The Human, Ethical & Tech Dimensions of AI-Driven Engagement

WEBVTT
00:00:04.799 --> 00:00:14.960
Hey everyone, I'm Nicholas Kontopoulos, Vice President of Marketing for Twilio, and I look after APJ, and I'm genuinely really excited to be here again today to decode trust with you.
00:00:15.119 --> 00:00:22.480
Now, a recent study published in May suggested that AIs, particularly ChatGPT, can hold its own in a debate.
00:00:22.640 --> 00:00:24.960
I really found that fascinating, actually.
00:00:25.199 --> 00:00:35.920
And it basically is stemming from its ability to really tailor its arguments in a more analytical, structured way that really adapts to each individual.
00:00:36.159 --> 00:00:41.280
It's making points that really resonate with the person and it's speaking to.
00:00:41.520 --> 00:00:43.439
I mean, that sounds pretty cool, right?
00:00:43.679 --> 00:00:46.079
But that should also make us a little bit uncomfortable.
00:00:46.240 --> 00:01:00.399
Today I'm joined by Natalie Nahai, who will spend the next hour or so exploring the topic of trust in the age of AI, what it means for us as individuals, and what organizations should be thinking about in its ethical applications of this technology.
00:01:00.560 --> 00:01:02.000
So again, buckle up.
00:01:02.079 --> 00:01:03.520
We're going to have a great conversation.
00:01:03.600 --> 00:01:06.799
And I'm looking forward to exploring this topic with you, Natalie.
00:01:06.879 --> 00:01:12.159
But before we kick start, can you share with the audience a little bit more about who you are and what you're passionate about?
00:01:12.400 --> 00:01:13.519
Sure, it's lovely to see you again.
00:01:13.680 --> 00:01:15.760
Last time we had a chat, it was very stimulating.
00:01:15.840 --> 00:01:17.280
So this is a treat for me.
00:01:17.840 --> 00:01:27.200
So my background broadly is at the intersection of persuasive tech, which is any kind of technology that influences our decision making, our behavior, our relationships.
00:01:27.359 --> 00:01:33.760
So it could be everything from apps and even things like marketing and websites to obviously artificial intelligence.
00:01:34.079 --> 00:01:36.719
And then also the human behavior side.
00:01:36.879 --> 00:01:40.959
So psychology, looking also increasingly at society, ethics, culture.
00:01:41.120 --> 00:01:58.640
So as we move into more persuasive, ubiquitous technology, so AI becoming part of not just business life, but daily life, I'm particularly interested in how we can broaden and deepen conversations around it to give us back a sense of capacity, agency, critical thinking.
00:01:58.719 --> 00:02:10.080
So when we're using these tools, we're using them we're using them in a way that supports the qualities we'd like to enhance rather than undermining our potential for a flourishing life.
00:02:10.479 --> 00:02:23.759
No, and um, yeah, you like you referenced our previous conversation, and and just for those that uh weren't in attendance, we had Signal last year in Singapore, and Natalie and I um had a fantastic conversation.
00:02:23.840 --> 00:02:27.680
And I'm a little bit starstruck because she's an amazing thinker in this space.
00:02:27.840 --> 00:02:41.840
So I just want to thank you for taking the time out and joining us uh on today's episode because you know the goal of this uh series is really to share insights with, you know, a lot of folks who are deeply passionate about, you know, building trust.
00:02:41.919 --> 00:02:45.599
And ultimately, you know, I see trust as the bedrock of customer experience.
00:02:45.680 --> 00:02:51.919
And the role of AI is playing a bit, there's a huge role in AI in terms of helping us build that trust.
00:02:52.000 --> 00:03:03.360
But that you know, there's some things we need to be considering when we venture into using AI and how we apply it and ultimately how do we ensure we are building trust and maintaining that trust with our customers.
00:03:03.439 --> 00:03:05.680
So I'm really looking forward to exploring this with you.
00:03:05.840 --> 00:03:18.240
And I think, you know, it's fair to say that, you know, everyone that's dialing in today, you know, me, yourself, everyone is going to be impacted by AI in one shape or form, whether we like it or not.
00:03:18.400 --> 00:03:24.719
You know, for instance, many companies are using AI, you know, or data to train their AI.
00:03:24.960 --> 00:03:41.280
And it really does ultimately rely on data that we're sharing publicly, um, certainly in terms of the evolution of the LLMs that are powering the AI capabilities today, have been really mining the data that we've been sharing through our different social feeds, our consumption patterns, et cetera, and beyond.
00:03:41.439 --> 00:03:45.439
And it's increasingly going to influence our perceptions.
00:03:45.599 --> 00:03:57.199
And that means it's going to be, you know, difficult time for brands when it comes to building trust as consumers begin to become more aware of this and what it potentially means for them.
00:03:57.360 --> 00:04:05.680
So I'd love for you to maybe um, you know, share your views on some of the nuances when it comes to the topic of trust in the age of AI.
00:04:06.400 --> 00:04:08.319
This is such a big question, isn't it?
00:04:08.719 --> 00:04:10.240
I thought I'd start with one.
00:04:10.719 --> 00:04:11.280
I know, right?
00:04:11.439 --> 00:04:12.479
In at the deep end.
00:04:12.639 --> 00:04:17.040
So there's lots of there's lots of things we could talk about um in terms of trust.
