1 00:00:00,040 --> 00:00:07,120 Today on episode #793 of CXO Talk, we're speaking about Data 2 00:00:07,160 --> 00:00:11,960 and a I. Our guests are Inderpal Bandari, 3 00:00:12,200 --> 00:00:18,520 the Global Chief Data Officer of IBM, and Anthony Scriffiniano, 4 00:00:18,560 --> 00:00:22,520 the former chief of Data scientists at Dunn and 5 00:00:22,520 --> 00:00:26,440 Bradstreet Inderpal, welcome to CXO Talk. 6 00:00:26,440 --> 00:00:27,680 It's great to see you. And please. 7 00:00:27,990 --> 00:00:31,950 Tell us about your work at IBM. I'm actually a four time Chief 8 00:00:31,950 --> 00:00:34,670 Data Officer. When I first became Chief Data 9 00:00:34,670 --> 00:00:38,070 Officer in 2006, there were just four of us globally. 10 00:00:38,630 --> 00:00:41,870 I was the first in healthcare and then the profession and the 11 00:00:41,870 --> 00:00:44,430 related professions like Chief Analytics officer, 12 00:00:44,430 --> 00:00:48,230 transformation officers, the century that took off and I 13 00:00:48,230 --> 00:00:51,990 happened to be fortunate enough to ride with it and I've done 14 00:00:51,990 --> 00:00:55,590 this job four times, IBM being the 4th and perhaps the most 15 00:00:55,670 --> 00:01:00,040 complicated. At IBM, my strategy, data, 16 00:01:00,040 --> 00:01:04,080 strategy has been to make IBM itself into an AI enterprise and 17 00:01:04,080 --> 00:01:06,960 then use that as a showcase for our clients and customers, 18 00:01:07,440 --> 00:01:09,440 because our clients look very much like us. 19 00:01:09,840 --> 00:01:12,520 So that's what I've been doing for the last seven years, 7 1/2 20 00:01:12,520 --> 00:01:15,800 years or so. And Anthony Scriffiniano, 21 00:01:15,800 --> 00:01:20,400 welcome back to CXO Talk. It's your good friend. 22 00:01:20,400 --> 00:01:24,240 It's great to see you and tell us about your work these days. 23 00:01:24,760 --> 00:01:27,600 Thank you very much, Michael. It's, it's great to see both of 24 00:01:27,600 --> 00:01:31,000 you. So as you mentioned, I was with 25 00:01:31,160 --> 00:01:33,840 Dunn and Bradstreet for quite a long time, over 20 years. 26 00:01:34,400 --> 00:01:39,120 Right now I'm doing a number of things and probably front foot 27 00:01:39,280 --> 00:01:42,680 is as a distinguished fellow with the Stimson Center, which 28 00:01:42,680 --> 00:01:46,000 is a think tank, think tank. I'll put in quotes because 29 00:01:46,000 --> 00:01:49,160 there's a lot of what I would call applied research or action 30 00:01:49,160 --> 00:01:52,040 research where they actually get involved in doing things, not 31 00:01:52,040 --> 00:01:54,120 just writing about them I've been involved with. 32 00:01:54,670 --> 00:01:58,070 Things that are called a I. The term's been around probably 33 00:01:58,070 --> 00:02:01,790 since the 50s but I've been involved with it in as it's 34 00:02:01,790 --> 00:02:06,510 become computational from its from its birth and and I know 35 00:02:06,510 --> 00:02:09,910 Interpol has as well lots of things going on in the world 36 00:02:09,910 --> 00:02:16,870 right now in terms of regulatory focus on a I as well as new 37 00:02:16,870 --> 00:02:20,910 types of a I becoming sort of the greatest new shiny object 38 00:02:20,910 --> 00:02:22,470 and everyone pays attention to them. 39 00:02:23,430 --> 00:02:26,990 And I I stand for the, the, the size behind it. 40 00:02:26,990 --> 00:02:29,790 What do you have to believe? What has to be true in order for 41 00:02:29,790 --> 00:02:32,950 you to do that thing that you think is so cool and and why is 42 00:02:32,950 --> 00:02:36,310 it better than what you're doing today and and what is the cost 43 00:02:36,310 --> 00:02:39,230 of it? So I try to ask those emperors 44 00:02:39,230 --> 00:02:42,710 new close kind of questions and that's the role I'm playing 45 00:02:42,710 --> 00:02:46,830 right now. So we're talking about data and 46 00:02:47,310 --> 00:02:49,640 AI. And I think where we need to 47 00:02:49,640 --> 00:02:56,080 start is when we talk about an AI data strategy, what actually 48 00:02:56,680 --> 00:03:00,240 is that Interpol, you wanna maybe take a crack at that to 49 00:03:00,240 --> 00:03:04,760 start? AI is only as good as the data 50 00:03:05,360 --> 00:03:09,840 that is used to train that AI, because AI has a training 51 00:03:09,840 --> 00:03:11,800 sequence and then an inference sequence. 52 00:03:12,400 --> 00:03:15,360 The training sequence has to do with seeing. 53 00:03:15,960 --> 00:03:21,320 All kinds of related data, so that it can actually then train 54 00:03:21,320 --> 00:03:25,000 itself to figure out what the right output is when it's shown 55 00:03:25,000 --> 00:03:27,240 an input that it may not have seen before. 56 00:03:27,840 --> 00:03:33,800 So if the data to begin with is flawed or low quality, the AI 57 00:03:33,800 --> 00:03:36,600 will not work effectively. It's the garbage in, garbage 58 00:03:36,600 --> 00:03:39,600 out, that kind of phenomenon that you would have. 59 00:03:40,000 --> 00:03:43,990 So they go hand in hand. And very often you think of 60 00:03:43,990 --> 00:03:48,110 people talking about AI, and if they haven't really looked at 61 00:03:48,110 --> 00:03:52,990 the data, but they embark on a data on an AI strategy that is 62 00:03:52,990 --> 00:03:55,590 going to be very high risk, it'll most likely fail because 63 00:03:55,590 --> 00:03:58,470 they'll have to go back and straighten out the data strategy 64 00:03:58,790 --> 00:04:00,950 first, just so that it's fit for purpose. 65 00:04:01,430 --> 00:04:04,590 Now when you say fit for purpose, what that means is. 66 00:04:05,480 --> 00:04:08,400 If you know what the business objective is that you're trying 67 00:04:08,400 --> 00:04:11,520 to serve, so it could be something quite narrow like a 68 00:04:11,520 --> 00:04:14,480 specific objective. It could be something like I 69 00:04:14,480 --> 00:04:18,160 want to understand what segments of my business. 70 00:04:18,810 --> 00:04:24,130 Should I try to try to expand to increase my top line then, in 71 00:04:24,130 --> 00:04:27,210 which case, if it's segments of business, you know, data about 72 00:04:27,250 --> 00:04:30,290 your clients, about your products, etc, those things 73 00:04:30,290 --> 00:04:32,210 become very important. You'd want to make sure that 74 00:04:32,210 --> 00:04:35,850 they are, that that data is of very high quality. 75 00:04:36,450 --> 00:04:40,050 And on the other hand, if it's something at a strategy level, 76 00:04:40,490 --> 00:04:43,250 which is what kind of what happened when I joined IBM, I 77 00:04:43,250 --> 00:04:46,130 mean, you know, IBM wanted to be a cloud and AI company. 78 00:04:46,990 --> 00:04:49,590 And to be a cloud and AI company, eventually we landed at 79 00:04:49,590 --> 00:04:52,910 the point that, well let's transform ourselves internally 80 00:04:52,910 --> 00:04:56,910 before we actually show this off to our clients and customers. 81 00:04:57,510 --> 00:05:01,870 That became a strategy that was enterprise wide and we realized 82 00:05:01,870 --> 00:05:05,830 that now while for instance not only do we have to make sure 83 00:05:05,830 --> 00:05:09,590 that our structured data is in order, but also our unstructured 84 00:05:09,590 --> 00:05:11,350 data because we are going to go after this. 85 00:05:11,910 --> 00:05:13,910 And transform ourselves into an AI company. 86 00:05:14,150 --> 00:05:17,110 So there are two, two aspects there that are relevant. 87 00:05:17,110 --> 00:05:20,310 One is at a strategy level when you're aligning, aligning to the 88 00:05:20,310 --> 00:05:23,790 business strategy or it could be more narrow to a specific 89 00:05:23,790 --> 00:05:28,230 business objective. Anthony, the challenge of 90 00:05:28,230 --> 00:05:34,150 aligning the data strategy to the business objectives is 91 00:05:34,150 --> 00:05:38,030 something that many organizations struggle with. 92 00:05:38,620 --> 00:05:42,780 What thoughts or advice do you have on making that work? 93 00:05:42,780 --> 00:05:45,180 You've seen so many different scenarios. 