WEBVTT - Microsoft Looks at Whether AI Will Fix Work

0:00:00.120 --> 0:00:03.960
<v Speaker 1>This is Bloomberg Business Week with Carol Messer and Tim

0:00:04.000 --> 0:00:08.240
<v Speaker 1>Stenebek on Bloomberg Radio. All right, Zoom was certainly an

0:00:08.280 --> 0:00:11.440
<v Speaker 1>indication of how we were working during the pandemic. But

0:00:11.480 --> 0:00:13.480
<v Speaker 1>we've really leaned on the team over at Microsoft three

0:00:13.520 --> 0:00:16.319
<v Speaker 1>sixty five for how we are working, yes, then, but

0:00:16.480 --> 0:00:20.360
<v Speaker 1>also today, and they've got a new report out that

0:00:21.040 --> 0:00:23.319
<v Speaker 1>finds that there's not necessarily any need to be worried

0:00:23.320 --> 0:00:26.080
<v Speaker 1>about AI. Of course, to be fair, Microsoft is deep

0:00:26.079 --> 0:00:29.280
<v Speaker 1>into AI, really awakening the world earlier this year with

0:00:29.320 --> 0:00:32.120
<v Speaker 1>its ten billion dollar investment into Open AI's chat GBT.

0:00:32.360 --> 0:00:34.440
<v Speaker 1>So even so, our next guest is here to help

0:00:34.479 --> 0:00:37.760
<v Speaker 1>answer the question will AI fix work? And great to

0:00:37.800 --> 0:00:40.320
<v Speaker 1>have back with us. Jared's Pataro. He is Corporate VP

0:00:40.479 --> 0:00:43.800
<v Speaker 1>of Modern Work and Business Applications at Microsoft three sixty five.

0:00:43.840 --> 0:00:47.199
<v Speaker 1>Once again on Zoom in Redmand Washington. Jared, how are you.

0:00:48.240 --> 0:00:50.360
<v Speaker 2>Oh, I'm doing great. Great to be back with you, Carol.

0:00:50.240 --> 0:00:51.959
<v Speaker 1>Yeah, great to have you here with us. So tell

0:00:52.000 --> 0:00:54.440
<v Speaker 1>us about this inquiry into AI? What you guys?

0:00:54.480 --> 0:00:54.680
<v Speaker 2>I mean?

0:00:54.680 --> 0:00:56.440
<v Speaker 1>The question is I've got the report in front of me,

0:00:56.480 --> 0:00:59.520
<v Speaker 1>Will AI fix work? And I think everybody's trying to

0:00:59.520 --> 0:01:01.480
<v Speaker 1>figure it out. I kind of can't wait because I

0:01:01.480 --> 0:01:04.440
<v Speaker 1>think it can be an assist in my job. But

0:01:04.520 --> 0:01:07.080
<v Speaker 1>tell us about what you looked into, questions you asked,

0:01:07.120 --> 0:01:08.319
<v Speaker 1>and who you talk to.

0:01:09.280 --> 0:01:12.119
<v Speaker 2>Maybe let's look at what's broken first. So this particular

0:01:12.160 --> 0:01:14.800
<v Speaker 2>survey is our annual survey thirty one thousand people across

0:01:14.840 --> 0:01:17.840
<v Speaker 2>thirty one countries, and the first number that popped out

0:01:17.880 --> 0:01:20.760
<v Speaker 2>to me is exactly that fix that's needed. Sixty four

0:01:20.880 --> 0:01:23.520
<v Speaker 2>percent of people who responded told us that they just

0:01:23.520 --> 0:01:26.320
<v Speaker 2>don't have the time or energy to get their jobs done.