00:04:17.279 --> 00:04:46.399
I think one of the things to start by saying is that the reasons, one of the biggest reasons that human civilization has been so successful is our extraordinary capacity for collaboration, which from a historical perspective rests quite heavily on our capacity to empathize with other people, whether that's, you know, perspective taking, so that cognitive structure of empathy, putting ourselves in someone else's shoes, or whether it's emotional empathy.
00:04:46.480 --> 00:04:58.160
So that kind of, you know, you see a kid crying, you feel that sadness, you pick them up, or that kind of somatic empathy, that kind of physical quality of, you know, if you see someone really hurt themselves, you feel an echo in your body.
00:04:58.319 --> 00:05:14.720
So the first thing to say is that empathy as a structure, or those three separate structures, um, is fundamental to us being able to collaborate, build trust, build rapport, organize on a huge level, and then be able to make decisions as larger collective groups.
00:05:14.959 --> 00:05:33.120
Now, when we're thinking about AI, and you mentioned at the top of this conversation, the ability for AI to start to debate or to tailor communications in a way that's really convincing, much of the time the capacity for large language models to do that is rooted in their ability to pattern match.
00:05:33.279 --> 00:05:36.480
So using phrases or obviously it's through tokens.
00:05:36.560 --> 00:05:50.240
So that the next probable word, by using language that we find convincing, we then project empathy or human-like qualities onto that system because it's performing in a very convincing way, human-to-human empathy.
00:05:50.319 --> 00:05:52.399
And this is called the ELISA effect.
00:05:52.480 --> 00:05:56.959
And it's a fundamental principle that's really useful to know about before unpacking trust further.
00:05:57.040 --> 00:06:06.800
And it's the idea that we infer these human qualities like understanding and subjectivity and empathy and feeling onto these models.
00:06:06.959 --> 00:06:38.480
And so if machines can do this well, even very simple ones like the one back from uh 1964 in MIT's AI Lab, which was a very early precursor to what we have today, if it can do that well, make us think that it understands us, then we are much more likely to trust that chatbot and therefore divulge personal information, which even with the best warnings in the world, you know, Chat GPT or Gemini or whatever, don't give away your private information, it doesn't matter because we've we've evolved socially to want to give information to that which we trust.
00:06:38.639 --> 00:06:48.720
So I think that's one of the biggest things to think about when we're talking about any kind of trust-building exercise or whether or not to trust platforms that emulate human communication.
00:06:48.879 --> 00:06:55.600
Um, and then there are questions around okay, when we're talking about trust, are we talking about uh within a professional context?
00:06:55.680 --> 00:07:00.480
So lawyers that might use AI that then get thrown out in court cases, like we saw recently.
00:07:00.639 --> 00:07:12.319
There's a case of um a guy called Mike Lindell, who was the MyPillow CEO, and two of his attorneys basically got fined for defamation because of using AI to cite court cases that didn't exist.
00:07:12.399 --> 00:07:23.439
So like there are very real examples where within professional contexts, trusting your tool too much or not doing the relevant checks can really cost you physically as well as reputationally, as well as in terms of profit.
00:07:23.600 --> 00:07:42.240
And then there's a flip side where you think about things like people becoming over-reliant and intimate and credulous of their AI, where you end up with these sorts of delusions that we've seen recently and some pretty high-profile examples of people feeling like they've been given some sort of keys to the universe through the AI.
00:07:42.319 --> 00:07:47.600
So, like, I'll pause there because I don't want to be sort of giving too much of an introduction.
00:07:48.480 --> 00:07:50.639
I think I'd I love where you started there.
00:07:50.720 --> 00:07:54.959
Uh and again, like I wanted to start with the human side of things really.
00:07:55.040 --> 00:08:07.839
I think, more often not, through people um processing technology, if I think of it in terms of business context and driving transformational change, you always want to start on the human side, the P, you know, and really think about trust through that prism.
00:08:07.920 --> 00:08:10.240
And I love where you led out with in terms of empathy.
00:08:10.399 --> 00:08:12.560
So just I just want to double-click on that.
00:08:12.720 --> 00:08:26.480
So is there the ability to apply uh empathy in your application of AI, in your deployment of AI, I guess, in terms of how it supports people, whether it's on the client side or employee side?
00:08:27.279 --> 00:08:28.720
I mean, I think so.
00:08:28.800 --> 00:08:31.360
This is where it gets into very ambiguous territory.
00:08:31.600 --> 00:09:20.960
From what I have seen in research looking at reducing hallucination rates, which is actually a fundamental question when it comes to trusting anything, um, if you're using a closed system that is built on proprietary data that you yourself as a company have acquired, it's clean data, it's well labeled, it's as transparent as it can possibly be, and you design a system within a closed framework, closed garden, that you can then overlay or instruct to be able to converse in an empathetic way, then that seems to carry a lot lower risk of unintended costly consequences than using large models that we all know, um, whose recent uh updates have actually had higher incidences of hallucinations than their predecessors.
00:09:21.120 --> 00:09:29.200
So I think there is a case to be made that if you have all these other structural elements in place, then yes, you could absolutely design more empathetic AI.
00:09:29.360 --> 00:09:44.399
The only issue then is thinking about the, I suppose, not the only issue, but one of the issues, how it impacts the person's interaction with the AI, whether they then start to disclose information you may not want them to.
