Pod Insights a16z Podcast AI Copilots and the Future of Knowledge Work with Microsoft's Kevin Scott About this episode Microsoft CTO Kevin Scott, in conversation with a16z's Bob Swan, explains how AI copilots are keeping developers longer in a flow state and why AI copilots more broadly could be the start of an industrial revolution for knowledge work.
[0:00] Microsoft's approach to AI
[5:55] The next Industrial Revolution
[10:27] Developer productivity & flow state
[15:26] Reprogramming the American Dream
[19:48] Advice to builders
This conversation is part of our AI Revolution series, recorded August 2023 at a live event in San Francisco. The series features some of the most impactful builders in the field of AI discussing and debating where we are, where we’re going, and the big open questions in AI. Find more content from our AI Revolution series on www.a16z.com/AIRevolution.
Transcript 0:00 Speaker 0 You know, when you're in flow state, you know it, like you're just killing it. Uh, and so how can you use tools to preserve that um as long as humanly possible?
0:40 Speaker 1 Look, Microsoft has played such a huge role in determining the future of AI. So maybe if you could just take a few minutes and talk about how you see the vision and the future of AI and what is the role of a large tech leader like Microsoft uh in this in this wave of AI.
1:00 Speaker 0 We really do see AI not as a product and not as you know, a research project, but as a real platform for lots of people to build many interesting new things on top of, particularly the past couple of years and maybe the past uh 12 months in particular with the um launch of Chat GPT and GPD4, you can sort of really see, I think the potential of the platform maybe in the same light as the personal computer or the smartphone, uh just in terms of uh a bundle of technologies that are going to make a whole bunch of new things possible that lots of people are going to be building things on top of, uh, most of which we probably don't even really understand right now.
1:50 Speaker 0 Um, and so like I think that's the exciting thing to me about building uh building a technology. Like I've always been a tool builder, and like the idea of building a tool and then having a bunch of people use it in ways that you didn't anticipate is like really uh super interesting.
2:11 Speaker 1 And in many ways, you're not doing this on your own. You've partnered with leaders in the industry, whether it's GitHub or OpenAI, as you mentioned, and most recently the meta uh Llama 2 announcement. Can you talk a little bit more about the role you see partners playing in your vision and how how you expect the ecosystem to expand over time? And maybe a question that you've I'm sure you've gotten a million times is where do you get all the GPUs to support going and expanding relationships?
2:45 Speaker 0 Yeah, I uh the GPU question is uh I don't know, it's it's probably not comical, but I have to sort of laugh uh just the frequency of that being the number one question right now. Um like it's not even uh you know weekly or monthly thing, it's uh you know, hours in the day that uh go by where someone's not asking for uh for GPUs is uh almost non-existent. You know, again, going back to this whole point of if you're building a platform, like the platform has to get built with partners and it has to get used by partners. And so uh like our partners are extraordinarily important uh to us. Um yeah, I think if you think about AI, like by far and away, the most important partner we have is uh is open AI.
3:36 Speaker 0 Like that partnership has been going for four years now, um, and started as a hey, we are a hyperscale cloud, like we're gonna build a bunch of AI supercomputing infrastructure uh in collaboration with you that is then going to get used to train some of the most advanced AI models and systems in the world, and then we're going to do some work together to figure out how to take those models, uh, those platform building blocks and get them deployed into products that both Microsoft offers like GitHub Copilot, as well as you know, how do you um you know deploy these things into environments like Azure and things like the Azure Open AI API, where uh people can just build uh their own software on top of it.
4:28 Speaker 0 Um, and then you know, these other partners are super important as well. Like GitHub, obviously is uh I think one of the most interesting places to go deploy AI uh right now. And it is uh like the the first illustration of this co-pilot pattern that we are trying to build out, which is how can you take a uh uh a bit of knowledge work or cognitive work that someone is doing um and use AI to help them be like dramatically more productive at doing that cognitive work? And if you look at developers, like we just don't have enough of them to write all of the software that the world needs, and a lot of the lot of the work that developers do is just toil, uh, and and so having a tool that can help both make the job of software development more delightful and to like help uh software developers just be more productive, and then you know, like back to your um you mentioned Meta and uh Llama 2, and you know, I I think Llama 2 is uh just like Lama uh going to be an important building block that people are going to want to use to build AI software.
