was managing that had all of their docs done on time. And they were excellent. And I'm like, what did you do?

• 21:16Geoff Nielson?

  Wrong.

• 21:18Speaker 01

  They were like, oh, we use chat. And I was like, first of all, why aren't the rest of you doing this? Second of all, I don't think that's apseka prud.

• 21:30Speaker 01

  And third, holy moly, we just had a huge cognitive shift in management.

• 21:40Speaker 01

  Even before I we fully felt the effects for the developer audience. So I don't think it's about our management giving permission. Because when you when you talk to a CFO and you give them a natural language interface to do ad hoc queries and analysis using, you know, things like Q for Quicksite, that that that accelerates their monthly business reporting review and gets them to happy hour sooner. So there is, you know, there, it's, it is literally affecting every job family that we touch. And so

• 22:22Speaker 01

  It's only the like really old stodgy people that are very like, get off my lawn. You can't use generative AI that is going to be left behind. It's not, this is just the new normal across everything now.

• 22:37Geoff Nielson?

  Yeah. So, old stodgy people, we had a, when we last spoke.

• 22:45Speaker 01

  So let's talk about the old people.

• 22:48Geoff Nielson?

  Let's talk about the old people, Rory, because when we, when we spoke last, you were saying, and you did say you didn't want to be ageist, but after you said that, you said that you found the biggest indicator of willingness to adopt generative AI tools was age and that younger people tended to be much more willing to experiment. Older, stodgier people tended to, you know, be more set in their ways. I'm already a master coder. I mean, two questions for you. Is that, is that still holding true in the time since we've had that conversation? And if it is like, what, what are the implications for that? Like, do we risk having kind of like a bifurcated workforce where you've got half of people

• 23:33Geoff Nielson?

  or, you know, some fraction doing it the new way, some fraction not coming along for the ride? And what do you do about it? You just have to wait for them to retire or is there some way to like take a happy path?

• 23:45Speaker 01

  I, one, it's not just ageist. I mean, because, I don't know, being stuck in your ways is not necessarily just age. So I, I was hyperbolic in my group.

• 24:00Speaker 01

  Me, hyperbolic, who knew?

• 24:01Geoff Nielson?

  That can't be right. That doesn't sound like.

• 24:06Speaker 01

  But there is definitely a divide and there, there's definitely a lot of fud.

• 24:12Speaker 01

  With certain populations, they're risk averse or, you know, they don't want to try a new way of doing things. But I don't, I don't think the divided will last because humans are wonderfully lazy. We're just lazy, right? We're, we don't do work for the sake of work. And when you have part of your workforce that's.

• 24:34Speaker 01

  Doing something faster and is more efficient. And the other one's like, no, I want to stay late and do this manually. No, that doesn't happen, right?

• 24:46Speaker 01

  I mean, and again, I'm gonna date myself here, but remember when we were all switching to email?

• 24:54Speaker 01

  You know, like, I know you don't, because you're a sweet young. I know. But, okay. In the 90s,

• 25:01Speaker 01

  just culturally, humans were shifting to email as the primary communication method for business. Instead of like faxes or typewriters or whatever, you know, it became email. And

• 25:18Speaker 01

  the question was like, does this destroy writing? You know, when you, when you have, we are switching to email. And there was a lot of, you know, senior executives that would want all of their emails printed out. I think Warren Buffett still does, but

• 25:37Speaker 01

  yeah, those, those, that, that didn't last. It was inefficient. Yeah. Right? And it just doesn't matter, like, humans adopt new technology because we're just wonderfully lazy people, you know? I don't, I don't see that, that, that is, mind you, the, this conversation has shifted pretty far away from technology. And now we're into, you know, armchair sociology, which I'm not qualified to answer. But I can see how humans are interacting with technology and different trends. But I don't see anyone being left behind in this. Mainly because it's so easy, darn easy to use, right? It's not like you have to learn a new programming language. It's, it's all through natural language.

• 26:23Speaker 01

  And that, that means that the adoption has been transformative over the past 18 months. Because it's just so darn easy. I was gonna say one other thing, and this is probably the most controversial thing that I can possibly say. I really hope I don't get fired, but

• 26:42Speaker 01

  Bezos is very famous for saying, there's no compression algorithm for experience. So you remember when I was talking about the, the younger developers becoming more performant?

