WEBVTT 00:00:00.000 --> 00:00:01.917 Hi, Trevor. >> How are you? It's great seeing you. 00:00:02.000 --> 00:00:05.277 Great to meet you. Likewise. Have a seat. 00:00:05.360 --> 00:00:06.997 Welcome to Australia. Thank you so much. 00:00:07.080 --> 00:00:09.597 It's been fun. We've got lots to talk about. You've made a lot of 00:00:09.680 --> 00:00:12.877 announcements. I want to talk about AI, where it's at. I've got some fun things 00:00:12.960 --> 00:00:16.597 to show you, but just I want to start by thinking about Microsoft at this 00:00:16.680 --> 00:00:18.237 position in time. I was thinking about 00:00:18.320 --> 00:00:20.837 it during the week. We went from command 00:00:20.920 --> 00:00:24.477 prompt MS-DOS to to Windows 3.11. This 00:00:24.560 --> 00:00:28.677 feels like a similar transition, you know, from being a 00:00:28.760 --> 00:00:33.557 a software operating system company to being AI first. Is that a a safe 00:00:33.640 --> 00:00:35.477 assessment? Yeah, I mean I always sort 00:00:35.560 --> 00:00:38.157 of think about Microsoft's core identity 00:00:38.240 --> 00:00:41.797 is a platform and a partner company, 00:00:41.880 --> 00:00:46.397 right? I mean I look back and, you know, um you know, the anecdote I was sharing 00:00:46.480 --> 00:00:50.117 at the keynote even is the first time I was in Sydney doing some demos was, you 00:00:50.200 --> 00:00:53.957 know, you're trying to talk about Visual Basic applications inside of Excel that 00:00:54.040 --> 00:00:57.837 others can use to build applications, right? I mean that is at the core. So, 00:00:57.920 --> 00:01:01.957 even though yes, it is a shift, a massive shift, very different, it may 00:01:02.040 --> 00:01:05.517 feel like the internet or um 00:01:05.600 --> 00:01:07.077 as you said, the move from DOS to 00:01:07.160 --> 00:01:12.557 Windows and maybe bigger even, quite frankly, but the core threadline for me 00:01:12.640 --> 00:01:16.597 is how have we enabled others to build more 00:01:16.680 --> 00:01:18.637 ambitious things than anything that we 00:01:18.720 --> 00:01:23.757 have built, right? That I think is the defining line of Microsoft. It was 00:01:23.840 --> 00:01:26.637 interesting listening to your keynote earlier today and there was people 00:01:26.720 --> 00:01:29.997 behind me literally saying, "Oh, that's cool." And you know, there those kind of 00:01:30.080 --> 00:01:33.837 things under their breath. And that's a crowd of people who work uh 00:01:33.920 --> 00:01:37.317 in software and they're they're developing tools. For the average person 00:01:37.400 --> 00:01:40.677 in Australia, let alone around the world, is AI something that they should 00:01:40.760 --> 00:01:43.597 be worried about or concerned about? 00:01:43.680 --> 00:01:48.717 I would say the most important thing for us, you know, collectively, both in the 00:01:48.800 --> 00:01:52.517 tech industry but in the broader society, is to make sure that the 00:01:52.600 --> 00:01:57.397 benefits of AI are really delivered to everybody, right? Because if we don't, 00:01:57.480 --> 00:02:01.597 uh then of course, it's not going to have the social permission. Uh what does 00:02:01.680 --> 00:02:03.277 that mean? It means you want to be 00:02:03.360 --> 00:02:06.197 really a good users of water and energy 00:02:06.280 --> 00:02:09.837 to create uh AI that then is helping 00:02:09.920 --> 00:02:14.037 with education outcomes, health outcomes, public sector is becoming more 00:02:14.120 --> 00:02:16.837 efficient, uh the companies that are 00:02:16.920 --> 00:02:21.077 using AI are becoming more productive, they're driving economic growth that 00:02:21.160 --> 00:02:25.517 then is making its way to people's, you know, lives because that's the way 00:02:25.600 --> 00:02:29.357 permission is earned, right? Because you I I feel like at this point the 00:02:29.440 --> 00:02:33.637 narrative around AI is, you know, such, which I'm glad, where we're talking 00:02:33.720 --> 00:02:35.237 about both the benefits and the 00:02:35.320 --> 00:02:38.117 unintended consequences versus waiting. 