5:44 Speaker 0 Uh, we want AI software running on uh on Azure, and so like we're we're doing everything that we can to like make that an easy thing for people to do.
5:55 Speaker 1 You know, the concept of of co-pilot and how you create these AI tools to assist people with the complex cognitive tasks.
6:05 Speaker 0 You
6:05 Speaker 1 know, how how do you think AI can and will change how we all work over time and maybe what's your advice for how we get the most out of tools like Copilot?
6:17 Speaker 0 If if you think about the industrial revolution, what happened is for the first time you had uh engineered devices, machines that were able to assist people with uh labor, and like most of the labor and the you know 19th and 18th century and like through most of the you know 20th century was physical, and uh the the invention of these new technologies all of a sudden allowed the transformation of uh the physical labor that almost everyone in the world was doing. Um, I I think a similar thing is going to happen now with uh with cognitive work uh and so knowledge work, so that which is becoming a bigger and bigger part of uh what the workforce is spending their time and energy doing.
7:10 Speaker 0 Um, I mean, like you and I are both knowledge workers, and so we we haven't really had a massive breakthrough in the productivity for knowledge workers. Maybe you
7:23 Speaker 1 know,
7:23 Speaker 0 maybe the PC was uh like a an accelerator, maybe the invention of the uh internet was an accelerator. I I don't think the smartphone helped much at all, just in terms of uh productivity and knowledge work, but I think this might be the biggest thing that's ever happened for this particular type of work. Um, and I think the way that it will get used in early days, and you can sort of see this right now is kind of twofold. Um, one is like you you just sort of have these big categories of knowledge work where you there is no way to do more of the particular type of work uh because productivity is basically not getting any better. Um, but where you know you've got some kind of deficit in society because uh like you actually could use more of it.
8:15 Speaker 0 Um and then you know, I think on the on the other hand, like you you just sort of have like tons of drudgery uh in a lot of the work that we all do, where if you had um some kind of productivity mechanism that could come in and do the you know your least favorite part of your job or the like the most repetitive, most redundant uh you know, sort of work that you uh you do. Like I think we all would be delighted to have those sorts of tools in our lives. And so I I think lots of companies are gonna get created uh very soon to uh you know really help uh tackle some of those problems of uh, you know, hey, where's the you know, where's the you know, sort of crap work uh that people would be desperate uh and like happy to have automated because it would make you know their experience of their work better and would let them get more done.
9:14 Speaker 0 Um and then but like the interesting thing though is if you think about some of these big platform shifts that we've had in the past, um the most valuable things that get done on the platforms are not the things that got deployed in the first year or two of the platform change. Um yeah, like if you think about the smartphone, the place where you spend most of your time on your smartphone is not the SMS app, it's not the web browser, it's not the male client. Uh, it's the new things that got created on top of the platform and you know, the years following the event. Availability of the platform, it's sort of the hard stuff that you you really have to sort of think about like what it what are the difficult things that have now become possible that were impossible before.
10:09 Speaker 0 Um and like I think that's the thing that people ought to be thinking about. Like what are the hard problems I can be working on? There's a lot of easy things that you can go do right now that are gonna be useful, but I don't think those are the most valuable things.
10:28 Speaker 1 One of the, you know, one of the use cases that has been pretty powerful, um, has really been code assistant within GitHub Copilot, kind of going inside the four walls of Microsoft. Can you talk a little bit about the tools that you're rolling out to your own engineers and how is it changing how engineers get their work done at Microsoft and any thoughts or ideas on how you track and measure that kind of productivity in the engineering community?
