E115 - What Must We Keep Human? - Podcast With Dr David Rock - Part Two

WEBVTT
00:00:00.160 --> 00:00:04.160
Look, the central question here is which skills are you okay to lose?
00:00:04.320 --> 00:00:06.240
That's what you have to be really clear about.
00:00:06.400 --> 00:00:11.359
And and know that using AI even as a helper, you will lose those skills.
00:00:11.519 --> 00:00:14.480
But the question is, whose brain is in the game?
00:00:14.720 --> 00:00:16.239
It's a bit like who's driving.
00:00:16.399 --> 00:00:19.920
Who's driving and who's like co-pilot just checking?
00:00:27.679 --> 00:00:29.679
Welcome to the inner game of change.
00:00:29.760 --> 00:00:31.839
I am your host, Ali Jumma.
00:00:32.399 --> 00:00:35.039
This is part two of my conversation with Dr.
00:00:35.119 --> 00:00:36.240
David Rock.
00:00:36.479 --> 00:00:50.320
In part one, we explored what continuous change does to the brain and why artificial intelligence can activate our sense of threat when many organizations are already nearing the limits of their capacity for change.
00:00:50.640 --> 00:00:53.679
Part two is anchored by one question.
00:00:53.920 --> 00:00:55.840
Who is doing the thinking?
00:00:56.320 --> 00:01:05.599
As AI becomes part of how we write, decide, and relate to one another, which capabilities are we prepared to lose and which must we protect?
00:01:05.920 --> 00:01:12.079
David challenges the advice that everyone should use artificial intelligence as much as possible.
00:01:12.319 --> 00:01:27.359
We discuss how it can increase confidence while weakening judgment, why feedback from a machine can feel easier to receive, and what happens when leaders use agreeable systems to navigate difficult to human relationships?
00:01:27.680 --> 00:01:29.359
The distinction is simple.
00:01:29.680 --> 00:01:34.560
Artificial intelligence should help us think more deeply, not think less.
00:01:35.439 --> 00:01:38.640
Here's part two of my conversation with Dr.
00:01:38.719 --> 00:01:39.680
David Rock.
00:01:43.200 --> 00:01:44.400
Here's the real problem.
00:01:44.719 --> 00:01:46.560
And the study just came out from BCG.
00:01:46.640 --> 00:01:54.079
Do you know what the one of the first skills people lose, one of the biggest skills people lose, is actually judgment of when to use AI or not.
00:01:54.719 --> 00:01:54.959
Yes.
00:01:55.200 --> 00:01:58.719
And judgment of like problem definition and all this kind of stuff.
00:01:58.879 --> 00:02:05.280
In other words, the more you use these things, the more your brain's going to default to using these things more and more and more.
00:02:05.439 --> 00:02:06.719
And you might say, well, who cares?
00:02:06.799 --> 00:02:07.439
Just keep using it.
00:02:07.519 --> 00:02:08.400
You know, it'll always be there.
00:02:08.479 --> 00:02:15.520
It's like, does it matter that you lose the ability to, you know, do complex math calculators are always around, right?
00:02:15.599 --> 00:02:16.800
We'll always have AI around.
00:02:17.039 --> 00:02:18.719
You know, the world's not going to explode, hopefully.
00:02:18.879 --> 00:02:19.039
Yes.
00:02:19.199 --> 00:02:20.719
There's an argument for that.
00:02:21.439 --> 00:02:30.879
But again, the problem here with that is with a small exception of people, you feel much more confident in your work.
00:02:31.120 --> 00:02:33.919
You feel like the quality of work is going up.
00:02:34.159 --> 00:02:39.280
You clear the speed of work is going up, all while your accuracy is plummeting.
00:02:40.719 --> 00:02:46.560
And in fact, pretty sure these numbers are right, but don't like haul me over calls if I'm wrong.
00:02:46.639 --> 00:02:59.919
But it was something like people who are now using AI all the time, in the past, were wrong about a quarter of the time, and sorry, were right about a quarter of the time about something, and now they're right about 9% of the time.
00:03:00.080 --> 00:03:05.520
So about one in, you know, one in four, they're right about something before, you know, AI, now they're right one in ten times.
00:03:05.599 --> 00:03:08.560
But they actually feel more confident in their decisions now.
00:03:08.800 --> 00:03:15.919
They feel like they're actually smarter and more confident and faster, but they're actually much more incorrect, much more often.
00:03:16.319 --> 00:03:21.039
Not to mention the the works really by definition average.
00:03:21.360 --> 00:03:21.680
Right?
00:03:21.840 --> 00:03:26.960
By definition, it's average work that every other person in every other company is getting the same answers.
00:03:27.120 --> 00:03:38.240
So if if you don't learn to like harness this thing properly, you are getting dumber, don't know it, doing even more average work, and probably eventually will be replaced.
00:03:39.039 --> 00:03:42.479
Because what are you adding if you're doing the same as everyone else is doing?
00:03:42.800 --> 00:03:45.199
You're not advocating that we don't use AI.
00:03:45.280 --> 00:03:49.759
You're advocating that we gotta look at this critically.
00:03:50.080 --> 00:03:54.080
I want to share with you something that I've been experimenting with myself.
