WEBVTT 00:00:00.000 --> 00:00:03.600 Welcome back to the Deep Dive. Today, we're really 00:00:03.600 --> 00:00:05.860 on a mission. We want to cut through all the 00:00:05.860 --> 00:00:09.160 noise, the hype around AI adoption. Yeah, there's 00:00:09.160 --> 00:00:11.419 a lot of it. There is. And dive straight into 00:00:11.419 --> 00:00:13.960 the practical stuff. Enterprise readiness. We've 00:00:13.960 --> 00:00:16.219 gathered sources from a major player in tech 00:00:16.219 --> 00:00:18.539 services. A leader. And our goal here is to give 00:00:18.539 --> 00:00:21.260 you a clear roadmap. How do you get from just 00:00:21.260 --> 00:00:25.000 wanting AI adoption to actually being ready and 00:00:25.000 --> 00:00:28.500 scaling it across the whole organization? In 00:00:28.500 --> 00:00:31.410 this Deep Dive, It really benefits from a perspective 00:00:31.410 --> 00:00:33.570 that's, well, deeply rooted in actually doing 00:00:33.570 --> 00:00:35.929 it, getting it done. Execution. Exactly. The 00:00:35.929 --> 00:00:37.890 specific insights, they come from Linda Yow. 00:00:38.149 --> 00:00:41.130 She's the COO for Lenovo's Solutions and Services 00:00:41.130 --> 00:00:43.990 Group. OK. Her background is, frankly, quite 00:00:43.990 --> 00:00:46.109 unique. She was employee number two at Linamar, 00:00:46.210 --> 00:00:48.990 so she has that real rapid growth foundational 00:00:48.990 --> 00:00:51.950 experience. Wow. And crucially, she spent five 00:00:51.950 --> 00:00:54.509 years incubating data science practices at Boeing 00:00:54.509 --> 00:00:57.969 Global Services. Managed teams in the US, in 00:00:57.969 --> 00:01:01.640 India. So real global perspective. Totally. And 00:01:01.640 --> 00:01:04.719 her focus is on services, the delivery, the integration, 00:01:05.120 --> 00:01:07.540 not just the hardware. That's why her views on 00:01:07.540 --> 00:01:11.180 what works are so well grounded, practical. OK, 00:01:11.180 --> 00:01:13.019 let's definitely unpack this. Before we jump 00:01:13.019 --> 00:01:15.299 into the methodology she shared, maybe let's 00:01:15.299 --> 00:01:17.459 frame the moment we're in. What's the single 00:01:17.459 --> 00:01:20.280 biggest trend that explains why technology adoption, 00:01:20.780 --> 00:01:24.180 especially AI, is suddenly a boardroom issue, 00:01:24.359 --> 00:01:27.260 not just, you know, an IT department thing anymore? 00:01:27.579 --> 00:01:29.299 Well, the sources are pretty clear on this. Yeah. 00:01:29.480 --> 00:01:33.060 The role of IT, it's fundamentally changed. How 00:01:33.060 --> 00:01:35.719 so? IT used to be, you know, the back office, 00:01:36.140 --> 00:01:38.219 a cost center, just keeping the lights on, keeping 00:01:38.219 --> 00:01:40.900 systems running. Right. But now, technology is 00:01:40.900 --> 00:01:43.000 moving rapidly out of that back room and directly 00:01:43.000 --> 00:01:44.840 into the hands of the business units themselves. 00:01:44.980 --> 00:01:47.239 OK, wait. So if IT is suddenly everywhere, what 00:01:47.239 --> 00:01:50.219 does that actually mean for the CIO? the person 00:01:50.219 --> 00:01:53.239 in charge. It means the CIO has basically transformed 00:01:53.239 --> 00:01:56.260 into an IT business partner. They're responsible 00:01:56.260 --> 00:01:58.579 not just for keeping servers running, but for 00:01:58.579 --> 00:02:01.560 frontline business outcomes. For P &L. For results. 00:02:02.340 --> 00:02:05.799 Absolutely. And this forces a new mandate. Technologies, 00:02:06.140 --> 00:02:08.340 especially something potentially massive like 00:02:08.340 --> 00:02:11.599 AI, must have demonstrable practical grounding. 00:02:11.849 --> 00:02:15.250 It has to work seamlessly from the headquarters 00:02:15.250 --> 00:02:18.530 boardroom all the way out to, say, a factory 00:02:18.530 --> 00:02:20.889 floor or somewhere else. That sounds like a huge 00:02:20.889 --> 00:02:23.090 shift in accountability, though, if the CIO is 00:02:23.090 --> 00:02:25.949 now on the hook for outcomes in P &L. I mean, 00:02:25.949 --> 00:02:28.210 aren't we just giving them massive new risks 00:02:28.210 --> 00:02:30.629 without maybe giving them more control over everything? 