94 00:05:45,580 --> 00:05:48,540 You would really have to unpack what Interpol just said quite a 95 00:05:48,540 --> 00:05:52,380 bit to really get at the essence of it, and I did when I was 96 00:05:52,380 --> 00:05:55,340 listening to him, but he was using some terminology very 97 00:05:55,340 --> 00:05:59,330 carefully there. A lot of times organizations 98 00:05:59,330 --> 00:06:02,850 don't have one strategy, I mean make more money, you know grow, 99 00:06:02,850 --> 00:06:04,810 grow, you know fill in the blank, right. 100 00:06:05,930 --> 00:06:08,690 The the things that we learn in Business School, you can serve 101 00:06:08,690 --> 00:06:11,450 your shareholders, you can serve your customers, you can serve 102 00:06:11,450 --> 00:06:13,530 your employees. Kind of hard to do all of those 103 00:06:13,530 --> 00:06:18,530 things at the same time because very often optimizing for one is 104 00:06:18,530 --> 00:06:20,850 is less optimal for one of the others. 105 00:06:21,650 --> 00:06:24,770 So the strategy of which we speak. 106 00:06:25,300 --> 00:06:30,100 And we start to talk about AI has some very serious 107 00:06:30,100 --> 00:06:32,900 implications, these methods that we're talking about. 108 00:06:33,220 --> 00:06:38,020 And I should say that these days it's rare that only one method 109 00:06:38,020 --> 00:06:40,660 gets applied very often. There are many methods being 110 00:06:40,660 --> 00:06:44,780 applied simultaneously. There are some commonalities. 111 00:06:44,780 --> 00:06:49,180 So one of the commonalities is that the quality of the data has 112 00:06:49,180 --> 00:06:52,740 many dimensions, so truth. If if your a I is going to 113 00:06:52,740 --> 00:06:55,340 ingest data, it's going to probably presume it's all true. 114 00:06:55,900 --> 00:06:59,300 Well, all data is not necessarily simultaneously true. 115 00:06:59,300 --> 00:07:03,340 It may have been true at the time that it was created, but 116 00:07:03,340 --> 00:07:07,140 maybe not so much anymore at the time that it's curated. 117 00:07:07,140 --> 00:07:10,940 So how old is the data? How is it still true? 118 00:07:10,940 --> 00:07:14,260 How would you know that it's still true it before you consume 119 00:07:14,260 --> 00:07:16,780 it into an algorithm or an approach that presumes that? 120 00:07:17,180 --> 00:07:19,860 I love to say that. When we go to court, we swear to 121 00:07:19,860 --> 00:07:21,980 tell the truth, the whole truth, and nothing but the truth. 122 00:07:22,660 --> 00:07:25,500 That's because those are three different things, and those are 123 00:07:25,500 --> 00:07:27,900 three different ways to manipulate veracity or 124 00:07:27,900 --> 00:07:30,340 understanding. So when you get back to this 125 00:07:30,340 --> 00:07:33,980 concept of strategy, well, whose strategy? 126 00:07:33,980 --> 00:07:37,580 What part of the organization? Specifically, what objective? 127 00:07:37,580 --> 00:07:39,940 How would we know when we were successful? 128 00:07:40,340 --> 00:07:44,180 And asking those questions is very often a source of intention 129 00:07:44,180 --> 00:07:46,300 because the people in the room that all think they want the 130 00:07:46,300 --> 00:07:48,650 same thing. Realize they don't. 131 00:07:49,130 --> 00:07:51,850 And when you unpack it a little further, they realize that to 132 00:07:51,850 --> 00:07:54,890 get what the guy on the left wants, you have to get less of 133 00:07:54,890 --> 00:07:56,330 what the person on the right wants. 134 00:07:56,770 --> 00:07:59,450 And it's not pretty. So it's not really a technical 135 00:07:59,450 --> 00:08:02,410 problem as much as it is an alignment problem. 136 00:08:02,410 --> 00:08:06,010 And, you know, sort of getting everybody to agree on what they 137 00:08:06,010 --> 00:08:09,330 want so that we would know that this strategy of which we speak 138 00:08:09,730 --> 00:08:13,250 is actually what this a I of which we speak, is delivering 139 00:08:13,610 --> 00:08:17,010 really difficult. These roles like the Chief Data 140 00:08:17,010 --> 00:08:19,450 Officer, the Chief Transformation Officer, the 141 00:08:19,450 --> 00:08:23,410 reason these are CXO roles is because of what Anthony just 142 00:08:23,410 --> 00:08:25,490 said. You have to be part of that 143 00:08:25,490 --> 00:08:28,490 discussion. It's not so much like there is 144 00:08:28,490 --> 00:08:31,850 the concrete business objective. Sometimes you get into those 145 00:08:31,850 --> 00:08:35,890 situations where it's very clear cut, but more often than not 146 00:08:36,169 --> 00:08:39,929 it's a strategic discussion in terms of a understanding, 147 00:08:39,929 --> 00:08:42,809 clarifying, perhaps even adding to the business strategy. 148 00:08:43,220 --> 00:08:45,620 And then relating it back to what you're trying to do with 149 00:08:45,620 --> 00:08:49,540 data and AI, and unless you're in a position to have that kind 150 00:08:49,540 --> 00:08:52,300 of conversation and you also have the wherewithal to pull 151 00:08:52,300 --> 00:08:55,220 that off, you know that you won't be able, you won't really 152 00:08:55,220 --> 00:08:57,900 be successful. So that's why these are CXO 153 00:08:57,900 --> 00:09:01,900 roles, because it's really part of the negotiation that goes on 154 00:09:02,140 --> 00:09:07,420 to align the business strategy to the data or the AI strategy. 155 00:09:07,940 --> 00:09:10,540 Maybe I can just add a little bit to that, that the whole 156 00:09:10,620 --> 00:09:14,300 concept of being in the room. It's so important back in the 157 00:09:14,300 --> 00:09:18,220 day that the the goals and the objectives would come down from 158 00:09:18,220 --> 00:09:22,380 on high and that the folks with the the keyboards will just make 159 00:09:22,380 --> 00:09:26,900 it so you know, it doesn't work that way anymore and it can't 160 00:09:26,900 --> 00:09:30,300 really work that way anymore. And so it's so critical that the 161 00:09:30,300 --> 00:09:34,420 folks in the roles that Inderpal and I have had have a seat at 162 00:09:34,420 --> 00:09:38,660 the table understand what went into the ask and not just the 163 00:09:38,660 --> 00:09:44,650 ask very often what. Folks want and what they need 164 00:09:44,650 --> 00:09:48,130 are two very different things. And so without being arrogant 165 00:09:48,130 --> 00:09:51,890 about it, you you asked a lot of questions and you get at what 166 00:09:51,890 --> 00:09:54,090 they really needed in the 1st place, which is probably not 167 00:09:54,090 --> 00:09:55,530 what they started out asking for. 168 00:09:55,970 --> 00:10:00,370 You're talking about organizational alignment with 169 00:10:00,490 --> 00:10:05,050 business strategy, and at a high enough level, this is true for 170 00:10:05,050 --> 00:10:07,370 every business decision that needs to be made. 171 00:10:07,730 --> 00:10:11,170 And yet, when you hear people talking about a I. 172 00:10:11,540 --> 00:10:15,380 And data strategy, the conversation turns and very 173 00:10:15,380 --> 00:10:18,220 quickly to what kind of data do we need? 174 00:10:18,220 --> 00:10:20,900 Where do we get that data? What's the technology that we're 175 00:10:20,900 --> 00:10:23,940 going to use to aggregate and to manipulate that data? 176 00:10:23,940 --> 00:10:26,300 What kind of models are we using? 177 00:10:26,300 --> 00:10:30,860 And so now I'm confused because you're talking about one thing 178 00:10:30,860 --> 00:10:32,820 and I hear the entire world talking about something 179 00:10:32,820 --> 00:10:35,140 different. The world tends to focus on the 180 00:10:35,140 --> 00:10:38,380 hammers and the nails, and it tends to focus on. 181 00:10:38,990 --> 00:10:42,830 The the tools that are going to be used for the the, the 182 00:10:43,030 --> 00:10:46,390 purpose. If I come to your house and I 183 00:10:46,390 --> 00:10:50,230 say and if then you want to put an addition on your house and 184 00:10:50,270 --> 00:10:53,070 I've met with the architect and I understand your objectives, 185 00:10:53,710 --> 00:10:56,510 let's talk and someone else comes to your house and says I'm 186 00:10:56,510 --> 00:10:59,030 going to build you a beautiful addition and I'm going to use 187 00:10:59,030 --> 00:11:02,180 the hammer, right. You don't really care about the 188 00:11:02,180 --> 00:11:04,740 hammer, so of course the hammer is important. 