00:02:38.200 --> 00:02:40.877 And that debate is a good debate to have 00:02:40.960 --> 00:02:45.197 uh versus just indexing on any one of them, right? Because just to say AI is 00:02:45.280 --> 00:02:49.357 all about sort of the fear of AI, that won't help because it's inevitable that 00:02:49.440 --> 00:02:53.717 this is a secular change. The question is how do we drive that change such that 00:02:53.800 --> 00:02:57.597 the benefits really are getting through to everybody. Feels like a secular 00:02:57.680 --> 00:02:59.357 change and also a societal change. 00:02:59.440 --> 00:03:01.637 That's right. Which also brings up one of the other concerns I think in 00:03:01.720 --> 00:03:05.517 society, which is jobs. People One of the most common questions I get when 00:03:05.600 --> 00:03:08.797 interviewed about it is, you know, but it's going to take jobs or cost jobs. 00:03:08.880 --> 00:03:12.797 How do you answer that to people in in the general public about where jobs are 00:03:12.880 --> 00:03:15.637 going to go with AI? >> Yeah, I mean I think the one thing is, 00:03:15.720 --> 00:03:20.597 you know, to start with, um I think in recognizing even right now in the state 00:03:20.680 --> 00:03:22.517 of AI and quite frankly even for the 00:03:22.600 --> 00:03:25.397 foreseeable future is more about what 00:03:25.480 --> 00:03:30.397 I'll call task level automation inside of jobs, right? You know, yesterday I 00:03:30.480 --> 00:03:34.117 was talking to some uh local Australian startup doing a 00:03:34.200 --> 00:03:37.197 fantastic job of building a scribe tool 00:03:37.280 --> 00:03:39.317 for doctors. And uh one of the things he 00:03:39.400 --> 00:03:42.277 was describing to me was how much it's 00:03:42.360 --> 00:03:46.477 benefiting the physi- physicians to by reducing their paperwork burden and 00:03:46.560 --> 00:03:47.877 helping them spend more time with 00:03:47.960 --> 00:03:50.197 patients, right? So, that's the type of 00:03:50.280 --> 00:03:53.837 task that they did that {quote} AI does 00:03:53.920 --> 00:03:58.277 so that they can do the real job, which is take care of patients. And that's 00:03:58.360 --> 00:04:00.237 happening even in office work, uh right? 00:04:00.320 --> 00:04:04.117 When I think about even my own job, right? You know, was I meant to mostly 00:04:04.200 --> 00:04:07.117 triage email or can I do 00:04:07.200 --> 00:04:09.717 uh things where I can focus my attention 00:04:09.800 --> 00:04:12.317 on being able to really get the deep 00:04:12.400 --> 00:04:16.837 insights, pass better judgment on things that I need to do so and so on. So, I 00:04:16.920 --> 00:04:20.197 think that that's one aspect of it. But at the same time, I'm clear-eyed that 00:04:20.280 --> 00:04:23.517 that's why I think even the skills announcement we had, which is about how 00:04:23.600 --> 00:04:27.957 do we really make sure that as even new job opportunities get created and maybe 00:04:28.040 --> 00:04:32.477 in different forms, we have the skills required to be able to take those jobs. 00:04:32.560 --> 00:04:35.517 >> And you've made a big commitment here in Australia to that skilling. That's 00:04:35.600 --> 00:04:38.517 right. We have now, you know, committed to, you know, training uh more than 3 00:04:38.600 --> 00:04:41.917 million people uh in Australia uh along with obviously 00:04:42.000 --> 00:04:45.757 all the other investments we're making around uh the data centers. So, big 00:04:45.840 --> 00:04:50.117 investment in skills for Australians and that's coming from Microsoft. Is that Is 00:04:50.200 --> 00:04:53.517 there a reason why Australia gets that investment? Is that a a 00:04:53.600 --> 00:04:57.077 an endorsement of our our people or our