0:01:26.319 --> 0:01:28.600
<v Speaker 2>And we thought that that was really interesting. Coming out

0:01:28.600 --> 0:01:30.640
<v Speaker 2>of the pandemic. People are tired, but it seemed to

0:01:30.720 --> 0:01:32.760
<v Speaker 2>be more than that. So we combine that with our

0:01:32.800 --> 0:01:35.440
<v Speaker 2>telemetry data. This is the cloud data where we see

0:01:35.480 --> 0:01:37.800
<v Speaker 2>how people are working, what they're up to. We found

0:01:37.840 --> 0:01:40.560
<v Speaker 2>that up to about sixty percent of the average day,

0:01:40.640 --> 0:01:44.800
<v Speaker 2>average workers day is spent communicating and coordinating. Only forty

0:01:44.800 --> 0:01:46.800
<v Speaker 2>percent is kind of spent on that day job, the

0:01:46.840 --> 0:01:49.000
<v Speaker 2>thing that they're supposed to provide to the organization. So

0:01:49.560 --> 0:01:52.680
<v Speaker 2>really interesting setup right now as people are feeling tired

0:01:52.720 --> 0:01:53.840
<v Speaker 2>and for good reason.

0:01:54.120 --> 0:01:56.680
<v Speaker 3>And Jery something that strikes me anytime I talk to

0:01:56.720 --> 0:01:59.200
<v Speaker 3>anyone they're always concerned about when it comes to AI,

0:01:59.240 --> 0:02:01.480
<v Speaker 3>maybe they could attenduly be replaced when it comes to

0:02:01.520 --> 0:02:03.800
<v Speaker 3>their job. Looking at your study, it's interesting as far

0:02:03.840 --> 0:02:06.760
<v Speaker 3>as how AI is poised to create this whole new

0:02:06.800 --> 0:02:09.960
<v Speaker 3>way of working, what would be the counter to that,

0:02:10.040 --> 0:02:12.640
<v Speaker 3>and the benefits of AI were you might not necessarily

0:02:12.639 --> 0:02:14.360
<v Speaker 3>be losing your job, but it can actually assist you.

0:02:15.280 --> 0:02:17.400
<v Speaker 2>Well, we'll start with the fear. There's still a little

0:02:17.400 --> 0:02:19.519
<v Speaker 2>bit of fear there. Almost fifty percent, it was forty

0:02:19.600 --> 0:02:21.520
<v Speaker 2>nine percent of people who responded told us that they

0:02:21.520 --> 0:02:23.720
<v Speaker 2>were afraid that AI would come and take their job,

0:02:24.280 --> 0:02:28.000
<v Speaker 2>but that was overshadowed by another number, a whopping seventy

0:02:28.040 --> 0:02:31.320
<v Speaker 2>percent told us that despite that fear, they actually would

0:02:31.440 --> 0:02:34.640
<v Speaker 2>outsource everything they could to an AI assistant just to

0:02:34.680 --> 0:02:37.680
<v Speaker 2>help them relieve the burden. So there's this sense that

0:02:37.720 --> 0:02:40.400
<v Speaker 2>their optimism, or at least their need for help for

0:02:40.560 --> 0:02:43.520
<v Speaker 2>AI based help, really outweighs their fears. And that's a

0:02:43.560 --> 0:02:46.120
<v Speaker 2>pretty interesting nuance on many the headlines that we've been

0:02:46.120 --> 0:02:47.280
<v Speaker 2>hearing for the last couple of months.

0:02:47.320 --> 0:02:49.760
<v Speaker 1>Well, I'm thinking about email alone. I mean, I'm going

0:02:49.800 --> 0:02:51.360
<v Speaker 1>to be quite honest with you. There's a ton of

0:02:51.360 --> 0:02:53.400
<v Speaker 1>email that I never even get to read because it's

0:02:53.520 --> 0:02:55.720
<v Speaker 1>just a ton of stuff that people are either pitching

0:02:55.919 --> 0:02:58.720
<v Speaker 1>or research. And I definitely have figured out ways to

0:02:58.760 --> 0:03:00.840
<v Speaker 1>kind of weed through do certain searches so that I

0:03:00.880 --> 0:03:03.200
<v Speaker 1>really make sure I don't miss some of the important stuff.

0:03:03.200 --> 0:03:06.680
<v Speaker 1>But how could AI ultimately help us in filtering through

0:03:07.160 --> 0:03:10.680
<v Speaker 1>what is dumped into our email boxes on a regular basis.