• 26:54Speaker 01

  What if there is? What if there is a compression algorithm for experience? And what if, what if this is it?

• 27:02Speaker 01

  Because we, we, we now have the ability to do pattern recognition on a really, really large scale, which prevents mistakes. And that's, that's what your most experienced developers do.

• 27:16Geoff Nielson?

  Right.

• 27:17Speaker 01

  So maybe we have it. Again, Bezos don't fire me.

• 27:22Geoff Nielson?

  So that's a really interesting point, Rory, and I don't know. I mean, maybe it's, maybe it's controversial and I'll knock on wood that, you know, you're still employed there next week, but.

• 27:31Speaker 01

  Thanks.

• 27:33Geoff Nielson?

  Well, when I talk to people on here, one of the things I hear more frequently is that it's actually junior staff that are most at risk of being disrupted by AI in the sense of like, oh, well, we don't need junior devs anymore because AI is just as good a coder as a junior dev is. And, you know, your approach is different. The, what I hear you saying is actually they have the most to gain because you can now get from a junior to, you know, being seasoned faster than ever before. So just to, to put a fine point on it, like if you're, you know, in your son's cohorts and you're going to be going, I mean, I'm not, I'm not suggesting he's going into CS, but if you're going into cop side, um, you know, should you be concerned?

• 28:18Geoff Nielson?

  Or is this like a more exciting time than ever?

• 28:22Speaker 01

  Okay. Whether or not I'm concerned or, or not as a mother, doesn't matter. No, I'm not really, I mean, because this is happening, y'all. Like you can be fearful about it or you could be happy about it, but it doesn't really matter because it's still going to happen. Like this is the new normal. I think

• 28:43Speaker 01

  what I try to emphasize with my kids is like, is grit and creativity. Because I don't think either of those things are going away as necessary for like being happy and successful for the rest of your life. So grit, creativity, those are, those are still very important. The most human things possible. As far as like being concerned, what I, what keeps me up at night or what, what I'm most concerned about is, um,

• 29:16Speaker 01

  It is bad actors doing really naughty things with this technology.

• 29:22Speaker 01

  And

• 29:25Speaker 01

  you know, white cats not being able to create and remediate fast enough. Because everything, this is like napalm. That's probably a bad word. This is like, this is a massive accelerant in both scenarios. So our ability to detect anomalies across a large data set, like say,

• 29:53Speaker 01

  I don't know, logs, massively improved, right? Because L L Ns are fantastic of taking massive amounts of data and then analyzing it super quickly. So our, that, that white cats push has been greatly accelerated. And so is the white cat activities that you could do. It's a massive accelerant. So the thing that keeps me up at night is just, I want to enable and infect people's imagination for doing good things because only then we'll have guardrails that actually protect people. I, I talk about my son a lot in the context of generated AI because well, it's fun to experiment on your kids when it comes to technology.

• 30:39Speaker 01

  And the metaphor that I'm going to give you is really apropos to generated AI. Um, the other day he came into the kitchen and he smelled like gasoline. And I'm like, why do you smell like gasoline? He's like, I made rocket fuel. And I was like, okay. Yay.

• 30:59Speaker 01

  Did you light it? And now he realizes there's something up and he's like, no. Of course he lit it. I did not have a no rocket fuel rule in my house, right? That like, I didn't even know I needed one. Well, that was not in the manual. Okay. So we had to come up with a new rule in order to accommodate

• 31:23Speaker 01

  like future experimentation, which is if you have a spark, you need an adult within 10 feet and a cellphone.

• 31:30Geoff Nielson?

  Right.

• 31:31Speaker 01

  Okay, like there needs to be a backup plan. I'm not saying no spark. I'm just saying backup plan. So it was about a guardrail than a discrete rule. And as we think about Jared debate AI, it's that same sort of mentality. You can't have specific rules like

• 31:48Speaker 01

  instead you've got to have guardrails for good sense. More so than specific traffic laws. Right? We're not traffic law stage at all. We're at, hey, don't run over people. That's a good rule. I mean, that's a good rule. But specifically what you need to do in order to ensure that you're not running over people, those discrete rules. I don't, I don't think we're going to get to anytime soon. Mainly because of the wildly creative aspects of what you're able to do with the technology.