industry? For sure. I mean I mean 00:04:57.160 --> 00:05:00.997 Australia is a very important market for us. Um and whenever I think about it, 00:05:01.080 --> 00:05:04.797 again, I go back to, you know, as a multinational company, how are we going 00:05:04.880 --> 00:05:08.957 to have a license to operate in Australia? It's primarily by making sure 00:05:09.040 --> 00:05:12.637 uh that we invest in this country, not only our own people but also our data 00:05:12.720 --> 00:05:16.277 centers and we're doing things like skilling and things that we're doing in 00:05:16.360 --> 00:05:19.357 cybersecurity with the Signals Directorate. These are all important 00:05:19.440 --> 00:05:21.997 investments ultimately to create local 00:05:22.080 --> 00:05:24.837 surplus, right? When that startup, when 00:05:24.920 --> 00:05:29.077 uh you know, Australian Post can use this in order to be able to, you know, 00:05:29.160 --> 00:05:31.757 have accessibility be core to them and 00:05:31.840 --> 00:05:34.677 their co-pilot adoption or 00:05:34.760 --> 00:05:38.997 um CBA uses it for their customer applications or what have you. And these 00:05:39.080 --> 00:05:41.677 are things ultimately that will drive our business. 00:05:41.760 --> 00:05:44.157 >> And did you get a sense You obviously met with the Prime Minister to make some 00:05:44.240 --> 00:05:47.837 of those announcements. Is Is the Australian government ready for AI? Uh 00:05:47.920 --> 00:05:50.717 are you helping them along the way or are they coming to you for help? 00:05:50.800 --> 00:05:54.317 >> Well, first of all, I think in Australia, I would say the government is 00:05:54.400 --> 00:05:58.597 taken uh really a a progressive uh you know, approach to sort of making 00:05:58.680 --> 00:06:02.397 sure that they're thinking about what is the the positives of AI, what are some 00:06:02.480 --> 00:06:05.637 of the things that they need to get ahead in terms of whether it's safety or 00:06:05.720 --> 00:06:09.437 whether it's around skills um and what have you. And it's fantastic 00:06:09.520 --> 00:06:12.077 to see their adoption and thoughtfulness 00:06:12.160 --> 00:06:16.677 uh when it comes to AI. Uh and also the human capital in Australia, right? Which 00:06:16.760 --> 00:06:21.037 is, as I said, is whether it's in these large-scale multinational companies that 00:06:21.120 --> 00:06:25.437 are operating or the startups that are getting born here and not only building 00:06:25.520 --> 00:06:27.837 for Australia but for the world, uh the 00:06:27.920 --> 00:06:32.477 high ambition here. This is the first time I'd say in the world uh and you see 00:06:32.560 --> 00:06:34.037 it in Australia that the rate of 00:06:34.120 --> 00:06:39.637 diffusion is such that there's no sort of I would say time lag between what may 00:06:39.720 --> 00:06:42.917 be happening on the West Coast of the United States or the East Coast of China 00:06:43.000 --> 00:06:46.557 and what's happening in Australia. Now, I want to take a minute. You spent some 00:06:46.640 --> 00:06:49.677 time yesterday with a lot of Aussie businesses looking at how they're using 00:06:49.760 --> 00:06:53.437 AI. And I think the average Aussie, let alone average person in the world, 00:06:53.520 --> 00:06:56.677 thinks of AI as ask a question, get a more thoughtful answer. It's very basic 00:06:56.760 --> 00:06:59.877 from a couple of years ago. But there's a thing called agents. You you know all 00:06:59.960 --> 00:07:02.837 this, but and you had the Commonwealth Bank talking about how they're using an 00:07:02.920 --> 00:07:07.837 agent in in a live chat. So, what I did, I took Copilot Studio and I tried to 00:07:07.920 --> 00:07:12.317 create an agent for my family. So, three kids, my wife and I, and I thought about 00:07:12.400 --> 00:07:15.957 my busy life, my busy diary, and we have a shared family calendar. So, I wanted 00:07:16.040 --> 00:07:17.237 to demonstrate it to you if I could. 