189 00:11:04,740 --> 00:11:07,060 It's very important that we have the right data, the right tools, 190 00:11:07,060 --> 00:11:09,220 the right technology, the right people. 191 00:11:09,420 --> 00:11:11,820 So it's people, process, tools and mindset. 192 00:11:12,140 --> 00:11:15,620 All of those have to be aligned in order to get this right, but 193 00:11:15,620 --> 00:11:17,700 it starts with making sure you're focused on the right 194 00:11:17,700 --> 00:11:20,300 mission. Yes, that number changes that 195 00:11:20,300 --> 00:11:23,060 piece of, you know, aligning back to the business it's. 196 00:11:23,750 --> 00:11:27,550 So the way I would put it is, yes, no matter how promising the 197 00:11:27,550 --> 00:11:30,990 technology, no matter how dramatic the advance, etcetera, 198 00:11:31,830 --> 00:11:36,150 it doesn't let the organization off the hook for coming up with 199 00:11:36,150 --> 00:11:39,950 a sound business strategy and then of aligning these elements 200 00:11:40,190 --> 00:11:43,230 to that business strategy. That's still going to be very 201 00:11:43,230 --> 00:11:47,910 much needed, in fact maybe even more so than before as you try 202 00:11:47,910 --> 00:11:50,270 to go after these new approaches and methods. 203 00:11:50,800 --> 00:11:55,120 Be sure to subscribe to our YouTube channel and hit the 204 00:11:55,120 --> 00:11:59,960 subscribe button at the bottom of our web page and you can 205 00:11:59,960 --> 00:12:02,720 subscribe to our newsletter and we'll tell you and notify you 206 00:12:03,040 --> 00:12:05,640 about our excellent upcoming shows and guests. 207 00:12:05,640 --> 00:12:08,880 We have lots of them. So would you say that the 208 00:12:09,040 --> 00:12:15,640 business side is more difficult or harder to achieve than the 209 00:12:15,640 --> 00:12:19,360 data and technology foundations in your experience? 210 00:12:19,820 --> 00:12:23,300 I would say that if you take something new, so you probably 211 00:12:23,300 --> 00:12:26,820 want to draw the distinction between a mature technology and 212 00:12:26,820 --> 00:12:29,940 a technology that's more recent or Macent or emerging. 213 00:12:30,660 --> 00:12:34,620 And if you take something new like that like the latter then 214 00:12:34,780 --> 00:12:37,260 it's you know there. There is a tremendous amount of 215 00:12:37,260 --> 00:12:40,700 complexity on the technology side as well. 216 00:12:41,300 --> 00:12:46,340 So early on when getting into this game, you know when you we 217 00:12:46,340 --> 00:12:52,380 were working on on. AI for instance, at IBM, it 218 00:12:52,380 --> 00:12:55,500 became very clear as we went forward that there were four 219 00:12:55,500 --> 00:12:59,180 elements that had to move in lockstep, data, technology, 220 00:12:59,500 --> 00:13:03,980 workflow and culture. And those four kind of had to 221 00:13:03,980 --> 00:13:07,300 move at the same time. Otherwise the adoption was not 222 00:13:07,300 --> 00:13:11,300 going to be not going to be effective and. 223 00:13:11,830 --> 00:13:15,230 The technology piece for an emerging technology, so at that 224 00:13:15,230 --> 00:13:20,430 time, you know the cloud was emerging, there was a lot of AI 225 00:13:20,430 --> 00:13:23,910 techniques that were emerging, the deep learning stuff with, 226 00:13:24,030 --> 00:13:26,270 you know, GPUs and things like that. 227 00:13:26,590 --> 00:13:28,430 You have to make all that stuff work together. 228 00:13:28,430 --> 00:13:31,270 So there is a significant complexity in the technology 229 00:13:31,270 --> 00:13:34,470 piece, but there is also a significant complexity in the 230 00:13:34,470 --> 00:13:37,110 data piece and the workflow workflow piece and then 231 00:13:37,110 --> 00:13:40,070 eventually in the culture piece of the organization. 232 00:13:40,510 --> 00:13:43,190 The the stuff that we were talking about in terms of the 233 00:13:43,190 --> 00:13:47,550 negotiation, working with the C-Suite, you know, there's a lot 234 00:13:47,550 --> 00:13:50,750 of the cultural aspect that goes into, there are many 235 00:13:50,750 --> 00:13:54,710 organizations one could go into and you would essentially not, 236 00:13:54,910 --> 00:13:57,510 Anthony said. They would still want to give 237 00:13:57,510 --> 00:14:00,790 you a set of objectives and say here, go off implement this. 238 00:14:00,790 --> 00:14:02,910 We really don't want to hear from you about anything else. 239 00:14:03,350 --> 00:14:05,830 These are your marching orders. Go off and implement this. 240 00:14:06,300 --> 00:14:09,340 Well, that's the wrong approach when you're trying to bring in 241 00:14:09,340 --> 00:14:12,020 an emerging technology and use it to impact the business. 242 00:14:12,460 --> 00:14:16,380 Anthony, we have a question exactly on this topic from 243 00:14:16,380 --> 00:14:19,660 Twitter from Arsalan Khan and maybe you can share your 244 00:14:19,660 --> 00:14:22,660 thoughts on this. He says we when we talk about 245 00:14:22,660 --> 00:14:26,140 alignment, there's business strategy, enterprise business 246 00:14:26,260 --> 00:14:32,820 architecture, change management culture and now data strategy. 247 00:14:33,920 --> 00:14:38,000 All right, Anthony, so what's your prescription than to make 248 00:14:38,000 --> 00:14:42,920 all these layers work together and align sounds almost 249 00:14:42,920 --> 00:14:45,880 impossible? Almost impossible is a synonym 250 00:14:45,880 --> 00:14:48,240 for possible. So if you said it was 251 00:14:48,240 --> 00:14:51,760 impossible, you know, now we have to, we have to talk, right? 252 00:14:53,160 --> 00:14:56,400 I think that first of all, thank you for the question. 253 00:14:56,480 --> 00:15:00,120 From someone who knows that I ask a good question, I would say 254 00:15:00,960 --> 00:15:04,690 it's really important. That you start with the question 255 00:15:04,690 --> 00:15:06,730 with the objective. Everybody wants to jump to the 256 00:15:06,730 --> 00:15:08,850 technology. They want to jump to the the 257 00:15:08,890 --> 00:15:11,290 date of the deal. The the thing that we're going 258 00:15:11,290 --> 00:15:16,010 to the the, the, you know that there's there's two factions in 259 00:15:16,010 --> 00:15:18,170 the room. The one faction is focused on 260 00:15:18,850 --> 00:15:21,650 the the revenue, the growth, you know what's going to happen to 261 00:15:21,650 --> 00:15:25,250 the organization and the other faction is focused on, all 262 00:15:25,250 --> 00:15:28,730 right, let's get going, Let's start, you know, doing stuff, 263 00:15:28,730 --> 00:15:29,930 let's start cooking in the kitchen. 264 00:15:31,200 --> 00:15:33,720 I'm usually the 1:00 somewhere in the middle of those two 265 00:15:33,720 --> 00:15:35,880 saying let's make sure we're answering the right question 266 00:15:35,880 --> 00:15:38,800 here. And I, you know, I'm not slowing 267 00:15:38,800 --> 00:15:40,960 you down. I'm actually making sure we get 268 00:15:40,960 --> 00:15:44,280 done in a way that we don't fall over the finish line. 269 00:15:44,680 --> 00:15:48,280 So it is very difficult to get all those factions in the same 270 00:15:48,280 --> 00:15:50,440 place. Probably the most important 271 00:15:50,440 --> 00:15:53,760 thing you have to do is be able to listen to each other and not 272 00:15:53,760 --> 00:15:56,440 start immediately talking about hammers and nails or immediately 273 00:15:56,440 --> 00:15:59,040 talking start talking about what color we're going to paint the 274 00:15:59,040 --> 00:16:02,280 finished product, right. But you know, somewhere in the 275 00:16:02,280 --> 00:16:04,480 middle is, you know, why are we doing this? 276 00:16:04,480 --> 00:16:06,520 What are we not doing while we're doing this? 277 00:16:07,200 --> 00:16:09,440 Do we know there's a big difference between can we do it 278 00:16:09,440 --> 00:16:11,720 and should we do it. So what are we giving up while 279 00:16:11,720 --> 00:16:13,920 we do it? What about compliance, What 280 00:16:13,920 --> 00:16:16,120 about regulatory, what about making? 