• 32:26Geoff Nielson?

  Right? So have you seen? You know, I'm curious and we're getting dangerously close to mixing metaphors here. But yeah, what have you seen rocket fuel situations with black hat actors in AI that you can share either firsthand or secondhand? But you're like, holy moly, like that was creative. Like, I didn't know you could do that. This is starting to open up like big concerns.

• 32:51Speaker 01

  To be honest, and this is just because I think I'm in a I'm in my own little bubble. I haven't seen that a lot. I think one of the biggest concerns I hear from CEOs and CTOs is the injection of malicious code. You know, because somebody said, oh, I thought this coding suggestion was great, so I cut it in.

• 33:15Speaker 01

  And they're terrified of it, but I haven't seen it to, I haven't seen the, uh, that happen.

• 33:23Geoff Nielson?

  Right.

• 33:24Speaker 01

  Really? Because here's the deal. If a developer wants to put malicious code into anything, it's just not that hard. I mean, you could get malicious code on Stack Overflow, cut, paste, you're done. Jared of AI is just an accelerant. Either way, and if I had a developer that was, you know, putting in malicious code, do I blame Jared of AI? Like I would a human? No, I fired the developer, right? I mean, like, it's still, this is not a,

• 33:56Speaker 01

  we use the word agent a lot and it annoys the crowd out of me because no one seems to have a consistent definition. But the thing about having agency is something we usually ascribe to humans. Um, because they are responsible and they have choice, but we're not at that stage yet with this technology at all. And if you, if you have a human that is using this tool to do naughty things, you still, you still blame the human. Okay. We don't blame the technology. I would still fire the developer for introducing that code. It's just this, there, there's a shift, right? From, like I said, authorship to orchestration doesn't,

• 34:37Speaker 01

  that doesn't mean that you have given up responsibility and ownership to technology, that we're just, we're just not there. Uh, that would be silly. And the CFO example that we were talking about, if the CFO reported bad numbers because he used Jared of AI, we would fire the CFO. We wouldn't fire the tool.

• 35:03Geoff Nielson?

  No, that's a great, uh, that's a great case in point to bring it home. Um, the CFO case is still, is still very interesting to me and I, I, you know, I hear about it and I imagine it that there's still CFOs out there that are saying like, but when, when can we fire 40% of developers, right? Like the tools are here. When, when can we fire them? What? What's the best antidote to that? Like, what's the best response to help CIOs better understand, or, sorry, rather CFOs better understand what developers actually do and where they're real?

• 35:35Speaker 01

  I ask them if they use managed databases. Like, a managed database, like RDS, 70% of what a DBA does is super boring. I know, because I was a DBA. Super boring, like patches and backups and failovers. This is not information architecture. This is majorly boring stuff. Managed databases like RDS do those things, 70%, right?

• 36:02Speaker 01

  And so my question to CFO is, like, when you switch to your managed database, because everybody has managed databases somewhere, did you fire 70% of your DBAs?

• 36:12Speaker 01

  And they were like, "No." So you still have the same number of DBAs

• 36:18Speaker 01

  as you did in the past? And they're like, "Yeah." And I said, "What are they doing?"

• 36:24Speaker 01

  They don't know. And I tell them what they're doing. Because what happened with managing databases was a cognitive shift, that these people that went to school and spent their lives in information architecture are free to do information architecture. In other words, and so what we saw was a democratization of data science on a massive scale and an accelerant for analytics. Data's getting cleaner. Like, information's more accessible. Insights are driving the company forward, not just, you know, a cost center that is like the tax that you pay for information.

• 37:07Speaker 01

  There was a shift in responsibilities once we gave graphing calculators to DBAs. We're going to see that same shift within all of our job families, I think, over the next two years. And what they actually do is probably going to change as well. Meaning, you're going to not see this area as, like, just a cost center that you have to maintain old code. What if we don't maintain old code anymore? 70% of a developer's time is spent maintaining old code. If we were able to retire all of that tech debt and get 70% time back...

