469 Why AI Adoption Is a Psychology Problem, Not a Technology Problem with Gleb Tsipursky の全文書き起こし

the future of work, the problem is more than technology. 00:04:48.959 --> 00:04:58.399 And people are leaders are pattern matching to previous technological disruptions and adoption and change management, which doesn't work for AI. 00:04:58.560 --> 00:04:59.439 Let me tell you why. 00:04:59.600 --> 00:05:11.279 Let's say you're adopting new technology, let's say you're adopting something like sales enablement like Salesforce, or a new CRM like HubSun, or a new ERP, so Resource. 00:05:13.360 --> 00:05:15.360 You're adopting a HR management system. 00:05:15.759 --> 00:05:18.000 The difficulty is the technology. 00:05:18.079 --> 00:05:19.680 You need to learn how to use. 00:05:19.920 --> 00:05:21.680 You need to learn which buttons to press. 00:05:21.839 --> 00:05:23.600 You need to learn how it works. 00:05:23.839 --> 00:05:29.600 And that technology is something that people have feel a lot of stress to learn about. 00:05:30.079 --> 00:05:31.839 They feel attached to their ways. 00:05:31.920 --> 00:05:33.360 They don't want to learn anything. 00:05:33.839 --> 00:05:39.920 And so there's a lot of tests and stress and learning, which is involved in change management. 00:05:43.600 --> 00:05:51.519 And so what happens is that the change management that leaders are used to has to do with the difficulty of learning. 00:05:51.600 --> 00:05:57.279 So and the hassle beliefs invest your staff members before associated. 00:05:57.519 --> 00:06:03.680 That's what has been in the past, but it's not the reality of AI. 00:06:03.920 --> 00:06:05.920 AI is actually much easier to learn. 00:06:06.240 --> 00:06:08.319 AI tools, you could ask them yourself. 00:06:08.399 --> 00:06:13.519 You could ask Claude, you can ask ChatGPT, you can ask Copilot, you can ask Gemini how to use it for your work. 00:06:14.079 --> 00:06:15.120 You can figure it out. 00:06:15.360 --> 00:06:18.480 I figured out, I figured out, other people have figured it out. 00:06:18.720 --> 00:06:23.680 It's actually not that hard, not nearly as hard as people make it out to be. 00:06:23.920 --> 00:06:26.560 And so these broad AI platforms are excellent. 00:06:26.639 --> 00:06:27.839 They're very flexible. 00:06:28.079 --> 00:06:31.680 You can use them to do nearly anything that you can do with a computer. 00:06:31.759 --> 00:06:33.600 So nearly all kinds of white collar. 00:06:34.000 --> 00:06:36.879 So that's not the block that people are experiencing. 00:06:37.199 --> 00:06:42.720 There are three big blockers that don't have an management, but it means technology. 00:06:42.959 --> 00:06:46.959 And so leaders are just making mistakes right and left, focusing on technology. 00:06:47.199 --> 00:06:48.319 It's about the psychology. 00:06:48.480 --> 00:06:50.560 So here are the three big blockers around. 00:06:51.199 --> 00:06:54.319 First big blocker is fear and anxiety. 00:06:54.639 --> 00:06:59.759 People aren't afraid of looking up information in a CRM. 00:06:59.920 --> 00:07:02.480 They feel it's a

lot of hassle to do that. 00:07:02.639 --> 00:07:04.879 It's annoying, but they're not afraid of it. 00:07:05.199 --> 00:07:08.720 People are afraid of training up AI to replace them. 00:07:09.120 --> 00:07:10.720 And so there's a lot of fear. 00:07:10.879 --> 00:07:24.639 And I talk about that in chapter six of my book, again, based on all this research, where I identify a specific psychographic profile of employees that I call the AI alarmists, for whom this fear is the primary emotional blocker. 00:07:24.800 --> 00:07:25.519 First one. 00:07:25.839 --> 00:07:30.000 Two, the second big blocker is identity. 00:07:30.240 --> 00:07:39.279 Now, people aren't frightened of their identity again by looking up new information in the CRM instead of looking up information at various Excel spreadsheets. 00:07:39.600 --> 00:07:43.839 But they're threatened by a tool which can write a great sales. 00:07:44.160 --> 00:07:53.120 They're threatened by a tool that can actually do a great marketing analysis or a financial report, create a great image, write a great article. 00:07:53.360 --> 00:07:57.040 Those things are threats to their professional identity. 00:07:57.360 --> 00:07:59.199 Let's not talk about identity. 00:07:59.519 --> 00:08:04.000 They replace things that people feel cloud, that people feel bring them meaning. 00:08:04.240 --> 00:08:11.839 And so the people who are less afraid for their jobs, but who feel an identity threat, I call them pragmatic resistors. 00:08:12.000 --> 00:08:14.319 So that's the second psychographic profile. 00:08:14.560 --> 00:08:20.879 And those two big psychographic profiles are what cause people to resist individual AI use. 00:08:21.360 --> 00:08:23.439 Now there's also a third blocker. 00:08:23.759 --> 00:08:28.560 This blocker has to do with a psychographic profile I call the reluctant authors. 00:08:28.959 --> 00:08:34.639 Now, these people, they use AI tools individually and extensibly. 