00:09:44.799 --> 00:09:51.759
That implicates privacy, uh, it implicates things like person-to-person interactions within organizations.
00:09:51.919 --> 00:10:04.320
So organizational culture can then come into play in terms of does it undermine people's ability to have more messy communications with peers that they might have a sense of greater friction with?
00:10:04.399 --> 00:10:06.000
And we're starting to see that too.
00:10:06.240 --> 00:10:08.879
Um also over-reliance on AI tools.
00:10:08.960 --> 00:10:17.039
If we have a frictionless, empathetic, positive experience with an AI tool that is then constantly there for us.
00:10:17.279 --> 00:10:30.879
Um are we more likely to give our tasks to that tool and reduce our own critical thinking, our own capacity for discernment, our own engagement with creative ideation just because it's so easy.
00:10:31.039 --> 00:10:33.039
And then actually it's kind of striking that balance.
00:10:33.120 --> 00:10:34.799
And that's a very hard thing to measure, right?
00:10:34.879 --> 00:10:44.879
Because in the short term, productivity might well go up, performance might go up, you might end up with teams that are outsourcing stuff to AI because they feel like they're understood and it's convenient.
00:10:45.120 --> 00:10:56.320
But over the long term, maybe the quality of um the output goes down because people have become, you know, there's this idea of cognitive debt, which came out in a paper recently.
00:10:56.559 --> 00:11:02.320
If you sort of upfront the costs, then later your thinking muscles get flabby, for want of a better metaphor.
00:11:02.480 --> 00:11:06.159
So there's lots of other things that then kind of, you know, one has to consider.
00:11:07.039 --> 00:11:25.360
I I love where you're going, and it's also going to be a nice little segue into sort of the the next area I wanted to explore because one of those areas that I do see um uh area of you know consideration for businesses is gonna be in particular younger talent coming through, right?
00:11:25.440 --> 00:11:35.440
As they come into the business, typically um they would have, you know, trial and error would have been part of the learning process, that interaction with other humans, face-to-face.
00:11:35.600 --> 00:11:40.639
Um, as we move to work from home models, ultimately we're now also using AI.
00:11:40.720 --> 00:11:54.399
I can see those having a dramatic impact in terms of people's ability maybe to build that critical thinking capabilities that you ideally want to see come through their their experience as they build that experience.
00:11:54.559 --> 00:12:09.039
So a recent study from Pearl.com conducted um you know census uh through uh census wide in December last year revealed that 41% of uh Gen Z trust AI more than humans.
00:12:09.279 --> 00:12:22.159
So there is certain uh is certainly a bit of hype around the capabilities of Gen I AI in the workplace that might have skewed uh the extent to in to which its accuracy can be trusted.
00:12:22.320 --> 00:12:24.639
I think you touched on a little bit of this already.
00:12:24.879 --> 00:12:31.360
But how do we know when to trust and when to take pause to examine that information that we're being fed?
00:12:31.519 --> 00:12:37.759
Um, you know, again, I this is an area I think is really important, in particular to the what I was just sort of alluding to.
00:12:37.919 --> 00:13:01.759
Um, because as we bring talent through, I I do sometimes worry, and I've got two sons, you know, that are 15 and 17, you know, I don't want them to over-rely on technology because obviously that I think will diminish their own ability to develop critical thinking, but also at the same time, there's significant um uh advantages to uh incorporating AI into the mix, right?
00:13:01.840 --> 00:13:04.960
I've seen that, I'm sure you've seen it in terms of what it can do for us.
00:13:05.039 --> 00:13:05.279
Yeah.
00:13:05.360 --> 00:13:07.759
So what are your thoughts there in terms of the gen AIs?
00:13:07.919 --> 00:13:21.600
Are you seeing any sort of um early signs of, you know, um, you know, positive impacts it's having for them or any potential for areas that we need to be thinking about as we as we bring that young talent through?
00:13:22.720 --> 00:13:33.679
So one of the things I always find uh as an interesting exercise is look at the decisions being made about education by the people who've designed the actual tools themselves.
00:13:33.919 --> 00:13:49.759
So here, for instance, if you think about the people who are now coming of age, or coming of age, who are in their parenting age in Silicon Valley, they're sending their kids to Montessori schools, they're not giving them screens, they're not giving them uh chatbots to work with.
00:13:50.080 --> 00:13:50.720
Why?
00:13:50.960 --> 00:14:21.759
And we can we can infer all kinds of reasons why, but when we look at data around the increased levels of screen time, increased levels of social media time, um, increased contact with AI chatbots, which have significant, all of which carry significant risks when overused, um, we can see that actually there's all sorts of reasons why we might want to put guardrails around and and also best education around how, when, and why we use AI.
00:14:21.919 --> 00:14:46.000
So you mentioned um Gen Z and how actually when people come in, I think this is like implied, do you uh challenge me if I've misunderstood, but when people come into an organization, let's say um a legal firm and their associates and they are being really driven hard to work super long hours, it's stressful, people are gonna have meltdowns, like it's just it's gonna be a really awful environment at times in which to work.
00:14:46.159 --> 00:14:47.279
I'm not endorsing that.
00:14:47.519 --> 00:15:04.799
However, on the flip side, it's part of that organizational context that allows the environment for these young people to be able to take on heavy caseloads, to develop grit, resilience, to manage stress.