00:02:30.810 --> 00:02:33.210 That's precisely why the conversation has escalated, 00:02:33.210 --> 00:02:35.509 right, to the board level. The stakes are just 00:02:35.509 --> 00:02:37.810 much, much higher now. OK. And the major catalyst, 00:02:37.930 --> 00:02:39.650 the thing that really accelerated this whole 00:02:39.650 --> 00:02:42.810 high -stakes shift, well, It's generative AI. 00:02:43.150 --> 00:02:45.930 Right. GenAI. Of course. It's given AI yet another 00:02:45.930 --> 00:02:48.830 reason to be front and center because GenAI makes 00:02:48.830 --> 00:02:51.669 it truly collaborative, productive, not just 00:02:51.669 --> 00:02:53.550 for the organization as a whole, but for every 00:02:53.550 --> 00:02:55.909 single individual user. And we're seeing real 00:02:55.909 --> 00:02:58.030 tangible effects already, aren't we? Not just 00:02:58.030 --> 00:03:01.050 theory. Exactly. Lindy Au specifically pointed 00:03:01.050 --> 00:03:05.569 out companies using GenAI right now to run, say, 00:03:05.729 --> 00:03:09.400 a much more efficient contact center. or to dramatically 00:03:09.400 --> 00:03:11.659 streamline how software engineering gets done. 00:03:12.419 --> 00:03:15.520 And because gen AIs by its nature sort of democratized 00:03:15.520 --> 00:03:18.520 it, you access it with natural language, right? 00:03:18.919 --> 00:03:21.979 It creates this incredibly powerful multiplicative 00:03:21.979 --> 00:03:25.460 effect. It scales knowledge instantly, broadly 00:03:25.460 --> 00:03:28.229 across the entire enterprise. OK, so that brings 00:03:28.229 --> 00:03:30.189 us right to the big challenge. You've got senior 00:03:30.189 --> 00:03:32.870 executives pushing hard for AI adoption. They 00:03:32.870 --> 00:03:36.069 are. And interestingly, unlike maybe some previous 00:03:36.069 --> 00:03:38.229 tech waves, people aren't really skeptical. They 00:03:38.229 --> 00:03:40.370 seem eager. They're taking courses, watching 00:03:40.370 --> 00:03:43.210 videos. That desire is definitely there. People 00:03:43.210 --> 00:03:45.169 want to embrace it, yes. So if the willingness 00:03:45.169 --> 00:03:47.389 is there, why are so many companies still kind 00:03:47.389 --> 00:03:50.389 of stuck, you know, stuck in pilot project mode, 00:03:50.650 --> 00:03:53.310 failing to actually scale this thing up? Because 00:03:53.310 --> 00:03:56.430 desire and, let's say, Infrastructure readiness 00:03:56.430 --> 00:03:59.530 are two very, very different things. Once organizations 00:03:59.530 --> 00:04:02.469 move past that initial, wow, this is cool excitement, 00:04:03.030 --> 00:04:06.050 they hit these massive, practical roadblocks. 00:04:06.729 --> 00:04:08.849 The willingness is there, sure, but when you 00:04:08.849 --> 00:04:11.270 look at large -scale enterprise adoption, the 00:04:11.270 --> 00:04:12.990 questions just become overwhelming. Like what? 00:04:13.009 --> 00:04:15.930 What are the big ones? Well, first, where exactly 00:04:15.930 --> 00:04:19.430 do we even start this whole process? Then. Is 00:04:19.430 --> 00:04:21.649 our security posture actually strong enough? 00:04:22.029 --> 00:04:25.009 Can we protect our proprietary data while we're 00:04:25.009 --> 00:04:28.769 training these models? Huge concern. Huge. Then 00:04:28.769 --> 00:04:32.029 policies. Are corporate policies mature enough? 00:04:32.310 --> 00:04:34.589 Can they handle responsible AI use? Things like 00:04:34.589 --> 00:04:37.290 bias, ethics. Right, the responsible AI piece. 