00:08:34.879 --> 00:08:37.120 But they feel ashamed. 00:08:37.360 --> 00:08:39.279 So the shame is the biggest shoe. 00:08:39.360 --> 00:08:43.440 They feel ashamed of talking to their team members about how they use. 00:08:43.919 --> 00:08:54.320 They don't want to talk about how they're using these AI tools to draft their emails, to draft their marketing pitch, to actually go and do a financial analysis for them. 00:08:54.480 --> 00:08:55.759 They don't want to talk about it. 00:08:56.159 --> 00:08:58.559 They feel social sleep around. 00:08:58.799 --> 00:09:00.159 Now, show you an example. 00:09:00.320 --> 00:09:01.039 There's an example. 00:09:01.120 --> 00:09:05.440 Last week you did a training for a peer group, visit group. 00:09:05.519 --> 00:09:09.039 So it's invested by a group of leaders, and then they in its work. 00:09:09.200 --> 00:09:17.519 And so I was doing a training, and during the training, one of the executives said that told me a story about the AI document. 00:09:17.840 --> 00:09:28.320 So he's a CR of a selezable company, and they had the VP of sales present the biggest client they had in the past year. 00:09:28.559 --> 00:09:33.120 And all the context, how they got them, and all the goods that they need to present. 00:09:33.679 --> 00:09:38.639 So previous years took the VP of sales about three days to get the presentation. 00:09:38.879 --> 00:09:43.039 Conduct for the operation, do the PowerPoint and so on. 00:09:43.360 --> 00:09:47.919 This year, the VP of Sales gave it to his executive assistant. 00:09:48.159 --> 00:09:52.480 And the executive assistant put together an hour and a half in using the actions. 00:09:53.279 --> 00:09:55.200 And that's their presentation, which felt great. 00:09:55.279 --> 00:10:00.000 The VP of Sales was praised by the other executives to whom he presented it. 00:10:00.159 --> 00:10:09.360 The VP of Sales met with the CEO and told the CEO that, hey, it usually takes me about three days to this time. 00:10:13.519 --> 00:10:18.000 Don't tell any of the other executives because they'll think I'm cheating and I waste. 00:10:18.559 --> 00:10:23.360 And so that's the kind of culture that takes place at many companies. 00:10:23.519 --> 00:10:24.879 And that's at the top level. 00:10:25.120 --> 00:10:29.279 Of course, everyone else does the same thing and follows the same pattern. 00:10:29.519 --> 00:10:36.320 So this is a serious problem where it results in AI tools having individual adoption. 00:10:36.480 --> 00:10:46.480 And people are individually productive, but they don't tell their supervisors about them, they don't tell their team members, and they spend their individual time, additional time gone and phased person. 00:10:46.960 --> 00:10:50.480 And so the company doesn't see the benefit from the AI tools. 00:10:50.639 --> 00:11:02.480 And so for all of these reasons, all of these recent graphic profiles, the AI alarmists, the fear of anxiety, the identity fed with the pragmatic resistors, and the reluctant adopters with the shame. 00:11:02.799 --> 00:11:11.440 That's why you have the MIT study recently showing that something like 95% of AI islands don't show return on investment. 00:11:11.759 --> 00:11:13.600 And so that's a serious issue. 00:11:13.759 --> 00:11:20.559 And so these are the kinds of blockers that leaders are paying attention to because they're too focused on the technology and not nearly nothing. 00:11:21.279 --> 00:11:25.120 You mentioned the fear that exists in some. 00:11:25.919 --> 00:11:33.200 I also wonder whether the narrative that people hear about AI contributes to that fear. 00:11:33.360 --> 00:11:52.080 There isn't a week that goes by where Dario Amade of Anthropic or Sam Altman of OpenAI don't talk about how it is going to make the vast majority of knowledge workers redundant. 00:11:52.559 --> 00:11:56.080 So is there some logic behind that fear? 00:11:56.559 --> 00:12:01.360 Or is it that these people are particularly more fearsome and shouldn't be? 00:12:01.759 --> 00:12:03.279 Of

course there's logic behind the fear. 00:12:03.440 --> 00:12:11.440 We recently, just on August 12th of this year, we had an update to the Stanford Canaries and the Coal Mines things. 00:12:11.600 --> 00:12:16.240 It's a peer-reviewed study that look at what's happening with the job market. 00:12:16.720 --> 00:12:48.480 And what they found a year ago in June 2026 or June 2025, they found that AI-exposed occupations, in Bay exposed occupations, young people, 22 to 45 specifically, had 50% less employment ability compared to people in non-AI exposed occupations, young people that's thinking about then that's a year ago, that they updated the study, and now they found that the rate is 90%. 00:12:48.960 --> 00:12:49.759 So it's growing. 00:12:49.919 --> 00:12:51.600 Well, from 13% to 19%. 00:12:51.919 --> 00:12:54.159 But of course, then it's going to keep growing. 00:12:54.320 --> 00:13:00.480 And so right now, young people are the ones whose careers are being disrupted, and we're seeing that differently. 