0:03:11.680 --> 0:03:14.160
<v Speaker 2>Well, many of your your listeners are probably familiar with

0:03:14.240 --> 0:03:18.360
<v Speaker 2>chat GPT, introduced last year, and it's particularly good, it's summarizing,

0:03:18.440 --> 0:03:20.320
<v Speaker 2>and so I've been using a product that we call

0:03:20.360 --> 0:03:23.280
<v Speaker 2>Copilot Microsoft through sixty five copil it essentially takes the

0:03:23.680 --> 0:03:27.200
<v Speaker 2>models behind chat GPT, these large language models and integrates

0:03:27.200 --> 0:03:30.280
<v Speaker 2>it directly into your email into Outlook. And this does

0:03:30.320 --> 0:03:32.560
<v Speaker 2>amazing things. It allows you, for instance, get a long

0:03:32.720 --> 0:03:35.080
<v Speaker 2>those long email threads where you're supposed to go read

0:03:35.080 --> 0:03:37.440
<v Speaker 2>from the bottom up, and it can summarize that just

0:03:37.520 --> 0:03:40.040
<v Speaker 2>in one click. And then once it's summarized, it actually

0:03:40.080 --> 0:03:42.520
<v Speaker 2>gives you a couple of different prompts or those are

0:03:42.720 --> 0:03:45.240
<v Speaker 2>different things you can choose from to answer, so it

0:03:45.280 --> 0:03:47.600
<v Speaker 2>will author an answer, allow you to make it longer, shorter,

0:03:47.960 --> 0:03:51.320
<v Speaker 2>more formal, more casual, it is changed the way I

0:03:51.320 --> 0:03:53.880
<v Speaker 2>do email. I never want to do email without it again,

0:03:53.960 --> 0:03:56.320
<v Speaker 2>so it definitely changes work habits and practices.

0:03:56.440 --> 0:03:58.400
<v Speaker 1>Well, so just just share with me, because, like I

0:03:58.400 --> 0:04:00.240
<v Speaker 1>said at the top, I mean to be fair. You

0:04:00.280 --> 0:04:03.040
<v Speaker 1>guys are all in on AI, no doubt about it.

0:04:03.080 --> 0:04:07.600
<v Speaker 1>And in many ways that news of the Microsoft investment

0:04:07.680 --> 0:04:09.720
<v Speaker 1>really did kind of wake up the world in terms

0:04:09.760 --> 0:04:12.720
<v Speaker 1>of what you folks were doing. Specifically, Now everybody's talking

0:04:12.720 --> 0:04:15.160
<v Speaker 1>about it. As you guys have been playing around with it,

0:04:15.240 --> 0:04:18.400
<v Speaker 1>what are you finding are the great capabilities? What are

0:04:18.400 --> 0:04:19.839
<v Speaker 1>still the things that need to be worked out?

0:04:21.520 --> 0:04:24.039
<v Speaker 2>The five seconds to wow type demo I do with

0:04:24.120 --> 0:04:26.440
<v Speaker 2>customers is in a meeting. So in a meeting like

0:04:26.480 --> 0:04:28.360
<v Speaker 2>this that would be just a normal business meeting, we're

0:04:28.360 --> 0:04:31.479
<v Speaker 2>trading kind of conversation back and forward. And what it

0:04:31.520 --> 0:04:33.760
<v Speaker 2>can do there is it can literally listen in as

0:04:33.800 --> 0:04:35.960
<v Speaker 2>if it were an assistant to the meeting and summarize.

0:04:36.000 --> 0:04:38.840
<v Speaker 2>That means it can write notes. It can respond when

0:04:38.880 --> 0:04:40.880
<v Speaker 2>I ask questions like what did Carol say again? Or

0:04:40.920 --> 0:04:43.479
<v Speaker 2>how did Jared answer this question? It can allow me,

0:04:43.520 --> 0:04:45.440
<v Speaker 2>for instance, to identify, you know, what are the two

0:04:45.480 --> 0:04:47.960
<v Speaker 2>sides of this argument and who's in favor of which

0:04:48.000 --> 0:04:50.919
<v Speaker 2>one it actually is incredibly valuable in a meeting. And

0:04:50.960 --> 0:04:52.680
<v Speaker 2>then even better than that, Carol, the thing that I

0:04:52.720 --> 0:04:56.080
<v Speaker 2>love is it allows me to attend meetings, so I

0:04:56.120 --> 0:04:59.000
<v Speaker 2>can not attend a meeting and then using that same

0:04:59.160 --> 0:05:02.200
<v Speaker 2>GPT larg language model, query the meeting after the fact,

0:05:02.279 --> 0:05:05.200
<v Speaker 2>did they ever mention my name or any decisions made?