00:04:37.529 --> 00:04:39.910 And then the immediate stuff. Has our talent 00:04:39.910 --> 00:04:41.629 actually been up skilled for this? Do people 00:04:41.629 --> 00:04:44.350 know how to use these tools effectively? And 00:04:44.350 --> 00:04:47.560 fundamentally. Is the underlying tech stack robust 00:04:47.560 --> 00:04:49.920 enough to even support these capabilities at 00:04:49.920 --> 00:04:52.879 scale? Hmm. This sounds less like a pure technology 00:04:52.879 --> 00:04:55.160 problem and maybe more like a leadership and 00:04:55.160 --> 00:04:56.779 process problem. I think that's fair to say. 00:04:57.220 --> 00:04:59.600 So if readiness is the big hurdle, what's the 00:04:59.600 --> 00:05:02.540 solution? Did Linda Yao and Lenovo share a kind 00:05:02.540 --> 00:05:04.660 of battle -tested methodology for getting over 00:05:04.660 --> 00:05:07.240 it? They did. They have a proven framework. It 00:05:07.240 --> 00:05:09.870 centers on four core pillars. And the guiding 00:05:09.870 --> 00:05:12.350 principle, sort of the operational mantra, is 00:05:12.350 --> 00:05:15.430 security first and people next. Okay, security 00:05:15.430 --> 00:05:18.350 first, people next. Let's break down those four 00:05:18.350 --> 00:05:20.550 pillars. This feels like where the really actionable 00:05:20.550 --> 00:05:26.110 insights are. Pillar one, security. Why start 00:05:26.110 --> 00:05:28.389 there? Why is that the immediate top priority? 00:05:28.949 --> 00:05:32.370 Security is defined, quite simply, as the thing 00:05:32.370 --> 00:05:34.959 that keeps leadership up at night. Huh, okay, 00:05:35.180 --> 00:05:37.100 fair enough. It's not just about warding off 00:05:37.100 --> 00:05:39.740 external hackers, right? It's also about managing 00:05:39.740 --> 00:05:43.699 internal risk, data security, privacy. For AI 00:05:43.699 --> 00:05:45.800 specifically, this means things like mitigating 00:05:45.800 --> 00:05:48.500 the risk of your own intellectual property leaking 00:05:48.500 --> 00:05:50.720 out when models are trained. Right, your secret 00:05:50.720 --> 00:05:53.529 sauce getting into the model. Exactly. and also 00:05:53.529 --> 00:05:55.930 guarding against the very serious threat of data 00:05:55.930 --> 00:05:58.269 poisoning, where someone could deliberately mess 00:05:58.269 --> 00:06:01.050 with your training data to compromise the AI's 00:06:01.050 --> 00:06:03.449 output, its trustworthiness. Okay, so you absolutely 00:06:03.449 --> 00:06:05.389 have to lay down those foundational security 00:06:05.389 --> 00:06:07.310 guardrails before you do anything else. You have 00:06:07.310 --> 00:06:09.730 to. It's non -negotiable. That clarity is crucial. 00:06:10.350 --> 00:06:12.610 So if security is the first really high -stakes 00:06:12.610 --> 00:06:16.410 barrier, the second pillar is people. How do 00:06:16.410 --> 00:06:19.910 they approach the human side of this huge transition? 00:06:20.269 --> 00:06:22.839 Pillar number two, people. This focuses really 00:06:22.839 --> 00:06:25.699 heavily on change management, on adoption, and 00:06:25.699 --> 00:06:29.420 crucially on training. To get buy -in to actually 00:06:29.420 --> 00:06:32.980 make this work, organizations must clearly define 00:06:32.980 --> 00:06:36.980 which tasks GEN .AI is going to augment. Make 00:06:36.980 --> 00:06:40.220 human workers better, faster, versus which tasks 00:06:40.220 --> 00:06:42.160 it might actually replace. That distinction feels 00:06:42.160 --> 00:06:44.399 critical. Augment versus replace. It's absolutely 00:06:44.399 --> 00:06:46.279 critical for trust and adoption. Can you give 00:06:46.279 --> 00:06:48.019 us the sort of concrete example of that? Augment 00:06:48.019 --> 00:06:50.480 versus replace. Sure. So augmentation might look 00:06:50.480 --> 00:06:54.300 like GNI drafting the first version of a complex 00:06:54.300 --> 00:06:57.560 legal brief maybe, or an internal memo. The human 00:06:57.560 --> 00:06:59.560 lawyer or the manager then takes that draft, 00:06:59.759 --> 00:07:01.740 refines it, approves it, it speeds things up. 00:07:01.819 --> 00:07:04.060 Got it. Faster, better work. Right. Replacement, 00:07:04.120 --> 00:07:06.540 on the other hand. Might be GNI handling, say, 00:07:06.540 --> 00:07:08.980 100 % of tier one customer support chats, routing 00:07:08.980 --> 00:07:11.259 issues, classifying common problems, maybe even 00:07:11.259 --> 00:07:13.500 resolving simple ones entirely on its own. Completely 00:07:13.500 --> 00:07:15.980 automated. Exactly. Getting employees to really 00:07:15.980 --> 00:07:19.160 understand which category their specific tasks 00:07:19.160 --> 00:07:21.699 fall into that's essential for building trust 00:07:21.699 --> 00:07:24.079 and making the transition smooth. OK, so if you 00:07:24.079 --> 00:07:26.939 manage security and people, arguably the two 00:07:26.939 --> 00:07:30.529 hardest, most human -centric pillars, What about 00:07:30.529 --> 00:07:32.709 the actual technology needed? Is that the easy 00:07:32.709 --> 00:07:35.769 part? Well, that's pillar number three, technology. 