00:13:00.720 --> 00:13:19.600 Now, a recent letter, not from people like Dario Alabunde or Sam Oldman who have an interest in this topic, but from over 200 economists or objective observers, 36 Nobel laureates, they said that in the next several years there will be massive AI-caused job displays. 00:13:19.919 --> 00:13:21.039 People have reasons. 00:13:21.120 --> 00:13:25.120 And of course, we have all these headlines talking about tech companies that are laying people off. 00:13:25.279 --> 00:13:28.720 Definitely people have reasons to be afraid. 00:13:28.960 --> 00:13:33.759 Especially right now, young people have reasons, but everyone has reasons to be afraid. 00:13:33.919 --> 00:13:39.120 So in order to address these reasons, leaders can take concrete steps. 00:13:39.279 --> 00:13:42.799 And so here is what I see works for AI alarmists. 00:13:43.039 --> 00:13:45.360 We can talk about the other two groups later. 00:13:45.519 --> 00:13:49.759 But for AI alarmists, what works is two things. 00:13:49.919 --> 00:13:57.200 So first is to say that, hey, we're not going to fire anyone because we're going to get more productivity due to AI. 00:13:57.600 --> 00:14:00.639 Instead, what we're going to do is focus on growth. 00:14:00.879 --> 00:14:03.600 So not cost-cutting, but growth. 00:14:03.840 --> 00:14:16.480 And we have recent peer-reviewed research came out just a couple of months ago showing that companies that are adopting AI effectively compared to other companies in the same industry are growing 10% quicker. 00:14:16.720 --> 00:14:19.039 10% quicker head count-wise. 00:14:19.440 --> 00:14:24.399 They're taking market share away basically from companies that are effectively adopting AI. 00:14:24.559 --> 00:14:32.879 Now those companies are also, of course, growing top line and bottom line revenue faster than they're growing their head. 00:14:33.440 --> 00:14:39.440 So they need less people to do the same amount of work, and that's showing up in

the results. 00:14:39.679 --> 00:14:41.200 But they're not firing. 00:14:41.279 --> 00:14:42.320 And so it's happening. 00:14:42.480 --> 00:14:43.600 Where are people being fired? 00:14:43.759 --> 00:14:45.039 Where are people being laid off? 00:14:45.200 --> 00:14:50.399 They're being laid off, but companies aren't adopting beta effectively and that are losing market share. 00:14:50.639 --> 00:14:54.879 So it's not the companies that are adopting effectively that they're losing people. 00:14:55.200 --> 00:14:56.240 People are being fired. 00:14:56.320 --> 00:14:59.440 It's the companies that aren't adopting better effectively that are being well. 00:14:59.679 --> 00:15:00.240 So that's one. 00:15:00.399 --> 00:15:02.000 You can make commitments to grow. 00:15:02.320 --> 00:15:09.519 And so one thing that I strongly encourage leaders of companies I work with to make commitments to is that focusing on growth rather than cost-creating. 00:15:09.759 --> 00:15:16.879 And let's commitment that the large majority of companies can make, not everyone, but that's certainly something that I strongly encourage. 00:15:17.360 --> 00:15:36.080 Two is something that all companies can commit to, which is that if you find employees, if you use AI effectively, if you learn how to use AI effectively, and if you build tools that make you more effective, your job will be safe because you will be much more productive and effective. 00:15:36.320 --> 00:15:39.360 And so that's something that all companies can and should make. 00:15:40.000 --> 00:15:43.440 So you want to flip the script for these AI belong lists. 00:15:43.759 --> 00:15:55.840 Instead of thinking that AI adoption means that you will lose your job, you need to have them understand that non-adopting AI at the company level and at the individual level will mean jobs. 00:15:56.000 --> 00:15:57.840 So talking about the company-wide level. 00:16:37.039 --> 00:16:48.720 I think now everyone that listens to this podcast and anyone in the working environment knows that they're most likely not going to be spending their career in one organization. 00:16:48.960 --> 00:17:01.200 Therefore, having marketable skills becomes critical, and this provides an opportunity for people to develop more of those marketable skills. 00:17:01.440 --> 00:17:06.480 So that addresses some of the fear that people have in the organization. 00:17:06.640 --> 00:17:12.480 Now, you also mentioned the second one is identity, and it is a challenge. 00:17:12.640 --> 00:17:16.079 I know you also mentioned a case study of a law firm as well. 00:17:16.240 --> 00:17:33.039 I had conversations with law firms in the DC region where some attorneys are somewhat hesitant, not because they don't trust the AI, not because they don't think the output is good. 00:17:33.440 --> 00:17:48.880 Because in the back of their mind, they say, wait a minute, now in a couple of minutes, I'm getting done what would have taken me and my junior associates hours or days.