00:07:17.320 --> 00:07:19.877 >> Cool. Um it's called >> you build it? Well, I used Copilot 00:07:19.960 --> 00:07:24.557 Studio and I integrated uh the the Outlook um calendars. And so, we've 00:07:24.640 --> 00:07:26.197 called it Charles, our family butler. 00:07:26.280 --> 00:07:27.957 So, I'll give you a quick example. 00:07:28.040 --> 00:07:30.957 What's my schedule look like on Monday? 00:07:31.040 --> 00:07:34.117 And so, basically, it's now going to go and check my personal schedule and see 00:07:34.200 --> 00:07:37.677 how busy my day is, but it's also, cuz I've engaged it to always check the 00:07:37.760 --> 00:07:40.997 family calendar because I want to know what I'm going to be able to do and how 00:07:41.080 --> 00:07:44.157 I'm going to help the family. So, it's looking at my calendar. I think people 00:07:44.240 --> 00:07:46.997 need to understand that it's not just reading a calendar, it's trying to 00:07:47.080 --> 00:07:49.117 understand the schematics of the day. 00:07:49.200 --> 00:07:51.957 And so, now it's giving me a whole output of my day telling me what I've 00:07:52.040 --> 00:07:55.957 got on, but by default, it's telling me about the clashes and and problems that 00:07:56.040 --> 00:07:59.477 I might have. So, considerations and even making suggestions, "Would you like 00:07:59.560 --> 00:08:03.277 me to send Amanda, my wife, a quick note about the netball pickup clash?" But 00:08:03.360 --> 00:08:05.877 here, it gets even better. So, what I can also do is say, 00:08:05.960 --> 00:08:07.557 "Move my planning meeting to earlier in 00:08:07.640 --> 00:08:12.997 the afternoon and let my wife know that I can pick Victoria up for netball." 00:08:13.080 --> 00:08:17.557 So, basically now, it's going to change my calendar and it's going to send my 00:08:17.640 --> 00:08:21.197 wife an email letting her know that I've now changed my calendar and I can kick 00:08:21.280 --> 00:08:24.917 pick up Victoria from netball. So, I know it's a really simple example. 00:08:25.000 --> 00:08:28.437 >> a fantastic demo demo. I wish you had gotten you into my keynote and demo 00:08:28.520 --> 00:08:33.237 that. Because I think what you just showed is like the pri- First of all, 00:08:33.320 --> 00:08:36.957 the pride you have, the sense of empowerment you have that you built this 00:08:37.040 --> 00:08:40.677 application just like in the past you may have built an Excel spreadsheet or a 00:08:40.760 --> 00:08:44.037 Word document. You just literally created a piece of software. And there's 00:08:44.120 --> 00:08:47.517 an agent. There's an email to my wife, which to be honest, probably not 00:08:47.600 --> 00:08:49.997 appropriate. I would normally send a text, maybe, but just letting you know 00:08:50.080 --> 00:08:53.277 that I've adjusted my meeting so I can pick Victoria up from school on Monday 00:08:53.360 --> 00:08:57.597 and take her to netball. So, it's it's a great I feel like is that the next phase 00:08:57.680 --> 00:09:02.797 for AI? So, people to see it in real world usage of it? 00:09:02.880 --> 00:09:06.597 >> you captured it well, which is what is the output that you care about. In this 00:09:06.680 --> 00:09:09.917 case, you you didn't sort of say, "Oh, well, you know, my this coordination 00:09:10.000 --> 00:09:12.877 task I have around my family, I want to 00:09:12.960 --> 00:09:17.437 basically build an agent agentic system that just does that so that I am not 00:09:17.520 --> 00:09:21.917 sending mails or text messages or keeping track mentally." And then the 00:09:22.000 --> 00:09:25.997 agent offloads all of that, right? That ability. And that's what CBA is doing, 00:09:26.080 --> 00:09:30.317 right? For example, is one the cus- the customer directly is interacting with 00:09:30.400 --> 00:09:32.197 the bot they built, uh but then more 00:09:32.280 --> 00:09:34.677 interestingly, their own agents are 00:09:34.760 --> 00:09:40.397 using agents in order to be able to do the escalation cases with higher 00:09:40.480 --> 00:09:43.997 quality, >> Yeah. with less drudgery. And that type 00:09:44.080 --> 00:09:45.757 of improvement is what's fantastic. 