00:07:36.370 --> 00:07:39.850 And interestingly, the expert expressed well. 00:07:40.009 --> 00:07:42.750 quite strong confidence here. Really? Yeah. The 00:07:42.750 --> 00:07:44.649 thinking is if you have your security parameters 00:07:44.649 --> 00:07:47.230 locked down and your people processes are clear 00:07:47.230 --> 00:07:50.810 and working, the technology itself can be iterative. 00:07:51.589 --> 00:07:53.670 Modern big tech players, they're agile, they 00:07:53.670 --> 00:07:55.610 can adapt, they can co -develop solutions, they 00:07:55.610 --> 00:07:58.350 can pivot if needed. The confidence level seems 00:07:58.350 --> 00:08:01.009 high, that if the enterprise knows what it needs 00:08:01.009 --> 00:08:03.829 the AI to do and how to secure it and manage 00:08:03.829 --> 00:08:06.750 the people side, then they can reliably engineer 00:08:06.750 --> 00:08:08.470 the tech solution to do it. That's interesting. 00:08:08.589 --> 00:08:10.509 So the tech isn't the main bottleneck if the 00:08:10.509 --> 00:08:12.649 other pieces are right. It seems to be the perspective. 00:08:12.790 --> 00:08:15.649 OK. And that brings us to the final pillar. Pillar 00:08:15.649 --> 00:08:20.250 four, process. Right. Process. So once security 00:08:20.250 --> 00:08:22.610 and people parameters are set, this is where 00:08:22.610 --> 00:08:25.149 deep industry experience really becomes paramount. 00:08:25.589 --> 00:08:28.350 You define the new workflows. You integrate those 00:08:28.350 --> 00:08:31.009 security guardrails and ethical checks. And then 00:08:31.009 --> 00:08:33.750 you standardize those processes so you can scale 00:08:33.750 --> 00:08:35.889 them across every business unit. So it's not 00:08:35.889 --> 00:08:38.330 just a one -off experiment? Yeah, exactly. This 00:08:38.330 --> 00:08:40.850 structure ensures that a successful pilot project 00:08:40.850 --> 00:08:42.889 doesn't just date a cool little experiment in 00:08:42.889 --> 00:08:45.169 one corner of the company, it becomes a repeatable 00:08:45.169 --> 00:08:47.710 blueprint for everyone. Okay, we have the methodology, 00:08:48.070 --> 00:08:51.490 security, people, technology, process. But, you 00:08:51.490 --> 00:08:54.230 know, in a world full of consultants selling 00:08:54.230 --> 00:08:57.500 frameworks, What did Linda Yao identify as the 00:08:57.500 --> 00:09:00.259 real secret sauce? What separates the companies 00:09:00.259 --> 00:09:03.820 that successfully scale this from those just 00:09:03.820 --> 00:09:06.279 stuck reviewing PowerPoint decks? Yeah, yeah. 00:09:06.840 --> 00:09:08.659 The secret sauce, according to her, isn't some 00:09:08.659 --> 00:09:11.539 magic bullet. It's internal expertise, or as 00:09:11.539 --> 00:09:13.460 they call it, eating its own cooking. Eating 00:09:13.460 --> 00:09:16.379 its own cooking. Meaning Lenovo acts as its own 00:09:16.379 --> 00:09:18.759 internal sandbox, its own guinea pig, basically. 00:09:18.960 --> 00:09:22.159 They implement and they refine AI use cases internally 00:09:22.159 --> 00:09:25.820 first in their own massive global contact centers, 00:09:26.240 --> 00:09:29.620 in their software engineering teams, across individual 00:09:29.620 --> 00:09:32.649 employee productivity tools. They do it themselves 00:09:32.649 --> 00:09:34.629 before they start rolling out similar solutions 00:09:34.629 --> 00:09:37.250 or advice to clients. OK, but how does being 00:09:37.250 --> 00:09:39.669 your own guinea pig actually prevent that common 00:09:39.669 --> 00:09:42.970 problem, where an AI solution looks amazing in 00:09:42.970 --> 00:09:44.970 a controlled lab setting, but then completely 00:09:44.970 --> 00:09:47.529 falls apart in a real messy enterprise environment? 