00:17:49.519 --> 00:17:52.240 What is the value I bring to it? 00:17:52.480 --> 00:18:07.359 The second part of it is the way they have monetized the value that they bring to clients, which has been on an hourly basis in many law firms measuring every six minutes and billing it to a specific client. 00:18:07.599 --> 00:18:25.200 So how can you tackle that identity when so much of the work that is being done both contributes to the sense of value the person believes they're bringing in, on the other end, to the way they are compensated based on how they bill clients? 00:18:26.079 --> 00:18:30.319 So when I work with law firms, one of the conversations changing the billing to clients. 00:18:30.400 --> 00:18:33.200 So I've talked about that specific one before I'm getting the back identity. 00:18:33.839 --> 00:18:37.920 And what I strongly advise them to do is change to value-based billing. 00:18:38.079 --> 00:18:48.960 So to the extent that they can make packages and change to a package-based billing system so that they're going to be aligned with client priorities. 00:18:49.119 --> 00:19:03.039 And so they can still fulfill their work and they can do their function and they can use AI effectively, and the client will be charged overall less than they would have been if they're on working on an hourly basis on without AI. 00:19:03.200 --> 00:19:05.039 So that's in terms of billing. 00:19:05.119 --> 00:19:08.559 We could talk about that more in Probably Professional Services. 00:19:08.880 --> 00:19:23.920 Now, going more deeply to the identity credit issue, what I see as really important is not focusing on using AI for the things that people feel attached to and what gives them meaning in their job. 00:19:24.160 --> 00:19:36.640 So instead of doing what might be the most effective thing from the company perspective, like asking having a survey of, oh, what are your most high-valued tasks that we can automate? 00:19:36.960 --> 00:19:38.880 Instead, you need to have a survey. 00:19:38.960 --> 00:19:46.799 So when I start working with a law firm or any other company, I do a survey on, hey, what are the tasks you eat the most? 00:19:47.119 --> 00:19:54.160 Why do you have to look up a whole bunch of information from a variety of systems to put together to do the stuff that you want to do? 00:19:54.400 --> 00:19:57.440 What do you find really annoying about your daily? 00:19:58.000 --> 00:20:07.519 And so then using those tasks, I prepare a library of prompts and agents that people could then be trained on. 00:20:07.680 --> 00:20:19.680 And I train people on those prompts and those agents which are focused specifically on the tasks that they find the most annoying, the most frustrating, where they gather information from all these systems to do what they really want. 00:20:20.240 --> 00:20:24.960 Now, this is not in the short term the most productive thing for the company. 00:20:25.279 --> 00:20:30.720 But that's where you have to have the psychological

judo move on going slow to go fast. 00:20:30.960 --> 00:20:33.200 Because then people will actually adopt these. 00:20:33.599 --> 00:20:37.359 They want to not be doing things that are frustrating and annoying. 00:20:37.440 --> 00:20:39.839 They want an AI tool to do it for. 00:20:40.079 --> 00:20:42.720 And the AI tools are incredibly adaptive. 00:20:42.960 --> 00:20:47.200 You could adapt them to do the things that people love to do or the things that people hate to do. 00:20:47.359 --> 00:20:57.519 And if you start with the things that people hate to do, people will then, let's say, in their psychology, they will then adopt these tools and they will start using these tools. 00:20:57.759 --> 00:21:07.039 Now, eventually, just once they learn how to use these tools and how victim they are, of course they'll be using them for other things, including things that they're excited about. 00:21:07.279 --> 00:21:11.599 But in the short term, you want to go slow to go fast. 00:21:11.759 --> 00:21:18.480 You want to have those tools especially and specifically aligned with things that people need to do. 00:21:18.640 --> 00:21:21.039 And that's how you address the identity threat. 00:21:21.279 --> 00:21:24.480 That's how you address the spragmatic resistors and get them on bolt. 00:21:25.039 --> 00:21:51.200 I love your approach, Gleb, because most specifically, you are centering it on how to meet the needs of the people within the organization who are a critical stakeholder in this process, rather than just jumping to the outcomes or the tools that can produce those outcomes, which a lot of organizations do. 00:21:51.440 --> 00:21:53.279 I am very outcome focused. 00:21:53.440 --> 00:21:54.319 I recommend that. 00:21:54.559 --> 00:21:59.759 However, that's also why then people get in the way of. 00:22:00.160 --> 00:22:02.960 Of the organization actually achieving that outcome. 00:22:03.279 --> 00:22:07.200 You also mentioned the reluctant adopters. 00:22:07.440 --> 00:22:13.599 There's a significant use of shadow AI in organizations. 00:22:14.240 --> 00:22:16.880 Sometimes it bubbles up to the surface. 00:22:17.119 --> 00:22:35.359 I was speaking to a CEO about a month and a half back, and he said there is almost not a single week that goes by when someone presents a brilliant strategic set of PowerPoints to me where when I ask them a question, they have no idea what's behind it. 00:22:35.519 --> 00:22:35.839 Yeah. 00:22:36.160 --> 00:22:40.000 Because the AI tools provide the opportunity to create those. 00:22:40.319 --> 00:22:51.359 So how have you seen that approached by organizations where people are more willing to embrace it and share the best practices with each other? 00:22:51.839 --> 00:22:55.119 So the first step is leadership modeling. 00:22:55.359 --> 00:23:15.440 So instead of the CEO agreeing with that executive, what I talked about in the group is that the CEO should have told the executive that, hey, what you need to do is march back out