0:05:05.240 --> 0:05:08.119
<v Speaker 2>What decisions? Why did they make them? So it opens

0:05:08.200 --> 0:05:11.000
<v Speaker 2>up these brand new possibilities for work because it kind

0:05:11.040 --> 0:05:13.720
<v Speaker 2>of blurs time and space. You know, it's as if

0:05:13.760 --> 0:05:16.320
<v Speaker 2>you could be there, but you're not there, and you're

0:05:16.360 --> 0:05:19.000
<v Speaker 2>able to use the power of AI here to kind

0:05:19.000 --> 0:05:21.279
<v Speaker 2>of get into the dynamics of people even just having

0:05:21.320 --> 0:05:24.200
<v Speaker 2>regular conversations. So just one example, but it's integrated across

0:05:24.240 --> 0:05:27.680
<v Speaker 2>all of our products. We talked about email or PowerPoint,

0:05:27.839 --> 0:05:28.640
<v Speaker 2>all of those things.

0:05:28.680 --> 0:05:31.440
<v Speaker 1>What's the thing though, that is still tricky because we

0:05:31.480 --> 0:05:34.120
<v Speaker 1>talk about AI kind of working alongside of us and

0:05:34.440 --> 0:05:37.120
<v Speaker 1>we don't have to go to the meeting Yahoo. But

0:05:37.320 --> 0:05:41.000
<v Speaker 1>I do wonder what are the things that you know,

0:05:41.080 --> 0:05:44.080
<v Speaker 1>people say you can't really you know, duplicate a human

0:05:44.120 --> 0:05:45.720
<v Speaker 1>in terms of their perception of things and so on

0:05:45.760 --> 0:05:47.720
<v Speaker 1>and so forth. So what do would you say that

0:05:47.760 --> 0:05:49.760
<v Speaker 1>we still need to be cautious about or we're still

0:05:49.760 --> 0:05:51.640
<v Speaker 1>figuring out, or you guys are still figuring out when

0:05:51.680 --> 0:05:52.320
<v Speaker 1>it comes to this.

0:05:53.440 --> 0:05:56.240
<v Speaker 2>Very stud question, because it turns out while these things

0:05:56.279 --> 0:05:59.480
<v Speaker 2>are incredibly useful, they do make mistakes. And these things,

0:05:59.520 --> 0:06:02.640
<v Speaker 2>I mean the AI assistance that we have, so they

0:06:02.640 --> 0:06:05.640
<v Speaker 2>get it wrong sometimes. In fact, the technical term is hallucination.

0:06:05.800 --> 0:06:08.040
<v Speaker 2>If you believe that, that literally is the term they

0:06:08.040 --> 0:06:10.960
<v Speaker 2>make stuff up. So the reason that we use copilot

0:06:10.960 --> 0:06:13.200
<v Speaker 2>as a name for our product is really the signal

0:06:13.240 --> 0:06:15.640
<v Speaker 2>to the user, hey, this is an autopilot. You can't

0:06:15.680 --> 0:06:17.600
<v Speaker 2>just take the answers and say there you go, I'm

0:06:17.640 --> 0:06:19.480
<v Speaker 2>going to fire this off as an email. You really

0:06:19.520 --> 0:06:21.320
<v Speaker 2>have to be in the driver's seat. You have to

0:06:21.360 --> 0:06:24.120
<v Speaker 2>make the decisions. But the great news is my experience

0:06:24.160 --> 0:06:25.960
<v Speaker 2>has been when it is wrong, it's what we call

0:06:26.080 --> 0:06:28.120
<v Speaker 2>usefully wrong. So it doesn't just make things up. To

0:06:28.200 --> 0:06:30.800
<v Speaker 2>make things up. Sometimes we'll feel in details where it

0:06:30.800 --> 0:06:32.680
<v Speaker 2>didn't have all the data. You can go in and

0:06:32.680 --> 0:06:34.920
<v Speaker 2>correct that, tweak that, and then it's put you ahead

0:06:34.920 --> 0:06:35.760
<v Speaker 2>of the game anyway.