00:03:54.639 --> 00:03:58.960
I got to the stage where I've been I was using AI 24-7.
00:03:59.680 --> 00:04:01.360
And then I did What'd you find?
00:04:01.439 --> 00:04:02.000
How was that?
00:04:02.319 --> 00:04:03.520
Good things and bad things.
00:04:04.080 --> 00:04:09.360
Good things that AI can help me, can give me feedback loop.
00:04:09.759 --> 00:04:13.280
Basically critiquing my ideas, that's good.
00:04:13.599 --> 00:04:16.959
I learned, so that's my cognitive ability, expands.
00:04:17.120 --> 00:04:18.480
That's really good for me.
00:04:19.360 --> 00:04:28.240
But also I discovered that there will be times, as you mentioned, when I don't use it, I feel like I'm in pain, like I need to use it.
00:04:28.480 --> 00:04:33.040
So what I've started doing now is my schedule, and I'm experimenting with this.
00:04:33.360 --> 00:04:37.040
But I'd like to hear your neuroscience lens on this.
00:04:37.360 --> 00:04:44.879
I've created a AI time, a human time, and AI in human time.
00:04:45.040 --> 00:04:45.199
Okay.
00:04:46.000 --> 00:04:48.319
Based on the activities that I do.
00:04:49.360 --> 00:04:56.000
I like my human time, but it's not easy because I've been using AI for quite a while.
00:04:56.240 --> 00:04:59.759
But I'm going through a self-corrected state.
00:05:01.120 --> 00:05:05.199
What am I doing here right that can help me going forward?
00:05:06.399 --> 00:05:10.720
Look, the central question here is which skills are you okay to lose?
00:05:10.959 --> 00:05:11.279
Okay.
00:05:11.680 --> 00:05:13.680
That's what you have to be really clear about.
00:05:13.920 --> 00:05:19.120
And and and know that using AI even as a helper, you will lose those skills.
00:05:19.360 --> 00:05:25.519
If you unless you're just using it as a light helper, but you're doing them with like 95% of the work, and you're just getting a little bit of input.
00:05:25.680 --> 00:05:29.279
But the question is whose brain is in the game, right?
00:05:29.439 --> 00:05:31.120
Like it's a bit like who's driving.
00:05:31.360 --> 00:05:35.120
Who's driving and who's like co-pilot, just checking, right?
00:05:35.279 --> 00:05:36.319
I was driving with my mom.
00:05:36.560 --> 00:05:43.600
I mean, I you know, I grew up in Sydney, my mom's still alive there and with us, and and we were driving, and I was driving because I she's 84.
00:05:43.680 --> 00:05:45.120
I don't trust her driving anymore.
00:05:45.279 --> 00:05:46.399
But sorry, mum.
00:05:46.560 --> 00:05:50.800
But we're driving and we're really close to her house, but just taking a slightly different route.
00:05:50.879 --> 00:05:53.519
And she's been there for like 18 years in this house, right?
00:05:53.600 --> 00:05:55.360
And in the area for even longer.
00:05:55.600 --> 00:05:57.519
And we're completely lost.
00:05:57.759 --> 00:06:01.120
And I'm like, Mum, how do you not know where we are?
00:06:01.279 --> 00:06:02.879
Like, it was nighttime.
00:06:02.959 --> 00:06:05.040
She said, Well, it's nighttime, I don't really know.
00:06:05.199 --> 00:06:08.800
And and it's we're driving around, and at some point I'm like, Mom, we're literally on your street.
00:06:08.879 --> 00:06:12.480
Can you not see that we're actually on your street now, just further up?
00:06:12.639 --> 00:06:14.480
And she's like, Oh, I see that.
00:06:14.639 --> 00:06:18.240
Now, you know, she doesn't have dementia or major cognitive problems.
00:06:18.399 --> 00:06:29.439
What the problem was, and she said it to me, she said, Well, when I'm not driving, I don't hold the whole map of where we are and where we're going next and where we've got, like, I just don't have the map in my brain of where we're going and what we're doing.
00:06:29.680 --> 00:06:31.360
So it's just all kind of noise.
00:06:31.839 --> 00:06:37.600
And and this is what happens if you're not like driving, if you're not building the chunks in your brain, right?
00:06:37.759 --> 00:06:44.160
If you're not the one with the, if you don't have your brain in the game, right, then it's a whole different story.
00:06:44.319 --> 00:06:45.600
So whose brain is in the game?
00:06:45.680 --> 00:06:46.560
Who's doing the driving?
00:06:46.720 --> 00:06:46.959
Yes.
00:06:47.040 --> 00:06:47.199
Right?
00:06:47.360 --> 00:06:48.240
Who's doing the driving?
00:06:48.399 --> 00:06:50.959
Who's doing the sort of commenting?
00:06:51.439 --> 00:06:56.319
And if you're letting AI do the driving and you're commenting, you will lose those skills.
00:06:57.360 --> 00:07:01.199
And the the question is, which skills are you okay to lose?
00:07:01.360 --> 00:07:02.800
And the answer might be lots of them.
00:07:02.959 --> 00:07:03.920
Great, that's fine.
00:07:04.079 --> 00:07:05.199
But don't lose all of them.