00:09:47.809 --> 00:09:50.049 Because they hit those real -world roadblocks 00:09:50.049 --> 00:09:53.320 first themselves. Ah, they feel the pain. Exactly. 00:09:53.639 --> 00:09:55.200 They learn the hard way how to integrate these 00:09:55.200 --> 00:09:58.240 new tools with messy legacy systems. They figure 00:09:58.240 --> 00:10:01.500 out how to actually train a diverse global workforce 00:10:01.500 --> 00:10:04.179 effectively. They find the policy gaps and the 00:10:04.179 --> 00:10:06.559 practical hurdles. So when they advise clients, 00:10:06.620 --> 00:10:09.340 they're offering a guidebook based on, well... 00:10:08.990 --> 00:10:12.110 failure and iteration and real experience, not 00:10:12.110 --> 00:10:14.029 just abstract theory. That makes a lot of sense. 00:10:14.169 --> 00:10:15.929 Okay, let's shift gears a bit and look at the 00:10:15.929 --> 00:10:17.870 history here. You mentioned AI has been around 00:10:17.870 --> 00:10:21.230 for decades in some form. So why now? Why are 00:10:21.230 --> 00:10:23.429 senior management and boards dedicating so much 00:10:23.429 --> 00:10:26.090 focus, so much capital to it right now? Yeah, 00:10:26.090 --> 00:10:28.090 that's a great question. We can understand the 00:10:28.090 --> 00:10:30.990 current, let's call it, excitement level better 00:10:30.990 --> 00:10:33.470 if we put it in context with the previous two 00:10:33.470 --> 00:10:37.000 sort of waves of AI enthusiasm. The core tech 00:10:37.000 --> 00:10:39.519 might have been there, but the enabling environment 00:10:39.519 --> 00:10:43.000 wasn't quite right before. OK. So what defined 00:10:43.000 --> 00:10:47.580 that first wave? The first wave, I think, roughly 00:10:47.580 --> 00:10:50.440 mid -2000s. It's often called the Flashboys era. 00:10:50.559 --> 00:10:53.539 Ah, high frequency trading. Exactly. HFT in the 00:10:53.539 --> 00:10:56.840 financial sector. Using complex algorithms, super 00:10:56.840 --> 00:10:59.320 high performance computing to execute financial 00:10:59.320 --> 00:11:01.809 trades at, you know, lightning speed. Right. 00:11:01.970 --> 00:11:04.549 This definitely proved AI's raw power and speed, 00:11:04.870 --> 00:11:07.090 but its impact was pretty niche, right? Highly 00:11:07.090 --> 00:11:09.789 specialized. It did, however, spark some of the 00:11:09.789 --> 00:11:12.850 first really serious conversations around needing 00:11:12.850 --> 00:11:15.389 new policies, new processes to govern this stuff. 00:11:15.629 --> 00:11:17.610 OK, so it showed power in finance, but maybe 00:11:17.610 --> 00:11:19.789 not broad applicability for the average company. 00:11:19.950 --> 00:11:21.870 What changed in the second wave? That was more 00:11:21.870 --> 00:11:25.169 the 2010s, and it was defined primarily by the 00:11:25.169 --> 00:11:28.029 rise of accessible cloud computing. Cloud. Cloud 00:11:28.029 --> 00:11:31.370 made computing power relatively cheap. scalable, 00:11:31.929 --> 00:11:35.470 easy to access for almost anyone. And this just 00:11:35.470 --> 00:11:37.789 fundamentally boosted awareness of the value 00:11:37.789 --> 00:11:40.690 of data. It spurred the adoption of sophisticated 00:11:40.690 --> 00:11:44.450 data visualization tools. And it led to, well, 00:11:44.750 --> 00:11:47.019 almost every industry... hiring data science 00:11:47.019 --> 00:11:49.080 departments. That's when DI really took off. 00:11:49.580 --> 00:11:51.360 Absolutely. Business intelligence, foundational 00:11:51.360 --> 00:11:54.259 data governance, that all matured significantly 00:11:54.259 --> 00:11:56.360 during that second wave. OK, so we went from 00:11:56.360 --> 00:11:59.419 niche finance innovation with HFT to widespread 00:11:59.419 --> 00:12:01.720 data awareness thanks to the cloud. And now we 00:12:01.720 --> 00:12:05.159 hit this third wave, gen AI democratization in, 00:12:05.200 --> 00:12:08.639 say, 2024. If gen AI is supposed to be the ultimate 00:12:08.639 --> 00:12:10.820 goal, the big prize, does that mean all that 00:12:10.820 --> 00:12:12.940 foundational BI work, the data governance from 00:12:12.940 --> 00:12:15.659 the last wave is that now obsolete, should companies 00:12:15.659 --> 00:12:18.200 just leapfrog that? Oh, quite the opposite, actually. 