0:06:36.000 --> 0:06:37.720
<v Speaker 3>So how long do you think this could take? But

0:06:37.720 --> 0:06:41.720
<v Speaker 3>before it's more widespread with users that could potentially use

0:06:41.760 --> 0:06:43.400
<v Speaker 3>this type of things in meetings when you do have

0:06:43.480 --> 0:06:45.799
<v Speaker 3>still some of these hiccups and glitches that could happen

0:06:45.839 --> 0:06:46.200
<v Speaker 3>with AI.

0:06:47.400 --> 0:06:49.279
<v Speaker 2>We just introduced a couple of weeks ago that we're

0:06:49.279 --> 0:06:51.839
<v Speaker 2>expanding what we call our Preview program to six hundred

0:06:51.920 --> 0:06:54.279
<v Speaker 2>enterprises around the world, so we'll have hundreds of thousands

0:06:54.320 --> 0:06:56.840
<v Speaker 2>of users here in the coming months. So that's kind

0:06:56.839 --> 0:06:58.640
<v Speaker 2>of the scale that it's out right at right now.

0:06:58.680 --> 0:07:00.960
<v Speaker 2>But it's moving very quickly. My team, we talk about

0:07:01.040 --> 0:07:04.040
<v Speaker 2>AI time, the cycle time of the industry has really

0:07:04.080 --> 0:07:05.880
<v Speaker 2>increased as we're able to build on top of these

0:07:05.920 --> 0:07:08.320
<v Speaker 2>technologies really quickly. So I don't think it'll be a

0:07:08.320 --> 0:07:09.920
<v Speaker 2>matter of years. I do think it'll be a matter

0:07:10.000 --> 0:07:12.040
<v Speaker 2>of months until you start to see these tools being

0:07:12.120 --> 0:07:13.080
<v Speaker 2>used all over the place.

0:07:13.200 --> 0:07:15.240
<v Speaker 1>You know, for someone who has been you know, focusing

0:07:15.280 --> 0:07:18.679
<v Speaker 1>on workplace trends and coming off a really crazy period,

0:07:18.800 --> 0:07:21.760
<v Speaker 1>you know, meaning the pandemic specifically, and we're able to

0:07:21.800 --> 0:07:24.560
<v Speaker 1>track a lot of things. How are you kind of

0:07:24.760 --> 0:07:31.120
<v Speaker 1>placing generative AI machine learning in kind of I don't know,

0:07:31.240 --> 0:07:35.040
<v Speaker 1>is it akin to some other great or interesting development

0:07:35.280 --> 0:07:37.120
<v Speaker 1>in terms of how we work, Like, how do you

0:07:37.160 --> 0:07:40.400
<v Speaker 1>position it historically or technologically for that matter.

0:07:41.280 --> 0:07:43.880
<v Speaker 2>We think it's as big as the PC revolution or

0:07:43.920 --> 0:07:46.040
<v Speaker 2>as big as the Internet. We think it's that big.

0:07:46.200 --> 0:07:49.040
<v Speaker 2>You know, if I again frame it, we would say, Wow,

0:07:49.080 --> 0:07:52.800
<v Speaker 2>you couldn't write a script like this if you were trying. Really,

0:07:52.800 --> 0:07:55.960
<v Speaker 2>what the pandemic did is is rewire the where and

0:07:56.000 --> 0:07:58.800
<v Speaker 2>the when of work, So it changed entirely around the world.

0:07:58.840 --> 0:08:01.559
<v Speaker 2>You know, those two mentions of work. All of a sudden,

0:08:01.600 --> 0:08:04.520
<v Speaker 2>AI is changing the how of work. By the time

0:08:04.560 --> 0:08:07.520
<v Speaker 2>we get kind of roughly four years out from the pandemic,

0:08:07.560 --> 0:08:10.720
<v Speaker 2>the combination of those two things will have really just

0:08:10.840 --> 0:08:13.680
<v Speaker 2>changed patterns practices associated with work. My kids, as they

0:08:13.680 --> 0:08:15.760
<v Speaker 2>go to work, they're not going to enter a work

0:08:15.840 --> 0:08:18.520
<v Speaker 2>environment anything like I did you know twenty five years ago?