00:07:05.439 --> 00:07:05.759
Yes.
00:07:05.920 --> 00:07:06.160
Right.
00:07:06.319 --> 00:07:07.759
And which skills do you not want to lose?
00:07:07.839 --> 00:07:14.319
You know, you probably don't want to lose the skill of, you know, building relationships with, you know, important new people.
00:07:14.480 --> 00:07:14.800
Right?
00:07:14.879 --> 00:07:24.399
You don't want to people to receive, you know, AI generated personal messages from you because they'll tell and they'll just they'll just shut you out.
00:07:24.560 --> 00:07:24.879
Right?
00:07:24.959 --> 00:07:32.319
You want to send even more personal, even more authentic communications to people that mat to people that matter to you, and don't use AI for that.
00:07:32.560 --> 00:07:32.959
Right?
00:07:33.120 --> 00:07:34.160
So as an example.
00:07:34.240 --> 00:07:36.720
So not telling you what to do, but just as an example.
00:07:37.040 --> 00:07:39.040
And I I actually like that, uh David.
00:07:39.199 --> 00:07:40.160
I am the driver.
00:07:40.319 --> 00:07:43.279
I think I picked up something from one of your webinars.
00:07:43.360 --> 00:07:45.519
You called it the human in the lead.
00:07:45.600 --> 00:07:47.040
And I really like that.
00:07:47.360 --> 00:07:51.279
You made me now, you just enlightened me that I am actually the driver.
00:07:51.600 --> 00:07:52.480
Yeah, you're the driver.
00:07:52.720 --> 00:07:56.399
I'm the driver of the interaction.
00:07:56.639 --> 00:07:56.959
Yep.
00:07:57.199 --> 00:08:03.360
And my generative AI or my AI models, they actually on the side.
00:08:03.439 --> 00:08:05.600
People actually helping you.
00:08:05.920 --> 00:08:10.959
Maybe give you some feedback, some nudges, but can can challenge you, but you're actually having the insights.
00:08:11.040 --> 00:08:11.439
This is the point.
00:08:11.759 --> 00:08:16.639
Because facing the music of my stakeholders who are actually human.
00:08:16.879 --> 00:08:19.920
And I need to understand that and I need to articulate it.
00:08:20.000 --> 00:08:26.319
And I need to continue protecting those cognitive abilities for me to actually look at things and think that does not make sense.
00:08:26.560 --> 00:08:26.800
Yeah.
00:08:26.879 --> 00:08:27.040
Yeah.
00:08:27.120 --> 00:08:32.240
So just to summarize, like be clear about which habits your and skills you're okay to lose.
00:08:32.399 --> 00:08:42.879
And the ones that you you don't want to lose, you can use AI, but you're having the insights, you're doing the problem solving, you're actually doing the processing, you're doing the majority of the build of anything.
00:08:42.960 --> 00:08:46.960
And you might get a little bit of feedback from AI, but you're doing the driving.
00:08:47.200 --> 00:08:49.360
And that distinction is really, really, really important.
00:08:49.440 --> 00:08:54.399
And people don't understand that enough that you know you're in the lead and the AI is in a slight support role.
00:08:54.480 --> 00:08:57.840
But it's it's so tempting, especially when AI pops up in every task we do.
00:08:58.000 --> 00:08:58.159
Yes.
00:08:58.320 --> 00:09:00.879
It's so tempting to let it do the majority of the work.
00:09:01.120 --> 00:09:04.960
And that's probably okay for things that you're okay to lose, right?
00:09:05.120 --> 00:09:06.559
The skills you're okay to lose.
00:09:06.639 --> 00:09:09.120
So just I just want people to be really thoughtful about that.
00:09:09.200 --> 00:09:13.039
And companies, I can't, I'm I'm kind of yelling at companies at the moment.
00:09:13.120 --> 00:09:19.120
I'm kind of yelling at boards because they're basically running around saying, we want everyone to use this as much as humanly possible.
00:09:19.360 --> 00:09:21.200
I'm like, stop saying that.
00:09:21.440 --> 00:09:22.559
That is not okay.
00:09:22.879 --> 00:09:29.759
So let's go deeper into that because what I'm observing is that we are asking people to experiment.
00:09:30.480 --> 00:09:34.720
But as an employer, I'm not gonna reap the benefits of those experimentations.
00:09:35.200 --> 00:09:39.039
And the other thing that and I do it in my practice too.
00:09:39.200 --> 00:09:42.399
I'm I'm I'm aware of that that particular challenge.
00:09:42.639 --> 00:09:51.279
I am helping people adopt a technology that we are not 100% sure how is gonna impact their their roles in the future.
00:09:51.600 --> 00:09:54.320
But I want to ask you a question around that this feedback.
00:09:54.480 --> 00:10:01.120
What does a neuroscience think of I find myself and I tried it with my son, teenager?
00:10:01.600 --> 00:10:08.080
He accepts feedback better from a chat GPT than me as his father.
00:10:08.559 --> 00:10:08.879
For sure.
00:10:09.039 --> 00:10:11.360
Probably from anyone more than his father, unfortunately.
00:10:11.440 --> 00:10:13.440
That's that's not a child relationship.