00:15:05.200 --> 00:15:12.720
And if we outsource all of that or we diminish that to a certain extent, I mean, I do think we need systemic change to make work environments healthier.
00:15:12.879 --> 00:15:14.159
That is without a doubt.
00:15:14.399 --> 00:15:37.279
But beyond that, if you outsource those um early experiences, or you you reduce access to those early experiences by saying to people, well, just run it through our, you know, AI system, those people are not gonna learn the skills they need to be able to qualify the output to manage complicated cases or difficult relationships with clients, what have you.
00:15:37.360 --> 00:15:46.960
And we saw there was a couple of years ago, it was um an article in the FT where a big legal firm was saying, actually, we're not letting any of our associates use AI because they're gonna be the lawyers of tomorrow.
00:15:47.039 --> 00:15:53.919
And if we do that, we're cannibalizing our potential to earn and to serve our clients in the long term because they need to have this training.
00:15:54.080 --> 00:16:01.120
So there's that aspect, and I think there are other aspects as well to take into account, thinking about the Pearl.com study that you suggested.
00:16:01.279 --> 00:16:08.159
Um, in it there was an Ipsos study that was carried out in the UK early this year that found it was something, it was nearly half, nearly 50%.
00:16:08.399 --> 00:16:14.480
I think it's like 47, 48% of people aged around, I want to say 18 to 21.
00:16:14.559 --> 00:16:19.600
So it's like young adults who said they wish they had been raised without the internet.
00:16:20.320 --> 00:16:27.919
Now, if you stack that, obviously there's the value action gap, like we say one thing, but then many of us have to be online because there's social network effects.
00:16:28.000 --> 00:16:32.480
If we're not there, we can't stay in touch with our friends, look for jobs on LinkedIn, all the rest of it.
00:16:32.559 --> 00:16:37.200
So there are many reasons why what we say we want and what we actually do might be different.
00:16:37.519 --> 00:16:52.879
But the kind of a lot of arrows are pointing towards the fact that the way that we're using our screens, technology, AI, social media is causing tremendous harm to relationships, mental health, critical thinking faculties.
00:16:53.039 --> 00:17:01.600
And so we have to put in guardrails for people coming into organizations where they absolutely get to use AI, but they're doing it in a way where they're testing how good this platform is.
00:17:01.840 --> 00:17:05.279
Does it work for solving X and Y problem?
00:17:05.759 --> 00:17:28.160
And so thinking about how companies can, you know, include this kind of more critical, more robust, more useful approach to using and testing AI, having two-week sprints, bringing young people in, go, right, how would you test these different suites of AI products to see if the solution you're getting is actually better than what you could do with other tools or in a team?
00:17:28.319 --> 00:17:34.880
Um, or for instance, I heard recently it was a fascinating conversation about the use of AI to write essays in universities.
00:17:35.039 --> 00:17:37.599
And one professor got up and she said, you know what?
00:17:37.759 --> 00:17:43.359
I tell my students, take whatever platform you're using, get it to write your essay.
00:17:43.519 --> 00:17:47.599
And tomorrow we all come into class and we critique the essays it's written.
00:17:47.759 --> 00:17:50.480
And that's it's brilliant because they're going to use it anyway.
00:17:50.720 --> 00:18:02.160
The whole thing is to kind of flex the muscle and get used to used to critical thinking and analyzing the output so that you're not saying to people, don't use it because it's here, but you are getting them to use it in a way that they're they're kind of learning its limitations.
00:18:02.400 --> 00:18:02.880
I love that.
00:18:02.960 --> 00:18:04.319
That's actually a really smart way.
00:18:04.640 --> 00:18:05.200
So smart.
00:18:05.359 --> 00:18:05.920
Oh, 100%.
00:18:06.240 --> 00:18:11.599
I think it's a really clever way of um um flipping, yeah, flipping it around.
00:18:11.680 --> 00:18:22.400
Uh what what I what I also loved about what you were sharing was actually, you know, without giving Twilio a plug, but you know, I have we have got a fantastic leadership team that enables us to get our teams together.
00:18:22.480 --> 00:18:25.119
So we actually regular get our teams together.
00:18:25.200 --> 00:18:32.319
Like just yesterday, there was about 50 of us in a co-working space, and we had a lot of different sessions running and a lot of interaction.
00:18:32.400 --> 00:18:33.920
And it's awesome, right?
00:18:34.000 --> 00:18:40.480
Because we we've also got access to the latest and greatest in tools and technologies that we're using, and AI tools are part of that mix.
00:18:40.720 --> 00:18:52.799
But I think the human connection element and being able to connect with your peers and whether it's through a Zoom session but or in person, that's when you get the ideas really sparking and and and taking shape.
00:18:52.960 --> 00:19:00.960
And to your point, that example you just pointed to, you may have used AI to generate the first draft of a document, but then we have that discussion and debate around that.
00:19:01.039 --> 00:19:04.079
So that's why that struck struck a call to me, what you just said there.
00:19:04.240 --> 00:19:13.839
So yeah, I think uh, you know, again, as employers, um, as managers of teams, creating that ability to bring your teams together and have that interaction is really important.
00:19:14.240 --> 00:19:19.920
Um, but again, it's not about one or the other, it's you know about bringing those together, but in a healthy way, right?