to the other executives and tell them that you used AI to create and you saved yourself so much. 00:23:15.599 --> 00:23:18.400 And of course, the executive knows all the strategy behind it. 00:23:18.559 --> 00:23:22.400 So it's not like that situation where you present the strategy and don't know it. 00:23:22.559 --> 00:23:23.839 He knows it in and out. 00:23:24.000 --> 00:23:27.279 He was the one who did the problem. 00:23:27.839 --> 00:23:31.359 He was the one who got the client and so on, managed the process. 00:23:31.680 --> 00:23:46.640 But it's much better to use his time to go out and get this year's next biggest client rather than spend three days on creating this PowerPoint presentation to present to these other folks, the other executives. 00:23:46.880 --> 00:23:55.759 So leadership modeling, and that needs to be a conversation that the VP of Sales has with the other executives and with all members of his team and be proud. 00:23:56.240 --> 00:23:58.240 So focusing on leadership tasks. 00:23:58.319 --> 00:24:02.559 Like the thing that leaders, for example, hate to do is performance evaluation. 00:24:02.720 --> 00:24:38.160 So instead of having the leader just think about what the last model was of somebody's performance and evaluate them based on that, which is realistically what happens in the annual performance evaluation, a much more effective task is having the AI look through the past year of deliverables by someone and then create a port based on those deliverables, which then the executive combines with everything that's not in the email and other systems that the AI has access to, and then creates the performance evaluation. 00:24:38.400 --> 00:24:41.119 So that is much more effective. 00:24:41.279 --> 00:25:00.559 And so that's another example of where the leader should talk to his team members or her team members about how they use the AI tool to do the performance evaluation and how it's great for everyone that they did, because now they can have a more objective and less biased performance evaluation that's weighted toward the last month of performance, right? 00:25:00.799 --> 00:25:02.240 And so that's great. 00:25:02.319 --> 00:25:03.599 And the leader needs to map. 00:25:03.759 --> 00:25:05.519 So that's the first leadership model. 00:25:06.240 --> 00:25:07.039 That's one. 00:25:07.359 --> 00:25:09.440 Two, I talked about the trade. 00:25:09.759 --> 00:25:15.359 So you want to do the training based on the things that people hate most. 00:25:15.599 --> 00:25:24.960 And that provides a basis for everyone agreeing to change the social norm to do at least those things using AI tools. 00:25:25.200 --> 00:25:35.599 And so you already get that change of base, the baseline of at least these activities will do using AI tools. 00:25:35.759 --> 00:25:44.400 And so now your people could talk about it because we have training, and everyone would agree that yes, that would be good to use the AI tools to do those things that we can eat.

00:25:44.720 --> 00:25:46.880 Let's do and three. 00:25:47.119 --> 00:25:55.440 You need to set up an asynchronous communication channel where everyone shares about new users they're using for AI. 00:25:55.680 --> 00:26:01.279 A Teams channel, a Slack channel, other venues, show card where you share about it. 00:26:01.440 --> 00:26:06.559 Now it's great if you could have a synchronous as well, meaning like a weekly meeting where you talk about it. 00:26:06.720 --> 00:26:10.079 But I recognize that's not something that all companies can do. 00:26:10.240 --> 00:26:13.359 So synchronous if possible, but at least asynchronous. 00:26:13.519 --> 00:26:15.039 So something everyone can do. 00:26:15.279 --> 00:26:26.160 Where the leader needs to actively praise and reward people who come up with innovations that save them time or make get them to be more effective in their work. 00:26:26.400 --> 00:26:28.079 So needs active praise. 00:26:28.400 --> 00:26:36.799 And ideally, it's a kind of reward bonus or something like that, if this is the tool that other people can use for their own. 00:26:37.759 --> 00:26:40.240 So this creates culture. 00:26:40.400 --> 00:26:48.880 You're shifting up culture to making it more acceptable to use AI for a wide variety of cases that weren't covered in the initial trick. 00:26:49.200 --> 00:27:05.039 So that's the pre-combination things that help address the reluctant adopters and help address sense of shame, leadership modeling, and getting over that shame and stigma there, the believership, the believer feels, and sharing it with others. 00:27:05.279 --> 00:27:12.960 Then the training, and finally asynchronous, deal with synchronous, but at least asynchronous communication, whether you're warned and praise from the leader. 00:27:13.440 --> 00:27:17.599 That's a great way of setting the right culture within the organization. 00:27:17.920 --> 00:27:28.400 Now, one of the things you also mention that I have seen is that middle managers are critical to this process. 00:27:28.640 --> 00:27:41.440 How have you seen middle managers play the most effective role in maintaining this AI experimentation and use as part of the culture of the organization? 00:27:42.160 --> 00:27:49.839 All three of these cases, middle managers need to be the ones who transmit and model these issues to all of their employees. 00:27:50.079 --> 00:27:58.079 So the executive might the course for the AI artists telling it's about communication, right? 