281 00:16:16,120 --> 00:16:18,960 How do we know that the data that we have is the right data 282 00:16:18,960 --> 00:16:22,680 to make the decision you want Just because you believe it and 283 00:16:22,680 --> 00:16:25,880 you have your confirmation bias and you found one or two pieces 284 00:16:25,880 --> 00:16:28,080 of data that support your hypothesis doesn't make you 285 00:16:28,080 --> 00:16:29,760 right. So we have to ask these 286 00:16:29,760 --> 00:16:33,600 difficult questions and there's a very fine line between being 287 00:16:33,600 --> 00:16:35,680 right and being dead. So you have to be able to ask 288 00:16:35,680 --> 00:16:38,880 them in a way that doesn't annoy, it can annoy them a 289 00:16:38,880 --> 00:16:41,400 little bit, but you have to annoy them just to the point 290 00:16:41,400 --> 00:16:44,040 where they don't kick you out of the room and and keep asking 291 00:16:44,040 --> 00:16:46,800 those, you know, help me understand kind of questions 292 00:16:46,800 --> 00:16:50,320 until we get to a shared understanding of what it is 293 00:16:50,320 --> 00:16:53,240 we're trying to achieve and the opportunity cost of all the 294 00:16:53,240 --> 00:16:54,320 other things that we're not doing. 295 00:16:55,400 --> 00:16:57,920 Interpol. But you're a technologist? 296 00:16:58,310 --> 00:17:05,069 So if this is strictly then a business issue of organizational 297 00:17:05,069 --> 00:17:11,550 alignment, why do technologists play such an important role in 298 00:17:11,550 --> 00:17:14,390 this discussion of such a foundational fundamental? 299 00:17:14,390 --> 00:17:17,230 Role. I think the best way to think of 300 00:17:17,230 --> 00:17:20,910 my role of people in similar situations is that of a 301 00:17:20,910 --> 00:17:24,790 changing. The catalyst for the change is 302 00:17:24,790 --> 00:17:28,750 the technology. The change has to be affected in 303 00:17:28,750 --> 00:17:30,790 the organization and in the business. 304 00:17:31,190 --> 00:17:36,270 So you have to be able to bridge those two to be able to do this 305 00:17:36,270 --> 00:17:39,270 successfully. So you know it's a 306 00:17:39,270 --> 00:17:42,990 transformation and the transformation typically has 307 00:17:42,990 --> 00:17:46,070 those elements that I talked about for what we do, what I do 308 00:17:46,590 --> 00:17:50,950 data, technology, workflow and culture and I'll give you one 309 00:17:50,950 --> 00:17:55,280 other thing, I mean it's. There is there is a lot to be 310 00:17:55,280 --> 00:17:58,480 done in terms of changing the culture of an organization when 311 00:17:58,480 --> 00:18:00,680 you try to bring bring about this change. 312 00:18:01,080 --> 00:18:06,040 What we what we saw at IBM when we pushed forward with our data 313 00:18:06,040 --> 00:18:11,960 and AI strategy was that the adoption of the platform was 314 00:18:11,960 --> 00:18:16,080 triggered far more by the bottom up measures that we put in 315 00:18:16,080 --> 00:18:18,280 place. So we actually had a team that 316 00:18:18,280 --> 00:18:20,840 was empowered to engage with other teams. 317 00:18:21,570 --> 00:18:24,610 That were working in the business, you know doing 318 00:18:25,370 --> 00:18:31,130 workflows, go to cash procurement, you know things 319 00:18:31,130 --> 00:18:34,570 like that supply chain. And so we have an empowered team 320 00:18:34,570 --> 00:18:38,410 on the technology side which was didn't really need to come back 321 00:18:38,410 --> 00:18:41,570 for direction or instruction but if they found a like minded team 322 00:18:41,570 --> 00:18:44,970 they could go ahead and and move forward with the transformation. 323 00:18:45,450 --> 00:18:49,130 We found that 85% of the adoption actually came from that 324 00:18:49,130 --> 00:18:51,720 part. As opposed to the top down path 325 00:18:51,880 --> 00:18:54,480 and so forth. So it really is all about how 326 00:18:54,480 --> 00:18:57,920 you effect the change. But obviously if the catalyst is 327 00:18:57,920 --> 00:19:01,560 the technology then you've got to be able to walk that walk as 328 00:19:01,560 --> 00:19:04,200 well. So but you you can't discount 329 00:19:04,200 --> 00:19:06,680 the other side of it, you have to really be the bridge. 330 00:19:07,160 --> 00:19:10,280 Like I smiled when you called into call a technologist and he 331 00:19:11,200 --> 00:19:14,320 very diplomatically didn't didn't respond. 332 00:19:14,920 --> 00:19:18,080 I think you can tell by that answer that you have to be much 333 00:19:18,120 --> 00:19:20,750 more than. Just an expert in the technology 334 00:19:20,750 --> 00:19:25,470 to get what he just said right in large organizations there, 335 00:19:25,470 --> 00:19:31,870 what's happening right now is a massive federation of data and a 336 00:19:31,870 --> 00:19:34,790 I capability. It's not like you go to the room 337 00:19:34,790 --> 00:19:37,910 where the people that know how to do that live and ask them to 338 00:19:37,910 --> 00:19:40,630 do it for you. Almost anybody can get these 339 00:19:41,390 --> 00:19:44,110 capabilities on their desktop. It doesn't mean that's the right 340 00:19:44,110 --> 00:19:46,950 place to do it, but they can start doing it there and 341 00:19:46,950 --> 00:19:48,550 everyone feels like they're an expert. 342 00:19:49,020 --> 00:19:54,100 Just like when we all first got, you know, I'm trying not to name 343 00:19:54,100 --> 00:19:56,940 a product, but I think I can say Harvard graphics or you know, 344 00:19:56,940 --> 00:19:59,780 like in the days even before PowerPoint where all of a sudden 345 00:19:59,780 --> 00:20:02,900 we could all, you know, lay things out on the screen, we all 346 00:20:02,900 --> 00:20:05,580 thought we were experts in design and layout and font 347 00:20:05,580 --> 00:20:07,980 selection and and all of that. And of course we weren't. 348 00:20:07,980 --> 00:20:10,540 And and there's an old joke where the punchline is death by 349 00:20:10,540 --> 00:20:13,860 PowerPoint. We we all know versions of that 350 00:20:13,860 --> 00:20:15,380 joke. And I'm not picking up 351 00:20:15,380 --> 00:20:18,840 PowerPoint. Federating A capability like 352 00:20:18,840 --> 00:20:23,200 that across an organization or across the world comes with some 353 00:20:23,200 --> 00:20:27,400 risk that those who really honor practitioners who know. 354 00:20:27,640 --> 00:20:29,600 What? The differences between what you 355 00:20:29,600 --> 00:20:34,000 can do and what you should do, who understand the implications 356 00:20:34,000 --> 00:20:37,680 of going down a certain path and the difficulty of changing 357 00:20:37,680 --> 00:20:39,520 course once you get too far down the path. 358 00:20:40,120 --> 00:20:42,560 They have to be able to hear what's going on so for. 359 00:20:42,980 --> 00:20:46,940 When Interpol is describing to work, well 85% of the time does 360 00:20:46,940 --> 00:20:50,020 require an organization that actually talks to each other or 361 00:20:50,140 --> 00:20:53,500 or at least talks up to people who talk to each other up and 362 00:20:53,500 --> 00:20:55,900 down. But you know that's not always 363 00:20:55,900 --> 00:20:57,660 the case. So you can't just throw 364 00:20:57,660 --> 00:20:59,620 everything out in the middle of the floor and say here you go, 365 00:20:59,860 --> 00:21:01,740 everybody play with this and do whatever you want. 366 00:21:01,740 --> 00:21:04,260 That will not work. That will end in tears. 