00:19:20.000 --> 00:19:26.480
And I think um, you know, what you know, again, it was interesting the children's uh the analogy to the raising kids.
00:19:26.559 --> 00:19:29.119
I mean, I remember when mobile phones first came out, right?
00:19:29.279 --> 00:19:34.640
And you know, I think Steve Job famously said his kids aren't touching a mobile phone until they were 15 or something like that.
00:19:34.799 --> 00:19:44.079
So yeah, but as we look forward and we think about consumers, I think it goes without saying consumer preferences are always changing, right?
00:19:44.240 --> 00:19:45.839
Especially when technology is involved.
00:19:45.920 --> 00:19:54.240
I mean, uh one of my favorite videos I used to show on um some of my keynotes was everyone chasing Pokemons across uh uh Central Park, right?
00:19:54.319 --> 00:19:56.319
And I remember I screened the video and saying, what's happening here?
00:19:56.480 --> 00:19:58.079
And people, some people would get it.
00:19:58.240 --> 00:19:59.519
But it was interesting, right?
00:19:59.680 --> 00:20:06.640
When that happened, we saw technology uh drive a real change in how people were using uh technology.
00:20:06.720 --> 00:20:14.400
And I think we're at that another sort of tipping point here where we're starting to see consumers, you know, um also see the benefits.
00:20:14.559 --> 00:20:22.240
Um, and obviously that ever-changing sort of preferences presents um, you know, real challenges uh for businesses.
00:20:22.400 --> 00:20:36.319
So, yeah, could you sort of you know explain sort of this current landscape we we now find ourselves um and how trust is shifting in consumer preferences in terms of what you're observing or seeing when you're working with different companies and and different geos?
00:20:36.799 --> 00:20:41.039
Are you seeing differences between Europe and APJ in North America?
00:20:41.759 --> 00:20:42.960
Yeah, so let's start there.
00:20:43.039 --> 00:20:59.440
So, one of the interesting things with Europe, and I don't know if you saw there was um an announcement that came out last week that a Swiss firm has created an open source, seemingly more ethical AI that people can use in Europe because obviously there are tougher compliance regulations here in Europe.
00:20:59.599 --> 00:21:14.000
And it's been really interesting to watch the kind of this fierce debate between the sort of more, I think, gung-ho attitude of the states of just like no AI regulation, let's just go for it, is kind of like the Wild West of technology.
00:21:14.079 --> 00:21:31.759
Um, and the European, I'm gonna generalize a bit, but the kind of the more European uh conservative approach to hang on, actually, we need superalignment if we, you know, if we completely ignore what we've learned about issues of data privacy not being upheld and all the rest of it, um, then we're gonna have problems.
00:21:31.839 --> 00:21:42.960
So there's like, we are seeing in real time different approaches to how we rule roll out and rule AI or create legislations that shift how AI is going to be deployed.
00:21:43.119 --> 00:21:49.839
So I think that's one thing to mention is that there are actually different zones where different approaches are being taken.
00:21:50.079 --> 00:21:59.519
Um I also saw, I think it was Denmark, that have rolled out a way in which citizens can upload images of their face.
00:22:00.079 --> 00:22:07.839
And copyright their biometric data so that if companies use their data online, so scrape anything that has that information.
00:22:08.000 --> 00:22:20.400
So imagine, for instance, a social media post on Instagram or TikTok or Snapchat, whatever it is or LinkedIn, that it that basically companies will have to change the way in which they train their data because that data is now copywritten.
00:22:20.559 --> 00:22:25.119
So we are seeing, and also and in um California they're starting to push bills through.
00:22:25.200 --> 00:22:36.960
So we're seeing specific regional differences in how countries, regions, and kind of blocks of territories are responding to AI and in terms of trust and go on the case.
00:22:37.119 --> 00:22:50.799
Yeah, and I I think I mean obviously GDPR was a great example of when that legislation came through, it forced businesses globally that wanted to transact in that part of the world to rethink how they organize themselves.
00:22:50.960 --> 00:23:03.839
So you what you're sharing there is some insights on, you know, early um grassroots sort of or um examples of of countries that are now coming out with their own legislation, which I think will manifest.
00:23:04.000 --> 00:23:08.559
Are you seeing that manifest into a general EU sort of um legislation?
00:23:08.640 --> 00:23:13.039
Um so we've got the EU AI Act, which is is now kind of rolling forward.
00:23:13.119 --> 00:23:18.799
And then there's also other things which are kind of the problem is because it gets its gets into everything.
00:23:18.960 --> 00:23:25.920
So other other areas which are now starting to actually see some traction that are adjacent to AI but also really connected.
00:23:26.079 --> 00:23:37.839
So thinking about social media, Macron has has basically pushed forward um a legislation where people under the age of 15 will have strong stronger regulations around social media.
00:23:38.799 --> 00:23:41.599
And there's under 16s, isn't it, in Australia?
00:23:41.839 --> 00:23:43.519
Yeah, they're under 16 in Australia.
00:23:43.680 --> 00:23:43.759
Yeah.
00:23:44.880 --> 00:23:54.000
But it's interesting because it shows that there is some real action being taken to reduce harms, whether you agree with that particular legislation.
00:23:54.079 --> 00:23:58.079
The point is it's kind of indicating that there are issues by not doing it.