10:57 Speaker 0 I'm I'm not sure whether uh any of the AI systems uh are creating the need for new measures for developer productivity. Um but I I think it's it's really really important. Uh, like if you are running a software development organization and you don't have good uh good metrics and visibility into the productivity of developers, like you need to get on it right now because uh it will be very hard to decide which of a bewildering variety of AI, you know, developer tools are coming your way that you want to adopt uh and like what order you want to adopt them in if you don't have those measurements in place. The thing that I will say is um developer productivity is not lines of code produced.
11:45 Speaker 0 Um it is uh how it is that you're measuring a developer's ability to very quickly deliver things to users and then measure whether or not those users are benefiting from the things that the developer is producing, and it's sort of you know instrumenting that entire feedback loop and making sure that you are able to identify points of friction, like throughout the entire uh product life cycle cycle development is just super important. So, I mean, the the thing that we're doing inside of the company right now, uh, you know, with AI tools, uh like first and foremost is like getting everybody to use GitHub uh GitHub Copilot, which is like a real big productivity win.
12:32 Speaker 0 The the thing that it really does in our observation with developers more than anything else is it helps keep them in flow state um longer than they otherwise would. So rather than you know, hitting a blocker when you're writing uh writing a chunk of code or you're trying to accomplish something and be like, okay, well, I don't know how to go get the next thing done. I've got to go consult documentation, I gotta go ask uh another engineer who might be busy with something, like being able to get yourself unblocked in the moment before you're out of flow state, uh is like just extraordinarily valuable. And I and I think that's a thing for folks who are thinking about the utility of these generative AI tools that you're building for things other than software development, it's a useful thing to look at.
13:29 Speaker 0 Like I I last fall before uh you know GPT4 was announced. I decided I was like, all right, well, could I use can I use these tools to help me write a science fiction book, which is something I've wanted to do since I was a teenager? It it was terrible at like writing all of the book, like it just wasn't good at like you know, giving you like chapters of well-formed pros with character development and like all of the things that you would want in a good book, but it was really terrific at keeping me in flow state. So, like if I measure my productivity of like how many words I could write in a day versus uh with the tool, like 2x easy. Um and so I think that is uh, you know, a really interesting thing that everyone should be thinking about in general is like this notion of of flow state, or like, you know, when what are the what are the conditions in which people are their absolute most productive, like they're most delighted at their job, like you know, they just sort of, you know, when you're in flow state, you know it, like you're just killing it.
14:44 Speaker 0 Uh, and so how can you use tools to preserve that um as long as humanly possible? Um, you know, the other place too where we're using a bunch of uh a bunch of AI tools is actually in the deployment of AI. So like a lot of the testing that we're doing right now, a lot of the responsible AI work that we do, uh, like are all using uh the AI tools them themselves to help do that. Uh so that that's another interesting thing that just wasn't obvious. Uh it hadn't been obvious to me that it was going to be as important as it is uh when we started doing deployments at scale about uh nine months ago.
15:26 Speaker 1 Can you talk a little bit about how you as a leader are effectively rolling out the tools to maybe to more motivate and inspire rather than stoke the fears that our employees have about what this technology means for our roles and and then on a related basis, how you know, how do you handle the resistance to change organizationally for uh disruptive tools like this and and your kind of daily job.
15:54 Speaker 0 Yeah. I mean, I think I think those are two very different questions, and I'll uh like I'll I'll sort of dig into them uh separately. So I I actually wrote a book about this, uh, like honestly, at the worst possible time. So uh books called Reprogramming the American Dream, uh, and it was published, uh it went to the bookstores uh and hit shelves uh on in March of 2020, uh the week that the shutdowns uh for the COVID pandemic uh started. Um, but but the whole point of the book was trying to assert that AI creates more opportunities for people by far than it creates um the potential for harm. And and it's not to say that the potential for harm is zero, uh, or that you can ignore it, or that uh you know, you can make that the you know the the umpteenth thing that you're thinking about uh with an AI product in uh you know among a very long list of things that you're thinking about.