• 37:52Speaker 01

  What's gonna happen? I don't think it like we're gonna be static and we'll just fire 70% of the developers that because what we've seen in various technology innovations is that just doesn't happen. Just humans change, they evolve and they do something different. Not stay static with the same demands on the business.

• 38:16Speaker 01

  There's been no evidence of it. People evolve.

• 38:24Geoff Nielson?

  So I'm thinking, I'm thinking about the technology itself, Rory. And, you know, it's like, we talked about this already and it's, you know, a truism that like the, like the, the pace of this is just like completely out of control. Right? Like as, as you said, it's like week to week here. Um, as that's happening, are you worried that like? The adoption curve is falling farther and farther behind that? And by the way, I mean, I'll, I'll share, you know, my story a little bit, which is when I talk to business leaders around here. They tend to be like, oh yeah, like AI seems to be this, you know, hot new thing. And in the past year, it's gotten really good, like, you know, and probably a year or two from now, it's going to do amazing things.

• 39:08Geoff Nielson?

  And then when I talk to the CIOs and CTOs around here, they're like a year to two years. Are you kidding? Like this thing, it's, it's weekly, like, it's like three months from now, we're going to be in a completely different world.

• 39:20Speaker 01

  Right.

• 39:20Geoff Nielson?

  And we're like struggling to stay on top of that. So, I mean, first of all, are you seeing that as well? And second of all, like, what do we do about that as, you know, organizational leaders, but also as, you know, for, you know, technology organizations like Amazon?

• 39:37Speaker 01

  Um, I recently saw a, an analyst report and I wish I could remember this guy's name and I, so I could give him proper credit, but my brain is fried.

• 39:52Speaker 01

  Where it did an analysis by job family of the impact of, uh, uh, of, um, at surveying people in different job families about the impact of generative AI. And you're dead on right, like the ones that thought that the, that generative AI was going to change their life the most were technologists.

• 40:14Speaker 01

  And it got me thinking about that a lot. And I think the reason why it's disproportionate by job family is that we saw the most stark shift in the effectiveness of these tools with technologists was because I think most of us started with code generation and first. And there's a really good reason why we started with code generation. It's because code is really consistent, right? There's no adjectives and it's highly structured language compared to

• 40:54Speaker 01

  English, right? So applying large language models onto a corpus of code that is already very, very, very structured is having exponential results. So I think it is easier for it to be real for the technologists and for them to see the impact because when you look at, you know, the earliest tools in 2023, like GitHub Copilot and for us, it was CodeWhisperer, that's where it started. And I still think that workload is probably the absolutely most mature of tools that have incorporated generative AI natively

• 41:40Speaker 01

  than anything else because it was the first. I think that's going to infect and go viral across the other job families. They just, they haven't done it yet. I mean, I was talking to

• 41:59Speaker 01

  a title company. I mean, title companies basically just process contracts, right?

• 42:07Speaker 01

  And what they wanted to apply generative AI to, to was we want it to write contracts. And I was like, but that's, that's the family jewels. That's literally what you guys do. Why would you want to apply to the most, the highest cognitive load and the most valuable humanistic thing that your company does? Instead, why don't you apply it, you know, to, to, to archiving? Like being able to read PDFs and search and index your treasure trove of data that you have in your previous contracts in order to, you know, learn from different patterns that you had there. You see, that's pretty low risk, right?

• 42:51Speaker 01

  Because that hasn't been done before and you'd be able to analyze massive amounts of contracts, not create net new contracts. So I think where a lot of people get kind of the wrong end of the stick is they want to go to the like, the hardest, riskiest, sexiest thing in their life and try to replace them. But where we're seeing the most success is when generative AI is applied to the muckiest, most boring parts of stuff you don't want to do anyway, or that's repetitive where, you know, foundational mistakes are made. Like, and for whatever reason, like the other job families, like outside of technologists

• 43:34Speaker 01

  are, are, are jumping to sort of the wrong conclusion of what they want to apply generative AI to first. And so it doesn't, they're not having the same effect. So if there's any advice I would, I would have for any job family outside of my technologist crazy brain is don't get generative AI to make art.