00:23:58.160 --> 00:24:04.400
And there's been a lot of pressure groups trying to make strong cases for why that should be the case.
00:24:04.559 --> 00:24:08.079
Um, but also even things like authentication at the device level.
00:24:08.240 --> 00:24:16.720
So there's this argument that's being kind of had at the moment as to whether applications should have age uh verification within each app.
00:24:16.880 --> 00:24:26.960
So when you log into Instagram or you log into, you know, WhatsApp or whatever, whether the whether the uh verification sits there or whether the verification sits on the device that the person's using.
00:24:27.039 --> 00:24:33.200
And then it will just say to any app on that device, this is uh an under 15 or under 16 person.
00:24:33.359 --> 00:24:37.039
And this matters to AI because AI is making itself felt across social media.
00:24:37.119 --> 00:24:41.279
There's so much AI-generated content and slop and fake influences.
00:24:41.359 --> 00:24:44.799
So these two worlds are actually incredibly tightly linked.
00:24:45.039 --> 00:24:50.400
Um, but thinking more broadly about trust, there's some really interesting things that we're seeing in the broader landscape.
00:24:50.559 --> 00:25:07.759
So I was seeing earlier in the Edelman Trust Barometer that low-income households, which are now increasing as wage gaps shift, uh, are much more distrustful of companies and social media and businesses and NGOs and organizations.
00:25:07.920 --> 00:25:17.759
Um, and also some of the consumer trends we're seeing that people, when they're spending, they tend to be making more deliberate decisions because money is tighter across most demographics.
00:25:17.920 --> 00:25:23.680
They're scrutinizing their purchases more, you know, is it actually a really must-have or is it just a luxury?
00:25:23.839 --> 00:25:25.599
Um, they're being more cautious.
00:25:25.680 --> 00:25:36.640
And so I think there's sort of an age of greater concern, of greater mistrust, of weighing up, you know, am I going to go with this brand or am I not?
00:25:36.880 --> 00:25:38.319
Is this brand more moral?
00:25:38.480 --> 00:25:39.359
Is it not?
00:25:39.599 --> 00:25:48.640
Um, we saw a recent boycott by many people of Spotify because so much of the funding is now going into military AI.
00:25:48.799 --> 00:25:57.119
And so there's a lot more of these complicated questions are directly being addressed in how and where we spend our money.
00:25:57.200 --> 00:26:00.000
So I think we're going to continue to see some of those trends.
00:26:00.319 --> 00:26:01.599
I I couldn't agree more.
00:26:01.759 --> 00:26:06.240
And again, you made me think of um uh that design.
00:26:06.480 --> 00:26:08.079
I'm a big fan of design thinking.
00:26:08.160 --> 00:26:16.559
And when I joined Twilio, actually, one of our uh values um uh at the time and still is, is this idea of walking in customer shoes, right?
00:26:16.640 --> 00:26:31.279
And really, I think in terms of building trust and and requires you to really think, and linking back to an earlier point you made, uh being empathizing with your customers and really thinking about the world through their eyes, right?
00:26:31.359 --> 00:26:39.200
Um especially to your point where trust is becoming um even more and more precious to maintain, right?
00:26:39.359 --> 00:26:42.240
Because of uh a lot of the what you just touched on.
00:26:42.400 --> 00:26:51.599
And I mean, it's a concept that's pretty easy enough to grasp, walking in customers' shoes, you know, but it's definitely hard to really, you know, walk the talk, you know.
00:26:51.759 --> 00:26:57.519
I have an equation I often use with um on stage, PM plus PK equals T.
00:26:57.759 --> 00:27:01.200
You know, promises made plus promises kept equals trust.
00:27:01.279 --> 00:27:03.680
And for me, trust is the bedrock of custom experience.
00:27:03.759 --> 00:27:11.119
So if you once you break that trust, obviously you really are breaking that custom experience that you're delivering to your customer.
00:27:11.440 --> 00:27:21.519
So, you know, where where it used to be a more implicit sort of uh agreement between brands and consumers, it now really does require a more intentional approach.
00:27:21.680 --> 00:27:30.400
Um, you know, when adopting AI within uh your tech stack, you know, how how should businesses be intentional about it?
00:27:30.559 --> 00:27:48.240
Um, what's your recommendations in terms of how they might want to think that that think about the utilization of that technology in ensuring trust is embedded into the process and really is at the heart of what they're trying to achieve when it comes to their engagements with customers?
00:27:48.559 --> 00:27:51.440
So we thinking, because we can we can talk about this from two different angles.
00:27:51.599 --> 00:27:58.720
Are we thinking like a framework for how people within a business can use AI in a more trustworthy way?
00:27:58.799 --> 00:28:00.000
And then we go to the consumer side.
00:28:00.240 --> 00:28:17.920
Yeah, I think we can start with the business because at the end of the day, a lot of the folks are going to be listening here, and what I want them to be thinking about is how can they use deploy a powerful technology like AI in a way that does help them main build, establish, build, and maintain trust.
00:28:18.079 --> 00:28:18.640
Yeah.
00:28:19.039 --> 00:28:27.839
So you talk, I love your equation, by the way, because it infers that you actually have to earn trust, that it's only if you've kept the promise that you actually earn that and that you build it up as a bank.