367 00:21:04,620 --> 00:21:09,060 So you have governance, you have focus on these foundational 368 00:21:09,820 --> 00:21:14,500 pieces. What about the interface between 369 00:21:14,500 --> 00:21:18,180 the technology and what you're describing the whole world and 370 00:21:18,260 --> 00:21:22,900 and organizations by and large tend to focus on that technology 371 00:21:23,340 --> 00:21:26,100 piece. And so can you now maybe talk a 372 00:21:26,100 --> 00:21:31,420 little about technology management as it relates to what 373 00:21:31,420 --> 00:21:35,340 you're just describing and also selecting the right kinds of 374 00:21:35,340 --> 00:21:39,610 technologies and especially? Selecting the right kinds of 375 00:21:39,770 --> 00:21:43,050 data to match with the problems that you're trying to ultimately 376 00:21:43,050 --> 00:21:45,530 address. And I would add time that it's 377 00:21:45,530 --> 00:21:47,570 still relevant. I think I have a good example 378 00:21:47,570 --> 00:21:50,770 for you. When when the pandemic broke 379 00:21:50,770 --> 00:21:54,930 out, I don't think anybody was really expecting that. 380 00:21:55,490 --> 00:21:59,610 All of a sudden organizations shifted to almost exclusively 381 00:21:59,610 --> 00:22:03,050 working from home. There's laws about what data you 382 00:22:03,050 --> 00:22:05,930 can access from home and what data you can access at your 383 00:22:05,930 --> 00:22:08,020 desk. You have a different firewall 384 00:22:08,020 --> 00:22:09,860 when you're working in the office than you do when you're 385 00:22:09,860 --> 00:22:13,100 working at home. You've got developers that used 386 00:22:13,100 --> 00:22:16,300 to be Co located that are not Co located anymore. 387 00:22:17,300 --> 00:22:21,580 Organizations had to absorb all of that change while still 388 00:22:21,580 --> 00:22:25,340 trying to serve their customers, and in some cases failure to do 389 00:22:25,340 --> 00:22:26,820 so could have been life and death. 390 00:22:27,380 --> 00:22:31,260 So you know there's an urgency about this as well. 391 00:22:31,260 --> 00:22:34,980 You can't take forever to do it and you have to have good 392 00:22:34,980 --> 00:22:38,550 discipline in place. So that when the unexpected 393 00:22:38,550 --> 00:22:40,990 happens in the middle of the other unexpected that was 394 00:22:40,990 --> 00:22:46,070 already happening, you have the resiliency survive that and come 395 00:22:46,150 --> 00:22:50,590 out of that stronger. I'm not going to suggest, 396 00:22:50,590 --> 00:22:53,830 although I could that I BM is one of those organizations, but 397 00:22:54,110 --> 00:22:58,510 you know mature organizations that get it and do that. 398 00:22:58,510 --> 00:23:01,750 We saw a lot of organizations that weren't so mature not 399 00:23:01,750 --> 00:23:04,270 getting it in the middle of all that disruption. 400 00:23:04,270 --> 00:23:06,390 So it's it's just a very big question you're asking. 401 00:23:06,760 --> 00:23:09,600 The example of the pandemic actually was particularly 402 00:23:09,600 --> 00:23:13,360 instructed, I think it goes to your data AI questions of the 403 00:23:14,240 --> 00:23:16,000 earlier in the in the segment as well. 404 00:23:16,640 --> 00:23:21,720 So when the pandemic hit in terms of being able to run your 405 00:23:21,720 --> 00:23:25,400 business, for instance, make financial forecasts, make 406 00:23:25,400 --> 00:23:28,160 forecasts about your supply chain, about your procurement 407 00:23:28,160 --> 00:23:32,960 abilities etcetera, all the models that were in play were 408 00:23:32,960 --> 00:23:36,480 essentially useless. Because we have now embarked on 409 00:23:36,480 --> 00:23:38,960 a situation that was completely new. 410 00:23:40,080 --> 00:23:44,120 And so no matter what technology we had in there from an AI 411 00:23:44,120 --> 00:23:47,240 standpoint or a model standpoint, it had been trained 412 00:23:47,240 --> 00:23:50,680 in a completely different world. And that's Anthony, was 413 00:23:50,680 --> 00:23:52,880 Anthony's point, right? I mean, it may not be true now. 414 00:23:52,960 --> 00:23:56,720 In fact, it wasn't true. What was true though was if we 415 00:23:56,720 --> 00:24:01,960 were able to get the data, accurate data, pristine data. 416 00:24:02,570 --> 00:24:05,050 Into the hands of the people who were running those different 417 00:24:05,050 --> 00:24:09,530 departments along with an overlay of what was actually 418 00:24:09,530 --> 00:24:12,770 happening in the pandemic, you know, where COVID-19 was 419 00:24:12,770 --> 00:24:15,850 breaking, what were the incident reports in different areas. 420 00:24:15,850 --> 00:24:18,050 So if you could like geographically, then overlay 421 00:24:18,050 --> 00:24:20,970 that on what these guys were working on, whether it be 422 00:24:20,970 --> 00:24:24,890 financial forecasts or sales with, you know, they expected to 423 00:24:24,890 --> 00:24:28,490 close or procurement sites that were endangered, things like 424 00:24:28,490 --> 00:24:30,370 that. They could make something out of 425 00:24:30,370 --> 00:24:33,900 it and move forward with it. And then so that that's I think 426 00:24:33,900 --> 00:24:38,060 also an instructive example of the relationship between data 427 00:24:38,060 --> 00:24:42,380 and AI and how that plays out as it as things really unfold that 428 00:24:42,580 --> 00:24:44,060 you know that are truly unexpected. 429 00:24:44,500 --> 00:24:47,780 Let me draw first blood on saying something super nerdy. 430 00:24:48,140 --> 00:24:51,460 There's a concept. I call it decision elasticity. 431 00:24:51,460 --> 00:24:53,340 I kind of stole it from economics. 432 00:24:53,340 --> 00:24:56,620 But how wrong can you be and still make the same decision 433 00:24:56,820 --> 00:24:58,940 effectively? So you don't have to be perfect 434 00:24:58,940 --> 00:25:01,350 to make a decision. And Interpol is talking about 435 00:25:01,350 --> 00:25:03,470 training. There's an implication there 436 00:25:03,470 --> 00:25:07,110 that you have longitudinal data data from the past that you can 437 00:25:07,110 --> 00:25:11,990 project into a near term future that looks reasonably similar 438 00:25:12,270 --> 00:25:15,070 and you can measure the elasticity of your decisions. 439 00:25:15,070 --> 00:25:17,230 How wrong are they? And then if they start getting 440 00:25:17,230 --> 00:25:21,190 wronger and wronger to coin a term, then you can stop and and 441 00:25:21,190 --> 00:25:24,190 reexamine those methods. The problem is when you have 442 00:25:24,190 --> 00:25:27,670 something completely disruptive, there is no data and the most 443 00:25:27,670 --> 00:25:29,870 dangerous situation you can find yourself in. 444 00:25:30,220 --> 00:25:32,620 Is when the world is changing faster than the data that 445 00:25:32,620 --> 00:25:35,380 describes it. That's exactly where we were at 446 00:25:35,380 --> 00:25:37,260 that moment. You can't just throw your hands 447 00:25:37,260 --> 00:25:39,820 up and say, well wait, when you have five years worth of data, 448 00:25:39,820 --> 00:25:42,460 come back and I'll, I'll retrain everything and and we'll be good 449 00:25:42,460 --> 00:25:44,340 to go. You have to have methods in 450 00:25:44,340 --> 00:25:49,220 place that are effective in a situation where and that's what 451 00:25:49,220 --> 00:25:53,580 this environment taught us, that you can't just rely on one type 452 00:25:53,580 --> 00:25:57,500 of learning, one type of projection into the future. 453 00:25:58,080 --> 00:26:00,680 At that time, I was very involved with watching bad guys 454 00:26:00,680 --> 00:26:03,560 do bad things well. When there's disruption, the 455 00:26:03,560 --> 00:26:06,000 best bad guys, especially if they think they're being 456 00:26:06,000 --> 00:26:07,680 watched, they change what they're doing. 457 00:26:08,000 --> 00:26:10,400 If you model based on what they were doing, you're modeling how 458 00:26:10,400 --> 00:26:13,280 the best ones are no longer behaving out of a dangerous 459 00:26:13,280 --> 00:26:15,320 thing to do, right? But we know this. 460 00:26:15,680 --> 00:26:20,600 And so the the flip side of that coin is if you know that the 461 00:26:20,600 --> 00:26:22,960 environment changed such that the bad guys are going to 462 00:26:22,960 --> 00:26:24,760 probably try to take advantage of it. 463 00:26:25,270 --> 00:26:28,390 That many of them are probably going to do that unarthfully. 