• 43:55Speaker 01

  That it's not really art them. Instead, you know, apply generative AI to all the stuff that you don't want to do so that you have more time to make art. That, that's where we need to shift our understanding. We're not trying to replace humans. We're trying to accelerate and democratize. And that's a shift in mindset that we got to get other job families to understand outside of just the technologists.

• 44:22Geoff Nielson?

  Right. And that was, that, that was like a real light bulb moment for me last time we talked too, is that like, you know, people complain about like, oh, like AI is going to do all the creative stuff. It's like, well, then don't use it for the creative stuff. Use it for the stuff you don't want to do. And I was like, yeah, it is good at that. That's a great, that's a great mindset. Yeah.

• 44:44Speaker 01

  They like, people ask me sociology questions all the time. It's like, what are you worried about? Or what are you concerned about? Well, I just don't think like that. And instead, you, the glass isn't half empty or half full. The glass has four ounces. Okay. What you choose to do with the, for those four ounces is kind of up to you. But worrying about the potential of what's going to happen with those four ounces isn't particularly useful. But going, hmm, four ounces. Well, it would be better if we had five. Uh, you know, like, use it as a tool. That's what the, that's what technology is. These are all just tools. And they're fairly neutral until

• 45:23Speaker 01

  humans get involved. Humans are the, the, the defining factor. And what, if I was a therapist or into the social sciences, I would, I wouldn't be working for Amazon or in doing what I'm doing. So I'm probably not the best person to be qualified to have those discussions. But I'm more in the, it's four ounces, and this is what it can do, and this is what it can't do, kind of mindset.

• 45:47Geoff Nielson?

  Right. Right. Spoken like a true engineer.

• 45:51Speaker 01

  I'm actually not an engineer. I think I was just born an engineer.

• 45:57Geoff Nielson?

  Really?

• 45:58Speaker 01

  Yeah.

• 45:59Geoff Nielson?

  I didn't know that about you. I thought, uh, but you had a math background or how to, no?

• 46:05Speaker 01

  Um, okay. Again, uh, it's gonna make me sound super old. I grew up in a really small town in Paris, Texas. And in Paris, Texas, you know, it was kind of late to the technology game. So by the time I went to college, uh, I went to the University of Texas at Austin, and you, you, you got the course descriptions, and it was a massive book. And for me, I was like a kid in Candyland. I would like every, I was like, ooh, anthropology, I'm gonna study anthropology, or I'm gonna study advanced chemistry. Because I love to learn. I was just that kid. That I always just, I had an insatiable appetite for learning. Um, so I think I graduated with literally twice as many hours as I needed for my degree.

• 46:48Speaker 01

  Because I just kept taking

• 46:51Geoff Nielson?

  Just follows it, right?

• 46:52Speaker 01

  Yeah, I just love learn. I don't, I, in my household, even though they were rednecks, I mean,

• 47:00Speaker 01

  my household just had an attitude that nothing was unknowable. And maybe it was because we, we just grew up so remote that we have a very can do attitude to literally everything. It's like, hey, the car broke down. Well, you should go figure out how, you should go figure out how to fix the car.

• 47:16Geoff Nielson?

  Yeah.

• 47:16Speaker 01

  Or, you know, this is a computer. You should go figure out how that works. It's just like it in the, I have a mental model that nothing is unknowable. Um, which means, and I don't, uh, I have a personal beef. When I'm interviewing candidates, if I ever get a candidate says, well, I'm really not technical. I'm like, that's like kiss of death for, for interviewing with me because I'm like, that's just, you, you, you're not born technical. You're not

• 47:45Geoff Nielson?

  Yeah, well, it's a lazy answer, right?

• 47:47Speaker 01

  It's not measurable either.

• 47:51Speaker 01

  It's like saying, well, I'm not literate.

• 47:54Geoff Nielson?

  Yeah.

• 47:55Speaker 01

  What?

• 47:58Speaker 01

  That's not a thing. There's just stuff you know, and there's stuff you don't know yet. That's it. Those are the only two categories we're allowed in life.

• 48:06Geoff Nielson?

  Yeah.

• 48:07Speaker 01

  There is no, and I have this similar, and by the way, this rant does relate to generative AI because when people say, uh, like, oh, I don't know, that's, that, I'm not really, you know, techie or I'm not really, like, that's not a thing.