00:28:27.920 --> 00:28:31.599
Um and we see this in relationships too, because it gives you all of this extra wiggle room.
00:28:31.680 --> 00:28:40.880
If you've earned the trust of people, they're more likely, consumers in particular, are more likely to give you that, I suppose, leniency when things go wrong and forgive your things go wrong.
00:28:41.039 --> 00:28:44.160
So thinking first about frameworks within organizations.
00:28:44.319 --> 00:28:57.200
So trust, obviously, as we've already touched on so many moments in this conversation in different forms, is about relationship and rapport and being able to rely on someone to be authentic, to make decisions that are hopefully for mutual benefit.
00:28:57.440 --> 00:29:04.319
So thinking about that in terms of frameworks for people within organizations, how they can use AI.
00:29:04.480 --> 00:29:10.720
Um, I think the first thing is to also strengthen the cognitive, psychological, and emotional skills.
00:29:10.960 --> 00:29:14.400
So cognitively it's going to be, okay, what's the output?
00:29:14.640 --> 00:29:19.599
If it's generating content, is it using uh sources that I can trust?
00:29:19.839 --> 00:29:21.519
So getting into that fact-checking.
00:29:21.680 --> 00:29:25.759
The psychological element might be, okay, is this tipping into sicker fancy?
00:29:25.839 --> 00:29:32.079
Is the tone just so generic that it just loses all that sense of brand distinctiveness and voice?
00:29:32.160 --> 00:29:39.759
Because when we trust a brand, it's often because, well, depending on the generation, for younger generations, it's because they feel like they're getting some kind of connection.
00:29:39.920 --> 00:29:48.319
For Gen X and Boomers, it's going to be more because there's a proof of the reliability of the brand's products.
00:29:48.400 --> 00:29:57.119
But so when we're thinking about younger people, you know, is the distinctiveness of voice there and the emotional skills within an organization, does this feel like it's landing the right story?
00:29:57.279 --> 00:29:58.640
Does it feel relatable?
00:29:58.799 --> 00:30:05.440
Is it unique enough for it not to feel so generic that people are just going to go, oh, that's dismiss it and that's just AI?
00:30:05.680 --> 00:30:09.039
Then I think there's also an element around building specific guidance.
00:30:09.119 --> 00:30:15.039
So enabling customers, sorry, enabling employees to experiment safely.
00:30:15.119 --> 00:30:20.319
So you're talking about Twilio's Zoom meetings or in-person meetings, getting together, experiment.
00:30:20.640 --> 00:30:22.480
You know, what are the markers of trust?
00:30:22.640 --> 00:30:26.799
What do we need to change in this output to make sure that it is hitting the right notes?
00:30:26.960 --> 00:30:32.480
Um, training employees, you know, this is what goes on in these different types of tools.
00:30:32.640 --> 00:30:34.480
This is how they were designed.
00:30:34.720 --> 00:30:39.039
This is how you might use AI ethically and responsibly.
00:30:39.279 --> 00:30:44.000
Are there biases that we want to scan for and then eliminate?
00:30:44.160 --> 00:30:47.359
Um, how to critically judge AI-generated content.
00:30:47.599 --> 00:30:58.720
Things like sanctioned use, authorizing specific AI tools from trusted vendors that you know that people can experiment with, and then you know, making sure that they're secure platforms.
00:30:58.960 --> 00:31:18.319
So I use Canva, for instance, and when I had to do a big rebrand, rather than spending loads of time on emails back and forth with my designer, we had a template we could adjust in real time, we could experiment, knowing that the assets that we're using are pre-approved, that we're not going to get hit with some kind of legal suit.
00:31:18.480 --> 00:31:19.759
So it's also all of these things.
00:31:19.839 --> 00:31:21.519
So giving people a bit of context.
00:31:21.680 --> 00:31:29.759
And then when it comes to the consumer-facing side, you know, as a brand, what is your stance on how your people use AI?
00:31:29.920 --> 00:31:39.440
We saw um earlier in the year, L'Oreal came out and they said, yep, we're going to use AI in our product ads, but we're not going to use AI-generated people.
00:31:39.599 --> 00:31:41.279
We're going to use real models.
00:31:41.599 --> 00:31:43.119
And that does two things.
00:31:43.279 --> 00:31:54.319
First of all, it enables them to use AI tools to be able to personalize their advertising, to um reduce their costs, which I mean it has all sorts of implications, but I get it.
00:31:54.559 --> 00:32:00.480
But by also addressing the concerns that so many people have are we all going to go out of a job?
00:32:00.640 --> 00:32:02.079
Are we all going to be replaced by AI?
00:32:02.400 --> 00:32:03.119
What are the values?
00:32:03.279 --> 00:32:13.759
When you can outsource everything, it says a lot about what you're not willing to outsource and what you will restrain from doing in order to make an ethical or values-driven point as a company.
00:32:13.920 --> 00:32:16.880
And that was really well received by L'Oreal customers.
00:32:17.039 --> 00:32:21.440
So it's also about your public position on when and how to use it and when not to.
00:32:22.160 --> 00:32:23.519
Yeah, I'll stop there for you now.
00:32:24.000 --> 00:32:24.480
No, no, no, no.
00:32:24.799 --> 00:32:25.440
This is brilliant.
00:32:25.519 --> 00:32:31.599
Um, picking up on that, I mean, you made me think about um the importance of transparency.