00:09:45.840 --> 00:09:50.597 >> Commonwealth also talked about how it then enabled the human staff to have 00:09:50.680 --> 00:09:54.557 more, I guess, insightful information by the time they got to the staff for for 00:09:54.640 --> 00:09:56.597 support. >> Exactly. Exactly. And I think that the 00:09:56.680 --> 00:10:00.597 you know, the the complexity of some of our products and others inside of an 00:10:00.680 --> 00:10:03.957 organization are such that sometimes when even you're on the front lines in 00:10:04.040 --> 00:10:09.197 customer service or as a banker even, you want to be able to have this agentic 00:10:09.280 --> 00:10:12.517 sort of support that's right at your fingertips, right? 00:10:12.600 --> 00:10:14.997 In fact, that is the way you know, you know, we used to have this thing called 00:10:15.080 --> 00:10:18.277 information at your fingertips and now going forward it's about agents at your 00:10:18.360 --> 00:10:21.677 fingertips that are really helping you be more productive. You also met some 00:10:21.760 --> 00:10:25.797 people from Cricket Australia who are using Microsoft Copilot to analyze stats 00:10:25.880 --> 00:10:27.797 in real time during cricket matches. 00:10:27.880 --> 00:10:29.877 You're a massive cricket fan. That must have been impressive. You've got a 00:10:29.960 --> 00:10:33.197 cricket pitch at Microsoft head [laughter] office in Seattle. We do. We 00:10:33.280 --> 00:10:36.837 do and it's well used. In fact, I get to overlook it. Uh but yeah, I mean what 00:10:36.920 --> 00:10:38.637 Cricket Australia has done is I mean 00:10:38.720 --> 00:10:40.597 first of all, you know, cricket is a 00:10:40.680 --> 00:10:44.317 very sort of stats rich uh sport and 00:10:44.400 --> 00:10:47.877 they were able to take what's been a treasure trove of data that they have 00:10:47.960 --> 00:10:50.877 had since 1897 or something like that 00:10:50.960 --> 00:10:53.477 and parlay that into now this Cricket 00:10:53.560 --> 00:10:58.117 Australia app and they have the stats guru functionality and to be able to 00:10:58.200 --> 00:11:02.197 follow along ball by ball where the stats it it's just think about it, 00:11:02.280 --> 00:11:03.957 right? You know, it's not you know, that 00:11:04.040 --> 00:11:06.317 ability to generate the most interesting 00:11:06.400 --> 00:11:10.197 stat relevant to a context of one, you 00:11:10.280 --> 00:11:12.157 know, ball. That's just extraordinary. 00:11:12.240 --> 00:11:15.797 If we rewind 5 years, I don't think anyone in the general public saw AI 00:11:15.880 --> 00:11:20.467 coming. You maybe did um and other other people in business, but I don't I don't 00:11:20.550 --> 00:11:23.117 [clears throat] know what the next 3 to 4 years looks looks like. Do you have a 00:11:23.200 --> 00:11:28.237 crystal ball of how much more is going to change for average Australians around 00:11:28.320 --> 00:11:32.997 the use of AI or what's next? >> Yeah, I think the the speed with which 00:11:33.080 --> 00:11:36.517 some of these new capability jumps are happening are pretty stunning and that 00:11:36.600 --> 00:11:40.197 to your point was what you know, when I first read even the scaling laws paper, 00:11:40.280 --> 00:11:43.797 I felt like wow, this if this happens, this would be really interesting and 00:11:43.880 --> 00:11:48.397 it's turned out that it's held that line. Uh but the more 00:11:48.480 --> 00:11:52.397 important aspect I think we're reaching now is the models and in fact, people 00:11:52.480 --> 00:11:55.637 describe this as the model overhang, right? I think the capabilities are 00:11:55.720 --> 00:11:59.397 there. The