00:12:18.340 --> 00:12:21.039 Linda Yao suggested that GNI is like the tide 00:12:21.039 --> 00:12:23.259 that's going to raise all boats. OK, explain 00:12:23.259 --> 00:12:25.700 that. How does the fancy new stuff help the older 00:12:25.700 --> 00:12:28.179 foundational stuff? Well, here's the key takeaway 00:12:28.179 --> 00:12:31.620 she offered. While loads of organizations are 00:12:31.620 --> 00:12:35.139 racing toward full GNI adoption, many quickly 00:12:35.139 --> 00:12:37.000 discover they aren't technically ready for it 00:12:37.000 --> 00:12:39.940 yet. They lack the data quality, the governance. 00:12:40.240 --> 00:12:43.519 Right, the stuff from Wave 2. Exactly. But the 00:12:43.519 --> 00:12:46.820 sheer push for GEN .AI from the top forces these 00:12:46.820 --> 00:12:49.059 companies to move up the data and analytics value 00:12:49.059 --> 00:12:51.960 chain anyway, even if they can't implement full 00:12:51.960 --> 00:12:54.960 GEN .AI tomorrow. So the ambition drives foundational 00:12:54.960 --> 00:12:57.360 improvement. Precisely. It means investment, 00:12:57.799 --> 00:13:01.769 education, executive attention. It's now massively 00:13:01.769 --> 00:13:03.830 boosted for those fundamental building blocks, 00:13:04.450 --> 00:13:06.490 data quality, data governance, even traditional 00:13:06.490 --> 00:13:09.750 BI. The fundamentals end up benefiting enormously 00:13:09.750 --> 00:13:11.629 from all the excitement around the advanced stuff. 00:13:11.950 --> 00:13:14.210 That connects the dots really well between organizational 00:13:14.210 --> 00:13:16.429 priorities. Okay, let's circle back to security 00:13:16.429 --> 00:13:18.289 for a moment, specifically thinking about the 00:13:18.289 --> 00:13:21.169 modern, often remote workforce. How on earth 00:13:21.169 --> 00:13:23.990 do you manage AI risks when employees are using 00:13:23.990 --> 00:13:26.549 these tools outside the company's physical walls, 00:13:27.029 --> 00:13:28.850 potentially using apps you haven't even approved? 00:13:29.000 --> 00:13:31.559 Yeah, that's a huge challenge. Yeah. And it sits 00:13:31.559 --> 00:13:34.659 right at that critical intersection of the security 00:13:34.659 --> 00:13:38.080 pillar and the people pillar. OK. The expert 00:13:38.080 --> 00:13:41.200 view here was that managing AI outside the company 00:13:41.200 --> 00:13:43.940 walls isn't about trying to build some giant, 00:13:44.120 --> 00:13:46.700 impenetrable digital fence. That's probably futile. 00:13:46.860 --> 00:13:49.799 Right. Instead, it requires identifying the user 00:13:49.799 --> 00:13:52.379 persona and the specific use case they're working 00:13:52.379 --> 00:13:55.419 on. That allows you to define appropriate tailored 00:13:55.419 --> 00:13:58.080 guardrails. Because everyone uses AI differently, 00:13:58.279 --> 00:14:00.799 right? Yeah. And their need for data access varies 00:14:00.799 --> 00:14:03.000 hugely. Can you elaborate on that? How different 00:14:03.000 --> 00:14:05.799 can those personas be within one company? Oh, 00:14:05.799 --> 00:14:08.299 absolutely. Think about three distinct users 00:14:08.299 --> 00:14:10.740 that were mentioned in the source material. First, 00:14:10.759 --> 00:14:12.799 you might have your typical office worker, maybe 00:14:12.799 --> 00:14:15.320 dialing in from an airport lounge, using AI for 00:14:15.320 --> 00:14:18.059 quick collaboration, maybe drafting emails, prepping 00:14:18.059 --> 00:14:20.360 for a big presentation. Pretty standard productivity 00:14:20.360 --> 00:14:22.759 stuff. OK, persona one. Second, picture someone 00:14:22.759 --> 00:14:24.879 on a manufacturing floor. They might be using 00:14:24.879 --> 