00:27:58.319 --> 00:28:01.839 Telling them that, hey, will I be fired at anyone because we're using AI? 00:28:02.240 --> 00:28:11.680 We'll be focusing on growth, and anyone who uses AI effectively, they will be the ones who will be promoted and they'll be the ones who will be successful. 00:28:12.000 --> 00:28:16.880 But the team members, the staffers, need to really hear it from their supervisors. 00:28:17.039 --> 00:28:20.400 So they hear that safe messaging needs to be consistent. 00:28:20.559 --> 00:28:22.480 And so that's going to be really important. 00:28:22.880 --> 00:28:24.960 Then, two, the trainings. 00:28:25.200 --> 00:28:33.359 The training is going to come first, the middle managers need to be trained on using AI tools, and they need to train their teams and using AI tools. 00:28:33.440 --> 00:28:37.200 And so that's going to address the clegmatic resistors. 00:28:37.519 --> 00:28:42.720 And finally, for the shaman stigma, that modeling needs to come, of course, from middle managers. 00:28:42.799 --> 00:28:46.400 So executives will be modeling for everyone. 00:28:46.640 --> 00:28:52.000 But the middle managers will be real role models for their staffers. 00:28:52.160 --> 00:29:02.960 And they will be the ones who, in that Slack channel or Teams channel, they'll be praising, rewarding their staffers for AI use cases. 00:29:03.119 --> 00:29:14.000 So the middle managers need to be doing all the things of AI adoption, at least separately from their own AI skills that they need to develop, which we talk about separately. 00:29:14.160 --> 00:29:17.359 But you're asking about those three problematic those three profiles. 00:29:17.440 --> 00:29:23.440 So that's what they need to do regarding those three profiles to ensure effective AI adoption for a company. 00:29:23.920 --> 00:29:25.599 I appreciate that, Gleb. 00:29:25.680 --> 00:29:33.839 And you share a lot more thoughts with respect to how to get AI adoption within the organization. 00:29:34.079 --> 00:29:48.559 Now, is it possible for AI to make individuals in the organization more productive while weakening their ability to determine whether the output is any good or not? 00:29:49.039 --> 00:29:51.039 I think there are a couple of things that are going on. 00:29:51.279 --> 00:30:09.839 So one thing, this was published in a recent Harvard Business Review article for research by a company called Betherop, found that something like two hours a month is wasted by colleagues fixing problematic output from their colleagues. 00:30:10.079 --> 00:30:12.799 And that's what they call AI work slaw. 00:30:12.960 --> 00:30:16.480 So two hours per month and the$296 on average per company. 00:30:16.559 --> 00:30:18.319 So it obviously adds up. 00:30:18.559 --> 00:30:20.160 Now, what happens? 00:30:20.319 --> 00:30:22.079 Why is there AI work slaw? 00:30:22.240 --> 00:30:23.039 What's the problem? 00:30:23.200 --> 00:30:24.000 What's the cause? 00:30:24.240 --> 00:30:26.720 One is malicious compliance. 00:30:27.039 --> 00:30:31.519 So you have people who are the photic resistors and the AI alarmists. 00:30:31.759 --> 00:30:33.759 They don't want to use AI. 00:30:34.079 --> 00:30:39.839 And if they're really forced to use AI, they use it in a malicious compliance fashion. 00:30:40.079 --> 00:30:44.480 Meaning they just press the AI button and then they get the output. 00:30:44.559 --> 00:30:49.599 They don't check it, they don't evaluate it, they just send it downward, and they're like, okay, I

did what you told me to do. 00:30:49.680 --> 00:30:51.200 I use the AI tool. 00:30:51.440 --> 00:30:54.400 And then you get what you pay for with these people. 00:30:54.960 --> 00:30:59.119 And so that's the results, this AI works like. 00:30:59.440 --> 00:31:02.000 So that results in a lot of problems. 00:31:02.240 --> 00:31:14.079 And that malicious compliance is a major reason why the 95% figure they took out, where 95% of AI pilots don't scale, they don't show it to our animals. 00:31:14.240 --> 00:31:14.960 So that's one thing. 00:31:15.279 --> 00:31:22.319 Now, the second problem is that the young people who are coming up through drinks, they're not taking footprint. 00:31:22.640 --> 00:31:28.160 They're teaching these young people is not focusing on the right things. 00:31:28.400 --> 00:31:35.200 Because the future of the workplace is not doing the initial generation of content. 00:31:36.079 --> 00:31:37.519 It's called generative AI. 00:31:37.599 --> 00:31:45.440 It's able to generate reports, it's able to generate writing, it's able to generate financial analysis, images, whatever you want it to do. 00:31:45.680 --> 00:31:52.559 The skill of the future is going to be evaluating the production of AI tools. 00:31:52.720 --> 00:31:54.160 And they're not being taught that. 00:31:54.240 --> 00:31:56.240 They're just being taught the initial generation. 00:31:56.559 --> 00:32:01.839 And that's very problematic because that's not the future of these skill sets. 00:32:02.000 --> 00:32:09.920 So that has to do with how young people are being taught and trained, even within companies, but especially in the educational system. 00:32:10.160 --> 00:32:21.839 So this is a big problem in the educational system and even within apprenticeships, because within apprenticeships, you know, they're still being taught in the old ways, they're not really being taught in how to use AI tools. 00:32:21.920 --> 00:32:27.440 So right now, what you're seeing is that AI tools are creating content fast and fast. 00:32:27.839 --> 00:32:33.599 Young people don't know how to evaluate the outcome of AI tools very effectively. 00:32:33.759 --> 00:32:42.400 And because previously they would have been evaluating the content of the tool during the process of creation as they're creating the content. 