464 00:26:28,910 --> 00:26:32,190 And so you may be more easily able to see them as they run. 465 00:26:32,470 --> 00:26:35,350 You know, you turn on the light and the the, you know, the 466 00:26:35,510 --> 00:26:38,430 little creatures run away, you can see that. 467 00:26:38,430 --> 00:26:41,430 And so there might be an opportunity there along with 468 00:26:41,430 --> 00:26:43,990 that risk. So it's it's sometimes these 469 00:26:44,190 --> 00:26:48,750 these situations are, I would say almost never are they all 470 00:26:48,750 --> 00:26:51,310 bad or all good. There's always something in it 471 00:26:51,310 --> 00:26:53,920 that can teach you. There's always something in it 472 00:26:53,920 --> 00:26:57,240 that can make what you're doing better if you have enough time 473 00:26:57,400 --> 00:27:01,280 to breathe and observe what's going on and use the energy in 474 00:27:01,280 --> 00:27:03,800 the best possible way. It doesn't mean the bad thing 475 00:27:03,800 --> 00:27:07,400 will stop happening, but it may mean that you emerge from it in 476 00:27:07,400 --> 00:27:09,920 a better way because you you took that time to be more 477 00:27:10,280 --> 00:27:13,280 thoughtful about it. So we have a question from 478 00:27:13,280 --> 00:27:16,760 Twitter. Elizabeth Shaw says the issues 479 00:27:16,760 --> 00:27:21,280 you're describing are true of any business or technology 480 00:27:21,280 --> 00:27:24,880 transformation. Are there particular points 481 00:27:24,880 --> 00:27:30,920 issues that are more problematic for AI enabled initiatives? 482 00:27:30,920 --> 00:27:32,880 Can you kind of drill down into that? 483 00:27:33,160 --> 00:27:37,120 If you look at the advent of AI, the progression of AI, it's 484 00:27:37,120 --> 00:27:42,200 moved very, very quickly in the consumer space, but not so fast 485 00:27:42,200 --> 00:27:45,160 on the business space. And that's because in the 486 00:27:45,160 --> 00:27:49,040 business context, people don't trust AI. 487 00:27:49,710 --> 00:27:52,150 And they don't trust AI for multiple reasons. 488 00:27:52,190 --> 00:27:55,310 I mean, there's the, we talked about some of the issues about 489 00:27:55,310 --> 00:27:58,350 the data. So the robustness of the data, 490 00:27:58,350 --> 00:28:01,030 the quality of the data, the currency of the data. 491 00:28:01,670 --> 00:28:04,950 Then you also get into issues that have to do with the 492 00:28:04,950 --> 00:28:08,190 fairness of the algorithms. You know that the results they 493 00:28:08,190 --> 00:28:11,630 produce are going to treat people fairly if they pertain 494 00:28:11,630 --> 00:28:15,750 to, if the, you know, the data pertains to people, you have the 495 00:28:15,750 --> 00:28:22,020 issue of privacy being invaded. In terms of the algorithms 496 00:28:22,020 --> 00:28:26,020 discovering something new, you know there's this famous example 497 00:28:26,020 --> 00:28:31,380 of or infamous example of retail retailer, large retailer 498 00:28:32,020 --> 00:28:36,380 actually looking at the shopping shopping data, shopping patterns 499 00:28:36,380 --> 00:28:39,900 and shopping data and then inferring that. 500 00:28:40,740 --> 00:28:44,380 That this person is is pregnant and actually mailing their home 501 00:28:44,420 --> 00:28:49,060 and it turns out to be a young lady and you know, it was, it 502 00:28:49,100 --> 00:28:51,420 was it was really a complete invasion of her privacy. 503 00:28:51,700 --> 00:28:54,940 So those aspects come in. Then there are the issues around 504 00:28:54,940 --> 00:28:57,860 the job displacement and things of that nature. 505 00:28:57,860 --> 00:29:01,980 You know, if you're applying AI in the enterprise, there are two 506 00:29:01,980 --> 00:29:04,180 flavors of it. There's the automation flavor 507 00:29:04,540 --> 00:29:08,020 which has to do with when things are kind of straightforward. 508 00:29:08,410 --> 00:29:12,250 And you go from one step to the other and you know what those 509 00:29:12,250 --> 00:29:14,610 steps are and you can automate all that. 510 00:29:14,930 --> 00:29:17,170 So there's job displacement associated with that. 511 00:29:17,410 --> 00:29:20,210 But even on the decision making side, where the AI is actually 512 00:29:20,210 --> 00:29:23,810 helping make a decision, there's a decision maker in play and 513 00:29:23,810 --> 00:29:25,890 they have to trust it. They have to say, well, this is. 514 00:29:26,480 --> 00:29:28,640 Going to, you know, this won't displace me. 515 00:29:28,640 --> 00:29:33,920 So and extending that further, the executives as you put AI, we 516 00:29:33,920 --> 00:29:37,320 kind of know by now that AI has to be infused into the major 517 00:29:37,320 --> 00:29:40,560 workflows of the business, things like procurement, supply 518 00:29:40,560 --> 00:29:43,240 chain, etc. That's the kind of IP that 519 00:29:43,440 --> 00:29:46,880 doesn't get published in papers or patented or anything. 520 00:29:46,880 --> 00:29:48,640 Those are the trade secrets of a company. 521 00:29:49,080 --> 00:29:52,280 So they have to be able to trust whoever the vendor is of this 522 00:29:52,280 --> 00:29:54,600 software that this is not something that's going to 523 00:29:54,600 --> 00:29:57,980 disintermediate. Furthermore, the decision maker 524 00:29:57,980 --> 00:30:00,140 that's working with the system has to understand it. 525 00:30:00,740 --> 00:30:03,980 You know, years ago we I did this computer program called 526 00:30:03,980 --> 00:30:08,060 Advanced Scout that ended up being used by every coach in the 527 00:30:08,060 --> 00:30:10,540 NBA. And I remember the first time it 528 00:30:10,540 --> 00:30:14,380 had a counterintuitive finding. It basically asked the coach to 529 00:30:14,380 --> 00:30:18,060 play 2 backup players in a playoff game that they were on 530 00:30:18,060 --> 00:30:21,600 the verge of elimination. And he, you know, he was very 531 00:30:21,600 --> 00:30:25,120 concerned about that because he felt if I make, if I do this and 532 00:30:25,120 --> 00:30:28,200 I lose, I'm gonna lose my job and reputation as well in 533 00:30:28,200 --> 00:30:31,320 addition to the series. And we kind of solved that 534 00:30:31,320 --> 00:30:34,560 problem by letting him see the video clips of when those two 535 00:30:34,560 --> 00:30:37,960 players were on court. But that's the explanation 536 00:30:37,960 --> 00:30:40,000 piece, right? So if you tell a doctor, 537 00:30:40,000 --> 00:30:43,020 amputate the left leg. They're going to have all kinds 538 00:30:43,020 --> 00:30:46,260 of questions, OK, why amputate? What other options were 539 00:30:46,260 --> 00:30:48,660 considered? Why is amputation the right one 540 00:30:48,660 --> 00:30:49,940 for this patient? Etc. 541 00:30:50,220 --> 00:30:53,660 So explanation is another big part of it, and the AI systems 542 00:30:53,660 --> 00:30:55,300 today don't do a good job of all that. 543 00:30:55,300 --> 00:31:01,380 So those are the special aspects of AI and trust that come into 544 00:31:01,380 --> 00:31:03,660 play. I think that was a fantastic 545 00:31:03,660 --> 00:31:08,820 list and I won't vain to add to it, but I will suggest another 546 00:31:08,820 --> 00:31:11,020 dimension to it. So great question. 547 00:31:11,020 --> 00:31:13,780 Like how how are the A, I issues? 548 00:31:13,780 --> 00:31:15,260 What's special about the A I issues? 549 00:31:15,260 --> 00:31:19,820 I would say another one is that you have the opportunity to fail 550 00:31:19,820 --> 00:31:24,300 faster and at larger scale. There's a tendency once these 551 00:31:24,300 --> 00:31:26,900 sorts of systems are implemented, someone says, well, 552 00:31:26,900 --> 00:31:32,260 it's 99% accurate, it's 92% accurate, it's 87% accurate and 553 00:31:32,260 --> 00:31:35,060 you assume that means that 87% of the time the prediction will 554 00:31:35,060 --> 00:31:37,100 be right. Well, no, that's based on the 555 00:31:37,100 --> 00:31:39,230 past. In the future, right? 556 00:31:39,750 --> 00:31:47,270 Very rarely do we measure fast enough to stop every conceivable 557 00:31:47,270 --> 00:31:51,350 bad thing from happening. Interpol hinted at something 558 00:31:51,710 --> 00:31:59,150 which is an observer effect that people, when told what to do by 559 00:31:59,470 --> 00:32:01,750 a quote UN quote machine, will sometimes. 