00:14:27.779 computer vision. powered by AI to inspect the 00:14:27.779 --> 00:14:29.639 quality of components coming down the line in 00:14:29.639 --> 00:14:31.919 real time. Very different context, different 00:14:31.919 --> 00:14:34.460 data. Totally different. Okay, third. Third, 00:14:34.960 --> 00:14:37.860 maybe someone in the ESG office. environmental 00:14:37.860 --> 00:14:40.799 social governance. They could be using AI to 00:14:40.799 --> 00:14:42.840 review vast amounts of documents, maybe legal 00:14:42.840 --> 00:14:45.940 transcripts, regulatory filings, to monitor environmental 00:14:45.940 --> 00:14:49.279 compliance. Again, highly sensitive data, specific 00:14:49.279 --> 00:14:51.700 task. And the guardrails, the security controls 00:14:51.700 --> 00:14:53.519 for each of those three must be fundamentally 00:14:53.519 --> 00:14:55.620 different. Precisely. The manufacturing worker 00:14:55.620 --> 00:14:58.080 needs guardrails around the integrity and accuracy 00:14:58.080 --> 00:15:00.259 of that vision model's output. Is it spotting 00:15:00.259 --> 00:15:03.639 defects correctly? The ESG analyst needs strict 00:15:03.639 --> 00:15:06.340 guardrails around data privacy, handling potentially 00:15:06.340 --> 00:15:09.320 sensitive of regulatory or legal documents. The 00:15:09.320 --> 00:15:11.360 office worker needs controls around maybe not 00:15:11.360 --> 00:15:13.919 pasting confidential company strategy into a 00:15:13.919 --> 00:15:16.639 public AI tool. The solution has to be tailored 00:15:16.639 --> 00:15:19.600 to the objective, the data sensitivity, and the 00:15:19.600 --> 00:15:22.399 user's actual need to know, which loops right 00:15:22.399 --> 00:15:24.620 back to why that people pillar understanding 00:15:24.620 --> 00:15:27.259 users, training them is so critical alongside 00:15:27.259 --> 00:15:30.659 security. Okay, finally then, let's tackle the 00:15:30.659 --> 00:15:33.019 broader, more general risks we always hear about 00:15:33.019 --> 00:15:35.960 with AI. Things like bias creeping into models, 00:15:36.259 --> 00:15:38.340 AI hallucinating or making things up, ethical 00:15:38.340 --> 00:15:41.200 concerns, privacy violations. How do we manage 00:15:41.200 --> 00:15:43.679 those? Well, the perspective shared was that 00:15:43.679 --> 00:15:46.299 managing these kinds of ethical and quality risks, 00:15:46.639 --> 00:15:48.980 it always points back to those four foundational 00:15:48.980 --> 00:15:51.860 pillars again. Security, people, technology, 00:15:52.220 --> 00:15:55.320 process. Exactly. Risk mitigation here requires 00:15:55.320 --> 00:15:58.139 careful, safe experimentation. It really comes 00:15:58.139 --> 00:16:00.240 down to fundamentally understanding the user, 00:16:00.440 --> 00:16:02.200 their objective, what they're trying to achieve 00:16:02.200 --> 00:16:05.440 with the AI. and then ensuring you hire or maybe 00:16:05.440 --> 00:16:08.659 retain the right internal or external experts 00:16:08.659 --> 00:16:10.899 who can advise on how to draw those specific 00:16:10.899 --> 00:16:13.539 parameters for safe use and testing. So it's 00:16:13.539 --> 00:16:15.179 about structured management, not just hoping 00:16:15.179 --> 00:16:18.259 for the best. Absolutely. In the experience shared 00:16:18.259 --> 00:16:20.860 in the sources, almost every single risk management 00:16:20.860 --> 00:16:23.100 concern you can think of, whether it's bias, 00:16:23.259 --> 00:16:26.909 hallucination, privacy, ethics. It can usually 00:16:26.909 --> 00:16:29.269 be traced back to a weakness in one or more of 00:16:29.269 --> 00:16:32.370 those four boxes. Maybe poor security protocols 00:16:32.370 --> 00:16:35.169 allowed bad data in. Maybe inadequate training 00:16:35.169 --> 00:16:37.950 meant people used it wrong. Perhaps an unstable 00:16:37.950 --> 00:16:40.169 tech architecture led to unreliable outputs. 