question now is to be able to harness them. Like that's why I'm more 00:11:59.480 --> 00:12:01.677 impressed with your demo than any model 00:12:01.760 --> 00:12:03.437 capability, right? Because at the end of 00:12:03.520 --> 00:12:06.957 the day, did we all use it do things 00:12:07.040 --> 00:12:09.637 that are useful for the outcomes we care 00:12:09.720 --> 00:12:12.317 about individually and as organizations 00:12:12.400 --> 00:12:17.037 throughout society. And that I think is more going to be the defining thing 00:12:17.120 --> 00:12:21.757 about what's the real frontier of this technology. How quickly finally, how 00:12:21.840 --> 00:12:24.637 will people know what's real and what's not in the future? I feel like that's a 00:12:24.720 --> 00:12:27.917 real concern for people especially with what's happening on the internet. 100%. 00:12:28.000 --> 00:12:29.637 I mean, I think the the digital, you 00:12:29.720 --> 00:12:33.317 know, the provenance and trust are going 00:12:33.400 --> 00:12:35.277 to be the most important aspects, right? 00:12:35.360 --> 00:12:38.797 It's not just about deep fakes, but everything, right? For example, when you 00:12:38.880 --> 00:12:42.637 have lots and lots of agents, uh that's why I think even what we are 00:12:42.720 --> 00:12:45.277 doing with agent 365 as a way to give 00:12:45.360 --> 00:12:50.157 organizations that core control plane to you know, have the observability. 00:12:50.240 --> 00:12:53.437 And this is the issue with any new technology, right? As the the rapid 00:12:53.520 --> 00:12:57.237 adoption comes on the unintended consequences and this time around, we 00:12:57.320 --> 00:12:59.117 have to really get ahead, right? Whether 00:12:59.200 --> 00:13:01.517 it is about identity, 00:13:01.600 --> 00:13:04.317 provenance, security and governance 00:13:04.400 --> 00:13:06.877 policies and so that way that we can be 00:13:06.960 --> 00:13:08.757 secure about it. Can Australia be an AI 00:13:08.840 --> 00:13:11.477 powerhouse? I mean, I think that we 00:13:11.560 --> 00:13:14.437 think about AI is more than an 00:13:14.520 --> 00:13:18.557 accelerant to the comparative advantage of 00:13:18.640 --> 00:13:21.117 countries and companies that already 00:13:21.200 --> 00:13:23.317 have that going, right? So in some 00:13:23.400 --> 00:13:26.517 sense, AI is not some abstract new thing 00:13:26.600 --> 00:13:29.237 that sits outside of what is the rest of 00:13:29.320 --> 00:13:31.677 the economy, right? Australia today is a 00:13:31.760 --> 00:13:34.757 powerhouse on many fronts and that those 00:13:34.840 --> 00:13:38.637 fronts will get more amplified because 00:13:38.720 --> 00:13:42.117 of their use of AI and of course, you know, when it comes to even just 00:13:42.200 --> 00:13:46.797 generating the startups that, you know, go on to capture large swaths of market 00:13:46.880 --> 00:13:49.917 share in different categories can come out of Australia as much as any other 00:13:50.000 --> 00:13:53.277 place. I've watched you bounce from place to place over the last 2 days. 00:13:53.360 --> 00:13:56.157 You've you're a very busy man. You're you're the you're the boss of one of the 00:13:56.240 --> 00:13:59.597 biggest companies in the world. Do you get a chance to just look around Sydney 00:13:59.680 --> 00:14:00.797 at any point while you're here? 00:14:00.880 --> 00:14:03.557 >> I that is one of my dreams is to just come 00:14:03.640 --> 00:14:07.077 here when I'm looking around. But look, it's it's always a pleasure to be able 00:14:07.160 --> 00:14:11.157 to come and visit and quite frankly, meet with people. It's energizing to me 00:14:11.240 --> 00:14:15.837 to be able to see the demo that you did or the demos from the Commonwealth Bank 00:14:15.920 --> 00:14:19.317 or the Cricket Australia and it's always fun. Thanks, Satya. Thanks so much, 00:14:19.400 --> 00:14:22.560 Trevor. Thank you. >> Cheers.