560 00:32:02,960 --> 00:32:06,200 Think they know better or not want to be told what to do by a 561 00:32:06,200 --> 00:32:10,000 machine and do something different just because a machine 562 00:32:10,000 --> 00:32:12,480 told them to do it? To prove that they can do 563 00:32:12,480 --> 00:32:15,520 something and not necessarily thinking it out loud. 564 00:32:15,520 --> 00:32:19,280 Like that Question I get asked a lot is, you know, what about 565 00:32:19,280 --> 00:32:23,400 someday when will people be reporting to robots or robotic 566 00:32:24,080 --> 00:32:26,640 bosses of some sort And you say, oh, of course not. 567 00:32:26,640 --> 00:32:28,840 I would never do that. And then the GPS tells you to 568 00:32:28,840 --> 00:32:32,210 turn left or right and you do. And Outlook tells you to go to a 569 00:32:32,210 --> 00:32:34,450 meeting and you go. We're already taking a lot of 570 00:32:34,450 --> 00:32:39,050 direction from automation. I won't call it a I necessarily, 571 00:32:39,050 --> 00:32:44,330 but from automation and the the the human fact of what we do as 572 00:32:44,330 --> 00:32:48,850 human beings to accept or reject that device is essential to get 573 00:32:48,850 --> 00:32:52,290 at trustworthy a I to get at making sure that we don't 574 00:32:52,290 --> 00:32:55,610 marginalize others that are already marginalized more 575 00:32:55,850 --> 00:32:57,930 because they don't have access to these technologies. 576 00:32:58,550 --> 00:33:02,630 This concept of good and and not good is is a very, it depends on 577 00:33:02,630 --> 00:33:04,910 where you're sitting sometimes whether it's good or not good. 578 00:33:05,630 --> 00:33:10,750 There are certainly lots of volumes, books, committees 579 00:33:11,710 --> 00:33:14,270 focused on trustworthy, A, I and explain ability. 580 00:33:14,270 --> 00:33:19,790 There's legislation as we speak being considered that will hold 581 00:33:19,790 --> 00:33:22,430 the feet to the fire of anyone who is implementing anything 582 00:33:22,430 --> 00:33:26,150 called a I. So you know to say that it's not 583 00:33:26,150 --> 00:33:32,930 being adopted by business. The adoption is lower, I think 584 00:33:32,930 --> 00:33:35,730 in some ways because of some of these human factors. 585 00:33:36,250 --> 00:33:42,090 It's not a lack of technology, it's a reticence to just push 586 00:33:42,130 --> 00:33:45,370 that button so quickly. And you know, technology will 587 00:33:45,370 --> 00:33:48,090 always outpace regulation. So you have to be careful where 588 00:33:48,090 --> 00:33:51,210 you could find yourself in a in a world of hurt where now 589 00:33:51,210 --> 00:33:53,890 they're coming after you because you use that technology that 590 00:33:53,890 --> 00:33:56,770 made a better decision. Good luck trying to prove that 591 00:33:56,770 --> 00:34:00,490 sometimes. This is a question from who Wong 592 00:34:00,530 --> 00:34:06,610 and he says we can sometimes measure the cost of implementing 593 00:34:06,610 --> 00:34:11,010 data solutions. But how can we measure the 594 00:34:11,010 --> 00:34:14,690 operational costs when a business decides not to 595 00:34:14,770 --> 00:34:19,370 implement certain solutions, such as governance or data 596 00:34:19,370 --> 00:34:22,889 quality? The opportunity cost, the cost 597 00:34:22,889 --> 00:34:26,139 of not doing something. And thank you way for that 598 00:34:26,139 --> 00:34:29,620 question that would. That's that's a big one and I 599 00:34:29,620 --> 00:34:32,900 think it's an important one. If we're going to decide not to 600 00:34:32,900 --> 00:34:37,460 do something, we should decide not to do it on purpose, not 601 00:34:37,460 --> 00:34:40,620 just because we got tired of arguing about it or because we 602 00:34:40,620 --> 00:34:43,699 didn't want to take the effort to get all the data that will be 603 00:34:43,699 --> 00:34:46,860 necessary to make that decision. So one of those annoying 604 00:34:46,860 --> 00:34:49,500 questions that I usually bring into the conversation. 605 00:34:50,060 --> 00:34:53,300 Is if we're going to decide not to do this because there's some 606 00:34:53,300 --> 00:34:56,420 other thing that we want to do and that other thing has been 607 00:34:56,420 --> 00:34:59,460 deemed more important, great let then let's make that decision. 608 00:34:59,780 --> 00:35:02,060 But let's understand the opportunity cost, the cost of 609 00:35:02,060 --> 00:35:05,020 not doing in. Many cases it does become clear 610 00:35:05,020 --> 00:35:09,580 cut because you might have regulations that then levy huge 611 00:35:09,580 --> 00:35:13,220 fines, for instance in the European Union GDPR for 612 00:35:13,220 --> 00:35:16,860 instance, if you don't have the right set up for governance and 613 00:35:16,860 --> 00:35:20,210 privacy and so forth. You'll be hit by a major fine. 614 00:35:20,570 --> 00:35:24,370 In other cases though, when you know they're making these 615 00:35:24,370 --> 00:35:28,330 decisions, they might choose not to do the governance of the 616 00:35:28,330 --> 00:35:32,410 data, but it'll end up reflecting in the actual output 617 00:35:32,410 --> 00:35:35,090 that's being produced, and then somebody has to go back and fix 618 00:35:35,090 --> 00:35:38,130 it. So keeping a keeping tabs of 619 00:35:38,130 --> 00:35:40,770 that, you know, I'm assuming here that you've lost the 620 00:35:40,770 --> 00:35:44,090 argument and they've gone ahead with it without actually, you 621 00:35:44,090 --> 00:35:46,890 know, then keeping tabs on that and raising that every time it 622 00:35:46,890 --> 00:35:49,240 happens. I think very quickly you'll be 623 00:35:49,240 --> 00:35:52,880 able to make a difference in the in the way people are viewing it 624 00:35:53,320 --> 00:35:55,560 because nobody wants, you know wants a disaster. 625 00:35:56,280 --> 00:35:59,600 And if the if the if what's if they've skipped that step which 626 00:35:59,600 --> 00:36:03,920 is which has major magnitude, you know, as sometimes that'll 627 00:36:03,920 --> 00:36:06,440 happen and the collaboration is the name of the game. 628 00:36:06,480 --> 00:36:10,080 So you just want to then keep an eye on it, warn people that this 629 00:36:10,080 --> 00:36:12,240 is going to happen. And every time it happens or 630 00:36:12,240 --> 00:36:15,440 even before it happens, you raise your hand and say look, I 631 00:36:15,440 --> 00:36:16,960 told you about this, now let's do it. 632 00:36:17,560 --> 00:36:22,440 This is from Jav Boshinov, who is a professor at the Harvard 633 00:36:22,440 --> 00:36:27,920 Business School, and he's also been a guest on CXO Talk, and he 634 00:36:27,920 --> 00:36:33,440 says this and Interpol. Maybe I'll ask you first, is 635 00:36:33,440 --> 00:36:38,480 there anything different between generative A I and more 636 00:36:38,480 --> 00:36:42,640 traditional A I? And how should organizations 637 00:36:42,800 --> 00:36:45,640 approach? This I think the best way to 638 00:36:45,640 --> 00:36:49,170 think about generative A I. The promise is that you can do 639 00:36:49,170 --> 00:36:52,130 things conversation. So just as you and I can have a 640 00:36:52,130 --> 00:36:55,450 conversation and we can discuss something and try to get get to 641 00:36:55,450 --> 00:36:57,330 some resolution. That's the hope. 642 00:36:57,330 --> 00:37:01,410 So now if you apply that in a large organization and say I've 643 00:37:01,410 --> 00:37:04,650 got some intelligence that can now conversationally help me do 644 00:37:04,650 --> 00:37:08,650 client support, employee support, my IT operations, etc. 645 00:37:08,890 --> 00:37:11,690 That's you know, hugely, hugely promising. 646 00:37:12,210 --> 00:37:15,650 On the on the other hand, I mean the way these systems work 647 00:37:15,650 --> 00:37:17,930 today. You know, the best way to 648 00:37:17,930 --> 00:37:21,010 understand generative AI that I've been able to get my mind 649 00:37:21,010 --> 00:37:27,490 around it is in a sense, each word is predicted, and then the 650 00:37:27,490 --> 00:37:30,530 word, essentially that word is fed back into the input and then 651 00:37:30,530 --> 00:37:33,610 the next word is predicted. It's almost like when you and I 652 00:37:33,610 --> 00:37:35,250 are talking, I'll sometimes do this. 653 00:37:35,290 --> 00:37:38,370 I'll go out on a limb, I'll start saying something and the 654 00:37:38,370 --> 00:37:41,610 thought hasn't fully formed. Usually I'll manage to come out 655 00:37:41,610 --> 00:37:43,990 of it. And but many times, you know, 656 00:37:44,070 --> 00:37:45,430 I'll end up with my foot and my mouth. 