00:10:13.759 --> 00:10:14.639
That's a child relationship.
00:10:14.879 --> 00:10:15.519
That makes me feel worse.
00:10:16.000 --> 00:10:18.000
No, no, this is normal parent-child relationships.
00:10:18.240 --> 00:10:22.080
You're you're you're smacking like the middle of like what every parent has to deal with.
00:10:22.399 --> 00:10:22.799
It's not you.
00:10:23.039 --> 00:10:23.519
Yeah.
00:10:24.080 --> 00:10:34.159
But I I also find it that somehow, you know, my claw gives me feedback, and some of it is actually harsh, and I'm fine with it.
00:10:34.320 --> 00:10:41.679
What I find myself doing with generative AI is actually I focus on the core point rather than the emotions around it.
00:10:41.840 --> 00:10:48.320
But if it was a human doing exactly the same thing, I probably won't be able to completely detach from the emotion.
00:10:48.639 --> 00:10:50.159
Yeah, no, we study this exact thing.
00:10:50.240 --> 00:10:59.679
We we studied feedback like really, really robustly some time ago, but but I've been working on feedback literally since like the year 2000.
00:10:59.919 --> 00:11:03.840
We've been we've been like building and delivering like feedback trainings in companies.
00:11:04.000 --> 00:11:04.240
Yes.
00:11:04.399 --> 00:11:08.799
And so we've been thinking about like how to get this right and modeling it and all this stuff forever.
00:11:08.960 --> 00:11:11.200
Literally, literally 26 years.
00:11:11.440 --> 00:11:16.799
But we we had a real breakthrough in understanding how to how to create a feedback culture.
00:11:17.039 --> 00:11:19.360
We had a real breakthrough around 2015.
00:11:19.440 --> 00:11:32.399
We did a study in Australia, actually, in in a company in a professional services firm, and we collected real biodata, real biological data from about 70 or 80 participants in the study that we set up inside a company.
00:11:32.559 --> 00:11:36.320
And we we saw what happens when people give and receive feedback.
00:11:36.720 --> 00:11:40.240
And what was interesting is unsolicited versus solicited feedback.
00:11:40.399 --> 00:11:42.159
So asking for feedback.
00:11:42.320 --> 00:11:44.159
And we we found three things.
00:11:44.480 --> 00:11:53.039
Firstly, that receiving feedback is really stressful, no surprise there, but we were able to get it to a level, like similar to your heart rate's similar at like you've just gone for a run.
00:11:53.279 --> 00:11:54.960
So like receiving feedback is stressful.
00:11:55.039 --> 00:11:56.559
We knew that, but we confirmed it.
00:11:56.879 --> 00:12:05.600
What was surprising was the the giving feedback was just as or more stressful than receiving it, which is why people give it so badly.
00:12:05.840 --> 00:12:06.879
So that was surprising.
00:12:06.960 --> 00:12:16.240
So from the biological data, and this is in a fairly trivial task, the biological data said, look, giving feedback is even more stressful than receiving it, or at least as stressful, one of the two.
00:12:16.720 --> 00:12:19.120
We we didn't find it less stressful for people.
00:12:19.360 --> 00:12:20.399
So that was surprising.
00:12:20.559 --> 00:12:26.320
And then the really big surprise was that when people ask for feedback, it's about half as stressful for both sides.
00:12:27.600 --> 00:12:30.960
And we looked at that and we realized, well, look, that's a status thing.
00:12:31.440 --> 00:12:31.679
Right?
00:12:31.759 --> 00:12:33.919
When you ask, it's much less of a status threat.
00:12:34.080 --> 00:12:36.559
In fact, you look kind of like smart.
00:12:36.879 --> 00:12:37.759
In control, right?
00:12:38.159 --> 00:12:39.440
And then it's an autonomy thing.
00:12:39.519 --> 00:12:41.519
You're making the choice, right?
00:12:41.840 --> 00:12:46.720
So, so the the combination of those things, and also it feels a bit more certain, right?
00:12:46.879 --> 00:12:50.480
The combination of those things just turn down the threat a lot.
00:12:50.639 --> 00:13:08.799
And then we dug into the research and we realized that's really, really important because for feedback to work, you have to do a really cognitively taxing activity, which is basically imagine what you currently do and imagine a new way of doing that something, and compare and contrast and then make a decision which one's better, right?
00:13:08.879 --> 00:13:11.759
So there's a lot of like that's a lot of like mathematics.
00:13:11.840 --> 00:13:13.600
It's like long division in your head.
00:13:13.759 --> 00:13:19.759
It's like a lot of steps, picturing, holding in mind the way you do something, holding in mind a different way of doing it.
00:13:19.919 --> 00:13:32.960
And if your brain is really taxed because it's basically threatened, you you you're you you don't you just don't have the resources for it, or your brain is full of you know frustration at the person giving you feedback or defensiveness or whatever.
00:13:33.120 --> 00:13:40.559
So anyway, we realize that the status and autonomy threats are really big factor in feedback as to why it just doesn't work very much.
00:13:40.799 --> 00:13:42.559
And yeah, absolutely.
00:13:42.799 --> 00:13:48.399
If you can if you can deal with a digital tool, you're gonna have much less of a status threat, right?