00:32:42.799 --> 00:32:51.680 But they don't have nearly as much experience in evaluating the outcomes you and I do, because we've seen outcomes that are good, outcomes that are bad. 00:32:51.839 --> 00:32:53.759 They've seen many fewer outcomes. 00:32:54.000 --> 00:32:59.599 And so they're not taught the skills of evaluating the outcomes, which is what they really should be taught to do. 00:32:59.839 --> 00:33:01.759 And so that is the second type of problem. 00:33:02.559 --> 00:33:11.680 Is that something that is trained that you can go in a new form of college or training for the company and gain? 00:33:11.920 --> 00:33:17.039 Or is it something that comes based on repeated exposure and experience? 00:33:17.359 --> 00:33:18.720 Of course, it could be trained. 00:33:18.960 --> 00:33:21.359 Right now, they're being trained on creating. 00:33:21.440 --> 00:33:25.759 But um, I think that that is not something that is going to be useful in the future. 00:33:26.000 --> 00:33:35.920 Instead, what they need to be trained on is here's the ideal final output, here's the context, here's how you edit it, here's how you use the AI tool to create it and edit it. 00:33:36.079 --> 00:33:37.359 Of course, it can be trained. 00:33:37.599 --> 00:33:42.960 And certainly something that people can be trained on, but they're not currently being trained on. 00:33:43.039 --> 00:33:47.839 The educational system is very much behind, much further than companies, on AI adoption. 00:33:48.000 --> 00:33:58.720 So I think this is something that leaders need to integrate into their onboarding programs and into their educational programs into their apprenticeships within a company. 00:33:58.880 --> 00:34:04.000 Right now, the onboarding and apprenticeship programs, they're not well set up either. 00:34:04.160 --> 00:34:10.000 So even companies are not really integrating AI into their apprenticeship and onboarding. 00:34:10.079 --> 00:34:11.760 And this needs to be something companies. 00:34:12.320 --> 00:34:19.280 Do you also argue, Gleb, that there is no final state of AI maturity? 00:34:19.519 --> 00:34:30.639 So if there is no stable destination, in your view, what organizational and leadership capabilities should we be trying to build? 00:34:31.199 --> 00:34:41.039 Well, it's prosaic to say we need to build continuous learning, but we definitely need to focus on continuous learning much more with AI tools than with previous technology. 00:34:41.199 --> 00:35:01.679 Because the future, it's not here right now, but the future in the next several years, here is something I can very definitively forecast because this is something that I'm very much seeing with companies that are at the forefront of AI adoption, is every individual staff member who will be at the company will not be working on individual tasks. 00:35:01.840 --> 00:35:06.559 Instead, they'll be managing agents who are working with individual tasks. 00:35:06.800 --> 00:35:10.880 So every individual will be a supervisor of AI agents. 00:35:11.039 --> 00:35:19.679 And they will be building AI agents customized to their specific workflows, which replace the individual tasks that they're doing right now. 00:35:19.920 --> 00:35:24.960 And so that is the skill of the future that people really need to learn. 00:35:25.280 --> 00:35:38.320 Managers, by contrast, will be managing both a set of agents that they're using themselves, like they talked about, performance evaluate agent, and they will be managing people who are managing agents. 00:35:38.480 --> 00:35:48.559 And so they need to be able to ensure that those agents are doing handoffs from one employee to another effectively, that they're

coordinating with each other effectively. 00:35:48.719 --> 00:35:52.000 That's a different skill that again managers currently have. 00:35:52.239 --> 00:35:56.480 And so managers, middle managers, need to learn those skills as well. 00:35:56.639 --> 00:36:02.559 And so there's a set of skills that employees need to learn, a set of skills that all managers need to learn. 00:36:02.800 --> 00:36:09.119 And as agents get better and better, people need to learn how to improve their current agents. 00:36:09.360 --> 00:36:23.039 Managers also need to learn how to improve their current agents and how to make sure their employees are learning about the kind of agents, the key, the capabilities and about increasing agents so that they're becoming more and more productive over time. 00:36:23.280 --> 00:36:26.559 So that's going to be the future in the next few years. 00:36:26.800 --> 00:36:32.719 And this is something that people need to focus on if they want to be successful in our increasingly AI disrupted world. 00:36:33.199 --> 00:36:34.559 It is incredible, Gleb. 00:36:34.719 --> 00:36:53.199 And we could spend hours talking about how managing an AI agent and how you ask for outcomes and set boundaries for an AI agent would be dramatically different than managing an individual that has a certain level of judgment. 00:36:53.440 --> 00:36:55.599 There's a lot that goes into that. 00:36:55.760 --> 00:37:10.559 So it's almost like a new set of management capabilities and skills that both individuals that have AI agents working for them, or the managers that have AI agents as well need to learn. 00:37:10.800 --> 00:37:18.400 Now, one other thing I wanted to touch on, Gleb, is there's so much with respect to AI that you can be doing. 00:37:18.559 --> 00:37:24.079 And one of the main things that I hear from people is like, I'm overwhelmed. 00:37:24.320 --> 00:37:27.199 I can understand how it can be overwhelming. 