17:00 Speaker 0 But it it does mean I think that we have to be really insistent and determined uh around these optimistic scenarios. Um and and like part of part of the book was rooted in my experience. I'm I grew up in rural central Virginia uh in a place where you know the the economy was powered by tobacco farming, uh furniture manufacturing and textiles, uh in all three of those industries by the time I was graduating from high school had just collapsed. And when folks in these communities have access to very powerful tools, like they tend to do remarkable things that create economic opportunities for themselves and for their families and their communities, they solve problems that you or I are not gonna solve just because we don't see the whole problem landscape of the world, like we we don't have their point of view.
18:02 Speaker 0 Um and I think these tools of AI are becoming more accessible now than they have been by a dramatic wide margin. Uh you can do interesting things with these tools right now to solve problems like create businesses, like you know, be an entrepreneur in small town Virginia, um, without having a PhD in computer science or expertise and classical AI. Uh, you just need to like be curious and and entrepreneurial. You know, your second question was about you know, how do you get people resistant to change to embrace some of the what what's happening right now with AI? What we have seen at Microsoft over the past nine months, and like I think you all are seeing this too in the entrepreneurial community is like as soon as you have large groups of people enthusiastically building things on top of this technology, it's very easy to like get on the I hesitate to call it uh you know, the bandwagon, but to like really sort of see like all of your peers uh adapting to the change and getting excited about it and building interesting things, and then just say to yourself, like I've got to go do this as well.
19:25 Speaker 0 And so I I think part of the problem in any organization is figuring out how to get yourself to that point of critical mass of adoption where then it just gets easy, and then you have the opposite problem, which is the thing you started the call with, which is people screaming at you that they don't have enough GPUs.
19:50 Speaker 1 So let me just kind of close with one last one last more general question. What do you see as the biggest open questions in AI? At the moment.
20:01 Speaker 0 We have proof points of you know a sort of a 20 watt organ that uh you know sits between our ears, uh, that that is uh artificial general intelligence, uh, that is many, many, many orders of magnitude more efficient than the digital uh artificial neural networks that we're building right now. Um, and so you know, I think they're you know, from a technology perspective, there's probably a whole handful of breakthroughs left to be uh discovered in figuring out how to close that efficiency gap. The unlock that you get when you can do something on your laptop versus uh, you know, things that where you have to go buy millions of dollars worth of uh worth of computing infrastructure to do something is really enormous.
20:51 Speaker 0 Like it, you know, does that thing for the product development that I was talking about earlier, which is like helps you turn the crank faster and like makes the iterations less expensive. Um I think just getting big efficiency breakthroughs, like be great for us, obviously, because we're spending you know huge amounts of capital on infrastructure to run uh run these models. Uh, but I think it'd be great for the whole world and the state of rate pace of innovation that we're seeing in uh AI products. You know, and like the thing I'll say to all of you know the entrepreneurs in attendance is just go find hard stuff to work on. Don't chase the trivial things.
21:37 Speaker 0 It's like don't don't be the AI equal moral equivalent of the fart app, uh, you know, fart apps that were flooding the app store. Uh, like just because a thing is has become possible doesn't mean that it's useful. Um, and and so just really, really focusing on the fundamentals of product building, I think is important. Like AI is like model's not a product. Um like your understanding as an entrepreneur of like, you know, who is your user, like what is their problem, like what can you do to help them, uh, and then like determining whether or not this AI stuff is actually a useful piece of infrastructure to go solve that user problem. Like that that is something that is unchanged.
22:25 Speaker 0 Uh like it it's it's yeah, the AI is just uh like a new and interesting piece of infrastructure that has come into existence that lets you solve a new category of problems or like you know solve an old old category of problems in better ways.
22:42 Speaker 1 Well, look, um Kevin, this is great. Thank you very much for taking the time and joining us today. I greatly appreciate you sharing the vision you have there at Microsoft, some of the tools and some of the human aspects of bringing this technology to life. It's great to see you. I look forward to next time.
22:59 Speaker 0 Thank you so much for inviting me.
Previous Engineering Rapidly Reusable Rockets Next AI Revolution: Top Lessons from OpenAI, Anthropic, CharacterAI, & More Page updated August 25, 2026 · transcript and summary are machine-generated