00:13:48.559 --> 00:13:48.799
Yeah.
00:13:49.360 --> 00:13:50.799
Excuse me, so and and and autonomy.
00:13:50.879 --> 00:13:54.320
And if also autonomy, you're actually asking, again, you're asking for the feedback.
00:13:54.399 --> 00:14:08.000
So I think I love that uh idea of the autonomy because it gives me a space to have a little bit of or sense of control that I'm actually t taking a proactive approach in here.
00:14:08.240 --> 00:14:23.279
But I want to go back to you've written an article recently around the crisis in leadership using generative AI to do performance discussions, and some of them actually probably prepare for feedback as well.
00:14:24.559 --> 00:14:25.679
What is the crisis?
00:14:25.840 --> 00:14:27.039
What's the issue here?
00:14:27.360 --> 00:14:34.879
Because generative AI can help us prepare for performance discussions and also provide feedback as well.
00:14:35.279 --> 00:14:39.519
Look, it it depends in part on the kind of tool that you're using.
00:14:39.679 --> 00:14:48.639
The problem is using the mainstream tools like Claude, ChatGBT, Copilot, Gemini, but all of the main tools, you know, all the other ones as well.
00:14:49.200 --> 00:14:54.799
They're basically their business model is keeping you engaged and keeping you spending tokens, right?
00:14:54.960 --> 00:15:03.919
Their business model is not based on rewards for you, you know, getting smarter or getting off there quicker or having an insight or something like that.
00:15:04.000 --> 00:15:05.440
They're rewarded by keeping you engaged.
00:15:05.519 --> 00:15:08.080
And the way they keep you engaged is by agreeing with you.
00:15:08.320 --> 00:15:13.679
And lots has been written about the sycophantic nature of these tools, but it's very insidious.
00:15:13.759 --> 00:15:16.799
It's not just that it says, oh, you're amazing, that's an amazing idea.
00:15:16.960 --> 00:15:21.600
It's much more subtle that it basically just agrees with you when you say something, right?
00:15:21.759 --> 00:15:27.519
But it can be something as banal as you say, like, you know, that employee is really underperforming.
00:15:27.840 --> 00:15:31.120
And it just takes that as fact, right?
00:15:31.360 --> 00:15:33.919
Let's talk about their underperformance, right?
00:15:34.240 --> 00:15:38.879
It's not like, oh, what are some factors, you know, do you think you could be a reason why they're underperforming?
00:15:39.039 --> 00:15:41.039
Do you think it could be a system reason, right?
00:15:41.279 --> 00:15:43.360
It's like, no, they're underperforming, right?
00:15:43.440 --> 00:15:48.159
So it just takes your whatever you say, it takes at face value and agrees with you.
00:15:48.399 --> 00:15:49.600
Um and it's very subtle.
00:15:49.759 --> 00:15:55.360
And so, so it was actually a study done in science, which is one of the most prestigious, you know, academic journals in the world.
00:15:55.519 --> 00:16:10.320
And what they did is they, they, they step by step took apart what happens when average people, it wasn't focused on leaders, when average people use these mainstream AIs, they didn't call out any one particular one, for interpersonal challenges.
00:16:11.360 --> 00:16:11.679
Right?
00:16:11.919 --> 00:16:21.759
So I'm having an argument with this person, or I don't like what this person's doing, or I need to give this person some feedback, or any kind of interpersonal challenge.
00:16:22.000 --> 00:16:28.639
And what they found, and they they actually scripted, like they showed and they they showed like the scripts that that ended up happening, it was kind of terrifying.
00:16:28.720 --> 00:16:42.000
They showed that in every instance the AI, you know, took the person's view as correct, and there was just a spiral down to this person needs to be fired, this person needs to a performance improvement plan, this person.
00:16:42.559 --> 00:16:49.840
It wasn't so focused on on that, but it it it you could see it spiraling down to a conflict, right?
00:16:50.080 --> 00:16:56.639
So rather than like heading towards some kind of reconciliation, it was heading towards more and more escalation, right?
00:16:56.720 --> 00:17:07.039
And what you what you see now is people using, you know, gen AI to actually have like arguments with each other, but it's their AIs arguing, right?
00:17:07.200 --> 00:17:07.359
Yes.
00:17:07.519 --> 00:17:19.200
And instead of like going to the water cooler and saying, look, I'm sorry that happened, let's sort that out and and and you know, like kind of repairing, right, they're escalating because the other person's always wrong.
00:17:19.440 --> 00:17:25.039
And so they're using AIs to articulate in great detail all the reasons why the other person's wrong.
00:17:25.200 --> 00:17:29.200
And then the other person receives that and they respond with all the reasons that they're wrong.
00:17:29.440 --> 00:17:45.519
So this is what the science study saw is that when people use gen AI for interpersonal issues, not just conflicts, but conflicts and concerns and feedback, it it rapidly devolves into escalation instead of repair.
00:17:46.240 --> 00:17:56.319
Now, you could use an AI that understands that and understands that managers are half the, you know, are the problem half the time and does challenge, but the mainstream AIs don't do that, right?
00:17:56.640 --> 00:18:01.759
So so the piece we wrote in Fast Company is called the coming wave of toxic leadership.