0:08:18.760 --> 0:08:22.520
<v Speaker 1>Is that good or bad? Bodily because you know, things

0:08:22.520 --> 0:08:25.400
<v Speaker 1>are never right black and white like it's it's that's right.

0:08:25.520 --> 0:08:27.960
<v Speaker 1>We're worried about people not really talking to each other.

0:08:28.160 --> 0:08:30.640
<v Speaker 1>I have a colleague, Matt Miller, who laughs that when

0:08:30.640 --> 0:08:32.480
<v Speaker 1>we all get on a zoom call and we're literally

0:08:32.559 --> 0:08:35.280
<v Speaker 1>like right next to each other, but I'm sitting there typing,

0:08:35.480 --> 0:08:38.839
<v Speaker 1>you know, notes, and you know, it's just interesting.

0:08:40.480 --> 0:08:42.160
<v Speaker 2>My response to that is I would say, you know,

0:08:42.240 --> 0:08:44.679
<v Speaker 2>AI giveth an, AI taketh away, Like, we have this

0:08:44.760 --> 0:08:46.760
<v Speaker 2>moment and the way we think about it is what

0:08:46.800 --> 0:08:48.920
<v Speaker 2>have we gained and what have we lost? And how

0:08:48.920 --> 0:08:50.800
<v Speaker 2>do we lose as little as possible and how do

0:08:50.800 --> 0:08:52.960
<v Speaker 2>we gain as much as possible? But you're exactly right,

0:08:53.080 --> 0:08:54.880
<v Speaker 2>these big if it's as big as the PC, you know,

0:08:54.880 --> 0:08:56.920
<v Speaker 2>if you really believe that it's that type of shift,

0:08:57.200 --> 0:08:59.280
<v Speaker 2>then you do have to recognize, Wow, this is probably

0:08:59.320 --> 0:09:01.880
<v Speaker 2>the beginning of an era, a new era, which is

0:09:01.920 --> 0:09:03.559
<v Speaker 2>be really thoughtful about what we want to make sure

0:09:03.600 --> 0:09:05.360
<v Speaker 2>we don't lose in that process. But that will take

0:09:05.360 --> 0:09:07.440
<v Speaker 2>all of us, you know, we're all learning as we go. Yeah.

0:09:07.520 --> 0:09:08.959
<v Speaker 1>Right, And the more we use it though, right, the

0:09:09.480 --> 0:09:12.280
<v Speaker 1>smarter or more specific it becomes. Jared, Nice to check

0:09:12.280 --> 0:09:15.000
<v Speaker 1>in with you again. Jared's Pataro. He's corporate vice president,

0:09:15.080 --> 0:09:17.440
<v Speaker 1>head of Modern Work at Microsoft three sixty five on

0:09:17.559 --> 0:09:22.040
<v Speaker 1>Zoom from Redman, Washington. Yeah, I feel like.

0:09:22.000 --> 0:09:24.040
<v Speaker 3>I have that same email problem as you, Carol.

0:09:24.320 --> 0:09:27.280
<v Speaker 1>It's crazy. Now. I've said to this though a million times,

0:09:27.280 --> 0:09:30.240
<v Speaker 1>like I've thought about AI the ability to you know,

0:09:30.320 --> 0:09:32.120
<v Speaker 1>I like to write kind of intros into our guests

0:09:32.160 --> 0:09:34.720
<v Speaker 1>and stuff, and just the ability to maybe shoot some information,

0:09:34.800 --> 0:09:36.880
<v Speaker 1>say here, just draft something, I'll edit it. I'll go

0:09:36.920 --> 0:09:39.320
<v Speaker 1>through it and make sure it's But to kind of

0:09:39.320 --> 0:09:41.800
<v Speaker 1>have that framework to try to kick off with would

0:09:41.840 --> 0:09:43.080
<v Speaker 1>be really cool. Yeah.

0:09:43.200 --> 0:09:45.920
<v Speaker 3>Yeah, all my distribution lists that I'm on for banks,

0:09:46.600 --> 0:09:46.640
<v Speaker 3>I

0:09:46.800 --> 0:09:48.920
<v Speaker 1>Just like to filter through staff like, would be really

0:09:49.000 --> 0:09:49.520
<v Speaker 1>kind of cool.