657 00:37:45,430 --> 00:37:48,590 So the generative AI techniques are essentially going out on a 658 00:37:48,590 --> 00:37:52,550 limb every time because it's, which is also why they're not 659 00:37:52,550 --> 00:37:54,310 always consistent with the response. 660 00:37:54,310 --> 00:37:56,070 You know, you might have the same problem to give you a 661 00:37:56,070 --> 00:37:58,550 different response because it's actually working off a 662 00:37:58,550 --> 00:38:01,790 probability distribution. So I think there's a tremendous 663 00:38:01,790 --> 00:38:04,870 amount of promise, but also a tremendous amount of work that 664 00:38:04,870 --> 00:38:08,870 needs to be done to address some of the issues that we've raised 665 00:38:08,870 --> 00:38:11,190 earlier. Anthony, Differences between 666 00:38:11,190 --> 00:38:14,100 Generative AI. And traditional AI and 667 00:38:14,100 --> 00:38:17,140 implications for the enterprise and pretty quickly please 668 00:38:17,660 --> 00:38:20,220 generative. AI is is making stuff that 669 00:38:20,220 --> 00:38:22,980 didn't exist before based on stuff that it observed. 670 00:38:22,980 --> 00:38:26,660 And that stuff can be taxed, it can be images, it can be 671 00:38:27,060 --> 00:38:29,220 anything that we as humans consume. 672 00:38:29,500 --> 00:38:32,420 So the the the challenge to it is that you look at all the 673 00:38:32,420 --> 00:38:35,100 stuff in the past and you you kind of compute on it and do a 674 00:38:35,100 --> 00:38:37,220 lot of math and then you generate something that looks 675 00:38:37,220 --> 00:38:40,420 like a human said it and a human didn't say it. 676 00:38:41,080 --> 00:38:45,000 And so when the world changes and the corpus of data that it's 677 00:38:45,000 --> 00:38:48,520 looking at didn't change fast enough, that nuance gets lost 678 00:38:48,560 --> 00:38:52,880 and we lose the ability to understand something nuanced. 679 00:38:53,040 --> 00:38:57,040 So if the if the purpose is to provide customer support based 680 00:38:57,040 --> 00:38:59,880 on frequently asked questions, or if the purpose is to 681 00:38:59,880 --> 00:39:02,120 summarize a whole bunch of things that you should have read 682 00:39:02,120 --> 00:39:04,800 and didn't have time, it's a fantastic idea. 683 00:39:05,160 --> 00:39:09,160 If the purpose is to to write some new thought leadership on 684 00:39:09,160 --> 00:39:11,800 something. Maybe it's a starting point, but 685 00:39:11,800 --> 00:39:14,440 it would be very careful when we consider that to be an ending 686 00:39:14,440 --> 00:39:18,640 point. Share final thoughts on advice 687 00:39:18,640 --> 00:39:24,040 that you would give to business and technology leaders who want 688 00:39:24,040 --> 00:39:29,240 to be more effective using data using AI. 689 00:39:29,600 --> 00:39:31,920 Interpol. You want to jump in with that 690 00:39:31,920 --> 00:39:33,800 one first. I've been doing this for the 691 00:39:33,800 --> 00:39:38,240 last 2025 years, starting from the days when I did that program 692 00:39:38,240 --> 00:39:41,720 for the NBA to now. And I've always felt that 693 00:39:41,720 --> 00:39:44,120 whenever I was doing it, I thought, oh, it can't get better 694 00:39:44,120 --> 00:39:47,200 than this, but it always seems to get better than that. 695 00:39:47,200 --> 00:39:50,360 And I think we're now in one of those moments where there is the 696 00:39:50,360 --> 00:39:54,040 potential and the opportunity to have a tremendous impact not 697 00:39:54,040 --> 00:39:56,800 just on business but also on society. 698 00:39:57,280 --> 00:40:01,240 And I think because of that implication that there are these 699 00:40:01,240 --> 00:40:05,280 major societal considerations as well, we absolutely have to get 700 00:40:05,280 --> 00:40:07,160 involved. And that would be my biggest. 701 00:40:08,310 --> 00:40:11,030 You know, advice to people either on the business side or 702 00:40:11,030 --> 00:40:14,030 on the technology side, you need to really get involved with 703 00:40:14,030 --> 00:40:16,590 what's happening here. And there's just tremendous, 704 00:40:16,590 --> 00:40:19,830 tremendous potential and it's there's never been a better time 705 00:40:19,830 --> 00:40:22,750 to be involved in data in there. Anthony, it looks like you're 706 00:40:22,750 --> 00:40:26,630 going to get the last word here. Number one, I would say ask why 707 00:40:26,670 --> 00:40:28,870 a lot? Why are we doing this? 708 00:40:28,870 --> 00:40:31,070 What do we have to believe? Why this data? 709 00:40:31,390 --> 00:40:34,310 Make sure that you understand before you jump into. 710 00:40:34,890 --> 00:40:38,130 That method with that data. Make sure that method and that 711 00:40:38,130 --> 00:40:41,530 data are in some way justifiable, like not only 712 00:40:41,530 --> 00:40:43,690 against what you intend to do, but against what you're not 713 00:40:43,690 --> 00:40:47,410 doing by by doing that instead. And then the second thing is 714 00:40:47,410 --> 00:40:50,610 make sure that you pay very close attention that how the 715 00:40:50,610 --> 00:40:54,210 environment is changing so that you don't get caught by the 716 00:40:54,210 --> 00:40:58,090 change that makes what made sense no longer sensible. 717 00:40:58,570 --> 00:41:01,250 And then the last thing is something I always advise, which 718 00:41:01,250 --> 00:41:04,370 is to be humble. It is extremely rare. 719 00:41:04,790 --> 00:41:07,110 When you know everything you need to know and have all the 720 00:41:07,110 --> 00:41:10,230 information you need without widening that circle and 721 00:41:10,230 --> 00:41:13,270 bringing in others that have some sort of expertise or some 722 00:41:13,270 --> 00:41:14,830 sort of perspective that you don't have. 723 00:41:15,070 --> 00:41:18,230 So inviting that expertise and that perspective is not a sign 724 00:41:18,230 --> 00:41:20,430 of weakness, It's a sign of great strength. 725 00:41:20,950 --> 00:41:23,870 With that, unfortunately, we are out of time. 726 00:41:23,990 --> 00:41:28,830 I just want to say a huge thank you to Anthony Scriffiniano and 727 00:41:28,830 --> 00:41:31,750 into Paul Bandari. Anthony, thank you. 728 00:41:31,910 --> 00:41:34,060 It's. Wonderful that you're you've 729 00:41:34,100 --> 00:41:36,340 been here again, and I hope you'll come back another time. 730 00:41:36,860 --> 00:41:38,460 Absolutely. Thank you so much, Michael. 731 00:41:38,860 --> 00:41:42,660 And Interpol, so honored that you joined us. 732 00:41:42,660 --> 00:41:45,340 And again, I hope you as well will come back and be a guest on 733 00:41:45,460 --> 00:41:47,900 CXO Talk again and on another date. 734 00:41:48,300 --> 00:41:50,740 Delighted to do that, Michael. Thank you for having me. 735 00:41:50,740 --> 00:41:53,380 And for those with unanswered questions, please, you know 736 00:41:53,420 --> 00:41:56,220 Lincoln and we can continue the conversation. 737 00:41:56,690 --> 00:41:59,450 Everybody, thank you for watching, especially those folks 738 00:41:59,450 --> 00:42:01,330 who just ask such great questions. 739 00:42:01,330 --> 00:42:07,850 You are such a smart and bright audience and we love your 740 00:42:07,850 --> 00:42:12,450 questions and keep watching. CXO talk, go to cxotalk.com, be 741 00:42:12,450 --> 00:42:17,450 sure to subscribe to our YouTube channel and hit the subscribe 742 00:42:17,450 --> 00:42:22,170 button at the bottom of our web page and you can subscribe to 743 00:42:22,170 --> 00:42:24,290 our newsletter and we'll tell you and notify you. 744 00:42:24,650 --> 00:42:27,650 About our excellent upcoming shows and guests, we have lots 745 00:42:27,650 --> 00:42:29,410 of them. Everybody, thank you so much. 746 00:42:29,410 --> 00:42:31,450 Hope you have a great day and we'll see you again next time. 747 00:42:31,530 --> 00:42:31,930 Bye, bye.