00:18:01.920 --> 00:18:04.559
We just made a very simple extrapolation.
00:18:04.640 --> 00:18:06.559
Well, this is what happens for normal humans.
00:18:06.640 --> 00:18:13.359
What if leaders and managers use mainstream gen AI tools when they have a performance issue?
00:18:13.519 --> 00:18:14.319
What does that look like?
00:18:14.480 --> 00:18:19.839
And we looked, we we extrapolated and went, oh, that's gonna escalate almost every time.
00:18:19.920 --> 00:18:22.640
And and it's gonna create even more toxic leaders, basically.
00:18:22.720 --> 00:18:29.680
It's gonna create leaders where who who blame everyone else for every problem because Chat GPT told them to.
00:18:29.759 --> 00:18:29.920
Yeah.
00:18:30.160 --> 00:18:30.400
Right.
00:18:30.880 --> 00:18:33.839
And that's that was essentially the problem we saw.
00:18:34.000 --> 00:18:37.759
So so I think, look, if you're a physicist, you don't use Chat GPT for physics problems.
00:18:37.839 --> 00:18:40.960
You use something that's tuned to the way a physicist needs information.
00:18:41.200 --> 00:18:45.680
If you're a lawyer, you use a legal AI that's tuned to how lawyers need information.
00:18:45.759 --> 00:18:54.000
And if you're a manager or leader, you should be using a manager or leader AI that's tuned to how these people should receive information, right?
00:18:54.240 --> 00:19:02.240
Which actually involves like spotting biases, spotting when the leader is the problem, spotting narcissism, like all sorts of things, right?
00:19:02.400 --> 00:19:07.759
So, so they they kind of need a different sort of AI for people issues.
00:19:08.000 --> 00:19:08.240
Yeah.
00:19:08.480 --> 00:19:15.119
Sure, use copilot for processing things and challenging you on conceptual things, but oh, I'm having a people problem.
00:19:15.200 --> 00:19:18.319
Actually, let's switch, you know, let's switch to Niles, right?
00:19:18.400 --> 00:19:19.920
Niles is the AI we built.
00:19:20.000 --> 00:19:27.200
I'm sure there's others that can do this well, but you know, Niles understands the brain really, really deeply and builds a model of your brain.
00:19:27.279 --> 00:19:36.880
And if you are, if you're being biased, it'll it'll feel it, it'll feel it, but it'll it'll detect it as one of 30 different variables it's detecting every second.
00:19:36.960 --> 00:19:39.279
Um, it'll detect it and challenge you.
00:19:39.519 --> 00:19:41.920
And it knows that the leader is the problem a lot.
00:19:42.079 --> 00:19:50.799
And that if you've you know, if the last two or three times you've been talking about everyone having you know low performance, it's gonna start to question you and challenge you.
00:19:50.960 --> 00:20:05.279
So anyway, the the short answer is leaders and managers need a leader and manager AI that's tuned to the way they should be receiving information, just like a doctor, lawyer, accountant, and physicist, you know, are now using their own kinds of tools as well.
00:20:05.519 --> 00:20:07.519
How are you using Gen A to PI?
00:20:08.880 --> 00:20:15.039
You know, when I look at what I what skills I want to lose, I don't have my list is very short.
00:20:15.279 --> 00:20:16.720
I have a fantastic assistant.
00:20:16.799 --> 00:20:19.200
I've had her for 12 years, God bless her.
00:20:19.440 --> 00:20:21.440
She says I'm I'm the biggest diva.
00:20:21.680 --> 00:20:25.039
She's she's annoyed at me half the time, but she has amazing work.
00:20:25.200 --> 00:20:26.799
I'm super specific about what I need.
00:20:26.880 --> 00:20:33.519
But like over the years, you know, through feedback and and this, like she's really, really good at just like taking a lot of annoying things off me.
00:20:33.599 --> 00:20:38.400
So I'm not scheduling, I'm not mapping, traveling, I'm not, there's a lot of things I'm not doing.
00:20:38.640 --> 00:20:39.839
So that's fantastic.
00:20:39.920 --> 00:20:43.680
Now, if you want to train an AI, an agent to do some of that stuff, good luck.
00:20:43.759 --> 00:20:45.119
It's gonna take you a few years.
00:20:45.279 --> 00:20:49.680
Yeah, it's gonna make a ton of mistakes, be very careful, sandbox things.
00:20:49.839 --> 00:20:54.799
But you know, so I'm very lucky in that I've spent some time working with the same person.
00:20:55.119 --> 00:20:56.960
Bless her soul, Janelle's her name.
00:20:57.200 --> 00:20:59.279
And so so I don't do other things.
00:21:00.160 --> 00:21:03.839
Do you use AI to generate ideas, do some research?
00:21:04.079 --> 00:21:06.000
Well, the first thing is it's impossible to avoid it.
00:21:06.240 --> 00:21:06.559
Exactly.
00:21:06.720 --> 00:21:06.960
Right?
00:21:07.039 --> 00:21:10.480
If you're a conscientious objector, you've got to work really hard to not use it.
00:21:10.640 --> 00:21:12.960
So anytime I'm doing a Google search, I'm using AI now.