• 48:24Speaker 01

  There's just stuff we know how to do now, and there's stuff we're going to know how to do tomorrow. There's not a person that is one thing or another. You, this is just all, there are some things you're going to learn faster than others. Fine.

• 48:41Speaker 01

  But that's where the grit comes in. Like, if you need to know this thing, whether or not your brain works that way or not, you, you need to fucking learn it. Like, if you have an alacrity for languages, great. If you don't, and you've moved to another country, too bad, you're gonna have work choices hard, but you're still gonna learn it.

• 48:58Geoff Nielson?

  Yeah.

• 48:59Speaker 01

  That's just, that's just human.

• 49:02Geoff Nielson?

  Yeah. Well, and you know, with generative AI, it just seems like it's like it's less of an excuse than ever, right? Because it evens takes less grit to answer a lot of these questions.

• 49:12Speaker 01

  Yeah, that like that drives me absolutely crazy because it's more about how fast you can learn, not if it's it's when and not if.

• 49:25Geoff Nielson?

  Yeah. Well, that's just that's just really resonates with me. And, you know, just a quick aside, you know, from my experience. So when I was in college, I actually used to be a guitar teacher. That was kind of like one of the ways I paid my way through college. Great gig, by the way, really good gig. But it was funny because you, you know, you have all these students and some people, and I'm sure you've encountered this, some people, I'm not musical. I'm not musical. I don't have a for it. I'm not musical. And some people like didn't have that. And what it came down to when you when you had students week after week, is the ones who got good were the ones who practiced, right? Like, yeah, like, forget what you like, you know, what mental model you have of yourself and what you can and can't do. Like, it's grit, right? Like, put in that

• 50:06Geoff Nielson?

  time, and you will get good. And do some people maybe get good 10% faster, 10% slower? Maybe. But if you want to be good, keep practicing and show up. Exactly.

• 50:18Speaker 01

  I you had no idea how much that resonates with me, because I was told my entire life I was tone deaf. Like, that was a physical deficit. By the way, there is no such thing as tone deaf.

• 50:28Geoff Nielson?

  No.

• 50:29Speaker 01

  I didn't. But you know, you're told your entire life, you're tone deaf, because you're like, hit this note, and you're like, and it doesn't it. And so, for whatever reason, I got a burr in my saddle that I wanted to learn violin when I was in college. Okay, I don't know, like, violins has no frets. So you are constantly tuning your violin, right? Which means you can't be tone deaf, right? In order to play the violin. So I sweet talked my way into a college civil violin course. I had never played the violin before.

• 51:03Geoff Nielson?

  Good for you.

• 51:04Speaker 01

  No, I'm just obstinate. Okay. So I came to this class, and this professor is like,

• 51:13Speaker 01

  you're tone deaf. And I'm like, no, not. And so what he did was is like, this is a pitch pipe.

• 51:22Speaker 01

  You've got to figure out how to do this. And I just kept trying until I finally got it. And then I got, I got the, I got the T65 of the violin, which by the way, I didn't know was a thing. And it's an electronic tuner. You, there's a thing with a graph. And then you just match it. And then you learn, and then you, and then you teach yourself how to hear different tones because you have a graphical interface for if you're sharp or flat. And I was like,

• 51:51Speaker 01

  wait, so everything I was told growing up about that you're a math person, or you're not a math person, or your tone depth, or your this is other BS?

• 52:03Speaker 01

  It is. It's total BS. We are wads of elastic clay that can literally do anything. And the idea that you were limited in your brain is absolute BS. We were taught that as kids, but none, it turns out it's not true at all.

• 52:20Geoff Nielson?

  Hey, I wanted to say a big thank you for coming on today. This has been a really, really interesting conversation. We like, I was gonna say we covered a lot of ground, but even saying that doesn't cover the fact that like, we worked in Pride and Prejudice and like, I'd give it to a conversation that was supposed to be about, you know, Amazon Q and generative AI. But I had a really, a really interesting time, lots of insights. And yeah, I really appreciated it.

• 52:43Speaker 01

  Awesome. Thank you so much for having me. I always enjoy these conversations.