00:16:40.789 --> 00:16:43.110 Or maybe there was just a lack of defined processes 00:16:43.110 --> 00:16:45.889 for checking errors or handling exceptions. That 00:16:45.889 --> 00:16:48.350 four pillar structure seems incredibly robust 00:16:48.350 --> 00:16:51.110 for diagnosing problems then. Okay, we've covered 00:16:51.110 --> 00:16:53.370 readiness, the history of the risks. One last 00:16:53.370 --> 00:16:56.309 big question. Does adopting these emerging technologies 00:16:56.309 --> 00:16:59.470 like AI, does it fundamentally change a company's 00:16:59.470 --> 00:17:01.639 business model? Or does it mostly just make the 00:17:01.639 --> 00:17:04.039 existing model run more efficiently? That's a 00:17:04.039 --> 00:17:05.759 challenging question. And the insight here was, 00:17:05.759 --> 00:17:07.779 well, quite pointed. It doesn't always change 00:17:07.779 --> 00:17:10.119 the business model. Yeah. But the strong consensus 00:17:10.119 --> 00:17:12.700 seemed to be that it should. It should. Why? 00:17:13.079 --> 00:17:16.039 Think about it. If a company puts this enormous 00:17:16.039 --> 00:17:19.500 effort, massive investment in time, money, focus 00:17:19.500 --> 00:17:22.960 into refining AI, making it secure, ethical, 00:17:23.200 --> 00:17:26.240 responsible, unbiased, adaptable, all that work, 00:17:26.779 --> 00:17:29.779 and yet their fundamental business model remains 00:17:29.779 --> 00:17:33.880 static, unchanged, that arguably that enormous 00:17:33.880 --> 00:17:36.480 effort was ultimately wasted, or at least under 00:17:36.480 --> 00:17:39.119 -leveraged. The potential for true business model 00:17:39.119 --> 00:17:41.819 transformation, that's seen as the real payoff. 00:17:42.039 --> 00:17:44.859 That's what justifies the immense effort involved. 00:17:45.019 --> 00:17:47.359 That really brings us to a compelling final thought 00:17:47.359 --> 00:17:49.759 for you, the listener. History kind of shows 00:17:49.759 --> 00:17:51.519 us, doesn't it, that whenever a powerful new 00:17:51.519 --> 00:17:53.680 technology emerges, I think the internet, maybe 00:17:53.680 --> 00:17:56.779 mobile phones, now AI, there's always this corresponding 00:17:56.779 --> 00:17:59.319 wave of concern about relevance. People worry, 00:17:59.319 --> 00:18:01.920 right? Will this replace my job? And it's understandable. 00:18:02.599 --> 00:18:05.299 But historical data, again and again, consistently 00:18:05.299 --> 00:18:07.380 shows that technology generally helps augment 00:18:07.380 --> 00:18:09.720 most workers. It tends to take over the tedious, 00:18:10.099 --> 00:18:12.400 the repetitive tasks. Freeing people up. Exactly. 00:18:12.799 --> 00:18:15.200 Making our jobs hopefully more productive, maybe 00:18:15.200 --> 00:18:17.779 even more enjoyable, and allowing us as individuals 00:18:17.779 --> 00:18:20.500 to focus our unique human expertise on higher 00:18:20.500 --> 00:18:23.839 value, more creative, more strategic tasks. That's 00:18:23.839 --> 00:18:27.200 a great perspective to maybe end on. So considering 00:18:27.200 --> 00:18:29.420 those three ways of AI, we talked about the niche 00:18:29.420 --> 00:18:32.279 HRT wave, the cloud and data awareness wave, 00:18:32.279 --> 00:18:35.740 and now this gen AI democratization wave. Here's 00:18:35.740 --> 00:18:37.970 a final thought for you, the listener. to ponder, 00:18:38.250 --> 00:18:40.829 where do you see the next big policy debate emerging? 00:18:41.390 --> 00:18:43.509 Will the political and corporate focus be mainly 00:18:43.509 --> 00:18:46.390 on security, on preventing IP leakage and misuse? 00:18:46.730 --> 00:18:48.730 Or will it shift more towards defining those 00:18:48.730 --> 00:18:50.849 tricky societal boundaries between automation 00:18:50.849 --> 00:18:53.269 and argumentation? And maybe more personally, 00:18:53.609 --> 00:18:56.450 what skill will you prioritize learning or honing 00:18:56.450 --> 00:18:58.569 next now that you know those four pillars of 00:18:58.569 --> 00:19:00.250 readiness? Some interesting questions to think 00:19:00.250 --> 00:19:02.230 about. Thank you for joining us on this deep 00:19:02.230 --> 00:19:04.490 dive into strategic AI readiness and leadership. 00:19:04.789 --> 00:19:05.450 Until next time.