00:21:13.039 --> 00:21:14.000
It's coming up, right?
00:21:14.160 --> 00:21:20.960
And what I'm finding is I'm I'm doing Google searches that are much more like my company's a Google shop more than a Microsoft shop.
00:21:21.039 --> 00:21:24.240
So we use, you know, Gemini, you know, we have Gemini and all the Google things.
00:21:24.400 --> 00:21:27.519
So I'm it's default for me to just search.
00:21:27.759 --> 00:21:30.960
And I what I'm finding is I'm searching much more like a prompt.
00:21:31.039 --> 00:21:31.440
Yeah.
00:21:31.680 --> 00:21:34.480
So I'm prompting when I'm searching, right, much more.
00:21:34.559 --> 00:21:35.839
So I'm using it that way.
00:21:36.000 --> 00:21:44.319
I definitely use it when I'm writing, I use it to research, but I've had to be really, really careful and like double check and sometimes triple check things because it hallucinates a lot.
00:21:44.559 --> 00:21:47.039
Like 40% of the time, it's wrong, you know.
00:21:47.119 --> 00:21:48.799
So so I use it to research.
00:21:48.960 --> 00:21:52.319
When I'm writing something, I say right away from it, to be honest.
00:21:52.480 --> 00:21:53.440
I've tried.
00:21:53.759 --> 00:21:54.960
Boy, I've tried.
00:21:55.200 --> 00:21:59.359
I'm like, look, if this could help, make my writing better, but it just makes my writing worse every time.
00:21:59.519 --> 00:21:59.759
Yeah.
00:22:00.319 --> 00:22:05.519
And maybe like just take like a paragraph.
00:22:05.680 --> 00:22:09.680
Maybe once a week I'll take a paragraph and run it through and just say, is there a better way to see this?
00:22:09.759 --> 00:22:11.119
And I still have never used it.
00:22:11.279 --> 00:22:13.599
It's just like got me thinking slightly differently.
00:22:13.680 --> 00:22:16.319
But I'll use it for research quite a bit.
00:22:16.640 --> 00:22:23.200
I asked a question just the other day, how many of the top 10 companies in the world are technology now?
00:22:23.519 --> 00:22:26.160
It turns out it's like eight or nine out of ten now.
00:22:26.400 --> 00:22:27.920
Five years ago it was six or seven.
00:22:28.079 --> 00:22:29.440
Ten years ago it was two.
00:22:30.160 --> 00:22:30.400
Wow.
00:22:30.640 --> 00:22:34.640
So AI was great at telling me that, but I'm gonna have to double check that before I publish that.
00:22:34.880 --> 00:22:35.279
Of course.
00:22:35.440 --> 00:22:37.119
But that's that's what I found.
00:22:37.359 --> 00:22:40.559
So, you know, that would take hours to work out other ways.
00:22:40.640 --> 00:22:42.960
So I've done I use it for research.
00:22:43.359 --> 00:22:46.880
I tried using it for like idea development and sort of stuff.
00:22:46.960 --> 00:22:51.279
I just I found personally that it not very helpful.
00:22:51.440 --> 00:23:00.000
And what I see is that a lot of really senior people, like chief human resource officers, like the CHRO or chief people officer, they're feeling guilty about how little they're using it.
00:23:00.240 --> 00:23:06.720
And I want to encourage them and say that's probably a good thing because the skills you don't want to lose involve communicating to influence people.
00:23:07.039 --> 00:23:08.160
That's a lot of what you do.
00:23:08.240 --> 00:23:11.119
Don't lose that skill, don't offload that skill, right?
00:23:11.359 --> 00:23:14.640
Thinking strategically about really deep, complex things.
00:23:14.880 --> 00:23:16.319
Don't lose that skill.
00:23:16.480 --> 00:23:16.799
Yeah.
00:23:16.960 --> 00:23:17.359
Right.
00:23:17.839 --> 00:23:22.000
So there's a lot of things that a very, very senior person does that they shouldn't be losing.
00:23:22.799 --> 00:23:35.839
And this is an interesting question is that companies should be thinking about roughly the percentage of time or the percentage of work people do that AI, you know, could be helpful and what those things are and aren't.
00:23:36.079 --> 00:23:38.480
Because right now the message is use this as much as possible.
00:23:38.559 --> 00:23:43.680
But I think companies should be saying, look, if you're encoding, you should use this to develop original stuff, but test it.
00:23:43.759 --> 00:23:48.720
And it probably looks like half your work is is AI assisted, but half is like really not.
00:23:48.799 --> 00:23:50.480
You know, if you're in this, it looks like this.
00:23:50.559 --> 00:23:52.000
And if you're here, it looks like this.
00:23:52.240 --> 00:23:56.880
And and putting some some guardrails, like you're never to do XYZ.
00:23:57.279 --> 00:23:59.920
So I think companies need to be a lot more prescriptive than they are right now.
00:24:00.240 --> 00:24:02.240
Enhance my experimentation with myself.
00:24:02.480 --> 00:24:02.960
Good job.
00:24:03.200 --> 00:24:07.759
I'm I'm trying to, you know, put scaffolding around using it.
00:24:08.000 --> 00:24:09.119
My summary is good job.