00:37:27.519 --> 00:37:41.679 So, in your view, how should leaders decide where they should lead, where they should experiment, where they should follow quickly, where they should wait, and how should they determine that? 00:37:42.400 --> 00:37:49.840 There's no reason I brought up the performance evaluation as an example, because that's a task that when they talk to leaders, it's the one that they hate the most. 00:37:50.079 --> 00:37:55.280 It's the one that they feel the most annoyed by, it's the one that they feel they want to get off their plate as much as us. 00:37:55.599 --> 00:37:58.159 So best of it's supposed to just focusing on it. 00:37:58.320 --> 00:38:02.159 Again, it has to do with psychology, it has to do with human motivation. 00:38:02.320 --> 00:38:08.079 You'll be most motivated yourself to focus on the work on the tasks that you hate the most. 00:38:08.239 --> 00:38:10.000 You will want to get that off your plate. 00:38:10.079 --> 00:38:12.079 Your employees will want to get that off your place. 00:38:12.400 --> 00:38:13.920 And so that's what you should focus on. 00:38:14.159 --> 00:38:22.239 Look at the tasks that you hate the most that you want to get off your plate, because AI can do pretty much anything that you can do with the computer. 00:38:22.480 --> 00:38:30.400 And so you want to use AI for the task, the biggest hassle and the biggest trouble in your day. 00:38:30.559 --> 00:38:32.559 And so focus on those tasks. 00:38:32.719 --> 00:38:36.639 Focus on getting rid of those tasks, focus on getting them off your plate. 00:38:36.880 --> 00:38:45.519 And once you know how to use AI tools effectively, you'll be able to make your own judgment about the other tasks, the ones that bring the highest value. 00:38:45.760 --> 00:38:52.159 But you want to start and focus early onward on those tasks that you hate, that your people have. 00:38:52.559 --> 00:38:54.960 And that way you'll go slow to go fast. 00:38:55.119 --> 00:38:56.719 Because remember, it's not about the technology. 00:38:56.960 --> 00:38:58.079 Technology is great. 00:38:58.239 --> 00:38:59.599 It's about the technology. 00:38:59.920 --> 00:39:02.159 That's a brilliant way of putting it, Gleb. 00:39:02.239 --> 00:39:11.840 And it connects to a Gallup survey I saw today, which at first can sound counterintuitive until we think about the point that you made. 00:39:12.239 --> 00:39:33.360 Where, as compared to organizations that haven't been as AI forward, the AI forward organizations seem to have both people's satisfaction had gone up and their dissatisfaction had gone up on average. 00:39:33.599 --> 00:39:36.639 And the reason is this was average across different companies. 00:39:36.800 --> 00:39:55.280 I would submit if we dig into the data, we would see that the ones that have done what you've talked about, which is approached it as a people-centric focus, their people are happier as a result of AI implementation. 00:39:55.519 --> 00:40:04.800 The ones that have purely focused on productivity for the benefit of the organization, the dissatisfaction has gone up. 00:40:05.039 --> 00:40:14.639 So I love the perspective you bring to it, which is a great way to bring people along and help them want to embrace this change. 00:40:14.880 --> 00:40:20.639 So, Gleb, for the audience to find your book and follow your work, where would you send them to? 00:40:21.119 --> 00:40:24.719 If you want to find me, the easiest way is LinkedIn. 00:40:24.800 --> 00:40:26.400 So I'm very available there. 00:40:26.559 --> 00:40:36.320 Just make sure to tell me when you reach out to me that you heard me on the partner leadership podcast because I get way too much spam and I'm not getting accepted otherwise. 00:40:36.480 --> 00:40:44.559 Now, for my book and my brother work, go to my website, designsteravoidance experts.com forward slash AI book. 00:40:44.800 --> 00:40:47.280 You'll find the for the book itself. 00:40:47.440 --> 00:40:52.400 You can get the Barnes Noble at Amazon, books, five bookstores everywhere, as people say. 00:40:52.559 --> 00:40:58.719 So it's out now, George Town University Press, Theater View, like I said, visually published, high quality. 00:40:59.039 --> 00:41:05.920 Or if you want to get a free introduction and a chapter of the book, you can go to the website that I mentioned before. 00:41:06.159 --> 00:41:10.320 Disasteravoidanceexperts.com forward slash AI book. 00:41:10.480 --> 00:41:13.360 So get book sample with a directional chapter. 00:41:13.599 --> 00:41:17.440 Disasteravoidance experts.com forward slash AI book. 00:41:17.599 --> 00:41:34.320 And if you already got the copy of the book pre-ordered or ordered, just put the receipt number from any bookstore into the website, and you'll get a free assist on adopting AI in your company and a manual on the civic mistakes that leaders make in AI adoptions. 00:41:34.559 --> 00:41:40.800 Again, that's going to be at disasteravoidance experts.com forward slash AI book. 00:41:41.599 --> 00:41:50.559 I appreciate the conversation with you, Gleb, and your book, The Psychology of AI Adoption at Work from Resistance to Results. 00:41:50.719 --> 00:41:52.639 Thank you so much for joining me, Dr. 00:41:52.719 --> 00:41:54.320 Gleb Zapersky. 00:41:54.960 --> 00:41:57.280 Thank you again, Van Han, for having me on the camp. 00:41:57.360 --> 00:41:59.039 It's been a pleasure to have another conversation. 00:42:00.800 --> 00:42:05.039 You have been listening to partnering leadership with your host, Mahan Tavakoli. 00:42:05.679 --> 00:42:21.280 If you enjoyed this episode, please leave a rating and review of the podcast on your favorite podcasting app and forward the conversation to a friend or colleague so you can help more people discover their purpose, grow professionally with meaning, and have a greater impact. 00:42:21.519 --> 00:42:27.679 For additional leadership insights and bonus content, visit us at partneringleadership.com.