00:24:09.440 --> 00:24:10.240
Keep doing that experience.
00:24:10.640 --> 00:24:13.519
It's it's well, I'm gathering my own insight, David.
00:24:13.599 --> 00:24:15.279
So I'm following your advice.
00:24:15.599 --> 00:24:19.039
I want to ask you one final question.
00:24:19.440 --> 00:24:30.640
What would be your advice to people like me working in the business of change and helping organizations now in this age where there's a lot of change happening?
00:24:31.039 --> 00:24:32.319
AI is in the mix.
00:24:32.480 --> 00:24:34.240
We just came out of COVID.
00:24:35.119 --> 00:24:41.119
What must be the things that we should be protecting now to help our clients?
00:24:42.319 --> 00:24:48.559
I mean, the last thing a company should do is get everyone using AI as much as possible because they'll be exactly the same as the next company.
00:24:48.799 --> 00:24:51.359
You know, what's your strategic advantage?
00:24:51.599 --> 00:24:54.720
Sometimes it's first mover advantage that doesn't last that long.
00:24:54.960 --> 00:25:05.599
Sometimes it's distribution, you know, like you've got a bit of a monopoly on distribution, you own, you know, those shelf space, like sometimes it's mind share, you know, you own that sort of mind share.
00:25:05.680 --> 00:25:22.720
So there's some moats, you know, physically and psychologically, but but you know, it it's your your people, there's this cliche, your people are your biggest asset, but your your people come up with original thoughts and through the culture you have, they they expand and evolve, right?
00:25:22.799 --> 00:25:25.519
And your culture determines how quickly people adapt.
00:25:25.759 --> 00:25:28.880
Your culture determines how much people innovate, right?
00:25:29.039 --> 00:25:32.079
Your culture will determine if people innovate or stagnate.
00:25:32.400 --> 00:25:35.759
You cannot innovate if everyone's using AI all the time.
00:25:35.920 --> 00:25:38.319
You'll come up with exactly the same as every other company.
00:25:38.799 --> 00:25:42.400
So your culture really matters, innovation really matters.
00:25:42.640 --> 00:25:44.720
These things are central for success.
00:25:44.880 --> 00:25:49.359
And just laying AI onto everything will diminish your culture and diminish your innovation.
00:25:49.519 --> 00:25:54.720
So it's just, it's just super important that people get the strategy right for using AI.
00:25:54.799 --> 00:26:10.720
And I think there's a pivot right now that companies have to make from everyone, please use this as much as possible all the time, to let's define the thoughtful usage so that we maintain our culture and so that we use the tool to maximize innovation, not minimize innovation, right?
00:26:10.799 --> 00:26:13.440
Not homogeneize, if there's such a word.
00:26:13.599 --> 00:26:16.880
We we don't make everything homogeneous with the company next to us.
00:26:17.279 --> 00:26:27.119
Talking to a big consulting firm, I'm like, look, if you, one of the big consulting firms, if you, everyone's using AI all the time, you're exactly the same as the next consulting company.
00:26:27.599 --> 00:26:28.400
You're exactly the same.
00:26:28.480 --> 00:26:34.319
You're doing the same work, you're coming up with the same ideas, people are going to stop paying you because they can just hire their own people to do that.
00:26:34.400 --> 00:26:34.720
Yeah.
00:26:35.200 --> 00:26:38.240
So it's it's important to get this pivot right.
00:26:38.480 --> 00:26:45.920
And as it relates to Gen AI in particular, less so for embedded AI, but for gen AI in particular, we have to teach people to use these tools well.
00:26:46.079 --> 00:26:52.960
Unfortunately, we've told people this is a tool to think less, and we have to change the story and say this is a tool to think deeper.
00:26:53.119 --> 00:26:55.759
And and there'll be lots of benefits to that to you.
00:26:56.000 --> 00:26:58.480
But this is a tool to think deeper, not to think less.
00:26:58.640 --> 00:27:01.359
And let's, you know, let's get on to the hard work of doing that.
00:27:01.680 --> 00:27:05.039
That is a great place to finish today's conversation.
00:27:05.200 --> 00:27:07.359
I am grateful for your time, David.
00:27:07.680 --> 00:27:09.039
Thank you so much for your time.
00:27:09.119 --> 00:27:11.039
And I hope I can get you back again.
00:27:11.279 --> 00:27:13.039
Thanks so much and have another conversation.
00:27:13.279 --> 00:27:14.480
Yeah, thanks so much for the opportunity.
00:27:14.640 --> 00:27:16.559
I appreciate the chance to share about this.
00:27:16.799 --> 00:27:17.680
Thank you.
00:27:21.599 --> 00:27:24.400
Thank you for listening to the inner game of change.
00:27:24.559 --> 00:27:35.039
If there is one thing you have learned through those conversations, it is that change is never just something that happens around us, it is something we participate in.
00:27:35.519 --> 00:27:42.480
Every conversation, every experiment, and every new perspective helps us navigate it a little better.
00:27:42.799 --> 00:27:50.400
And remember, one of the best ways to have a sense of control over any change is by simply embracing it.
00:27:50.720 --> 00:27:58.000
Until next time, stay curious, stay open, and keep playing your inner game of change.