00:32:31.920 --> 00:32:41.200
And I mean, I would you agree it's incredibly important that brands really think through how they communicate the way AI is being used.
00:32:41.359 --> 00:32:48.319
Um, you know, whether you're talking to a human or agent is a good example, like a you know, should we be signaling?
00:32:48.559 --> 00:32:55.279
Because obviously the AI agents now, it's very difficult to tell that that's an AI agent unless someone lets you know.
00:32:55.440 --> 00:32:58.079
So again, give me, could you share me some thoughts on this?
00:32:58.160 --> 00:33:14.240
Because I do think, you know, you know, again, we talk a lot about the importance of accountability, responsibility, and transparency when it comes to how we use AI, both in terms of the way we build our products and ultimately how we communicate with our customers about the use of those.
00:33:14.480 --> 00:33:26.559
Do you see that as being quite an important area of investment that businesses need to be um you know uh making in terms of thinking through how they apply, say, transparency, responsibility, and accountability?
00:33:26.640 --> 00:33:31.039
Or are there and I and also are there any other areas that they should be thinking about?
00:33:31.599 --> 00:33:33.440
No, I think you're absolutely bang on.
00:33:33.519 --> 00:33:37.119
And I like to think about it as the art of ethical communication.
00:33:37.279 --> 00:33:40.240
So accountability, responsibility, and transparency.
00:33:40.319 --> 00:33:42.960
That's the art of good communication and building and earning trust.
00:33:43.200 --> 00:33:57.200
And so thinking about um ways and contexts in which you might want to flag that you're using AI, there's a really interesting research piece that came out um fairly recently that was looking at the AI authorship effect.
00:33:57.440 --> 00:34:16.159
And this was a study that found that when companies use or an organization uses AI to craft emotional messages, so emotionally resonant messages, anything that's a piece of marketing or an ad that's there designed to pull on your heartstrings or elicit some kind of emotional response.
00:34:16.320 --> 00:34:26.880
If you're using AI to do that, to create that content, it can induce a really powerful moral disgust response in the recipient.
00:34:27.119 --> 00:34:40.480
So if I'm someone who's receiving an ad and I know that it's been authored by AI and it's it's just really like sending me off, fly, you know, flying off the handle, is going to obviously undermine that brand's reputation, my trust in the brand, et cetera.
00:34:40.639 --> 00:34:47.679
However, and it also reduces positive word of mouth, reduces loyalty, all these sorts of impacts that then negatively impact the bottom line.
00:34:47.920 --> 00:34:53.280
However, if you're using AI as a chatbot and you say, this is an AI chatbot, what's your problem?
00:34:53.440 --> 00:35:16.880
Like fraud detection in banks, great example of positive use where I can just, if I have the right banking app, go in and say, I think there's been a fraudulent activity on my account and it's an AI chatbot that is using information to be able to check that really quickly, and I can resolve it in a way that is reliable, secure, and swift, then that's a brilliant example of how to use AI.
00:35:17.039 --> 00:35:28.000
Um now we see backlashes happening in the context of ads when I'm sure many folks listening or watching this uh will remember the Coca-Cola ad that came out in 2024.
00:35:28.159 --> 00:35:29.599
The holidays are coming, like that.
00:35:29.760 --> 00:35:34.880
And I know, like whatever we think about the hyper commercialization of Christmas, which has its issues.
00:35:35.119 --> 00:35:45.199
Actually, I like the creative around these Coca-Cola ads because it's festive, it's sweet, there's a lot of thought that's gone into it, and you know, it's it's something creative and imaginative.
00:35:45.360 --> 00:35:53.039
And as soon as it got outsourced to AI and they jumped the bandwagon, jumped on the bandwagon like everybody else, it was such a crushing disappointment.
00:35:53.119 --> 00:36:00.480
It's like, oh, this just feels flat and uninspired, and it's like everyone else, and what a shame, right?
00:36:00.639 --> 00:36:13.679
And so I think there is something there about um, you know, yes, of course, it's fine to test something, but understand what it is that makes your brand voice so appealing, and then make sure you protect it.
00:36:13.920 --> 00:36:27.920
And I think it's those companies that are able to protect the brand voice, that distinctiveness, and use AI in ways that are clear, that are transparent, where it serves a really obvious purpose and are restrained in other areas.
00:36:28.000 --> 00:36:37.199
Like with customer service, we've seen so many examples, time and again, where teams get fired only to get rehired three months later because the chatbot just failed spectacularly.
00:36:37.280 --> 00:36:38.400
It was spewing garbage.
00:36:38.559 --> 00:36:41.840
Like this is not something that has gone away as a problem.
00:36:42.000 --> 00:36:54.639
So it's just being discerning about you know where to optimize with AI, how to do it, how to signal it, and where actually you have the heartbeat of your business, don't outsource it because it's just gonna decimate you.
00:36:54.880 --> 00:36:56.320
You made so many great points there.
00:36:56.480 --> 00:36:59.280
I think um, and look, there's no silver bullet here.
00:36:59.360 --> 00:37:02.719
I think we're yeah, we're all still learning and figuring that out.
00:37:02.880 --> 00:37:15.199
And uh I I think in terms of the point you just made about um, you know, some people approach AI about, you know, I can take all these heads out of my, out of uh, out of the organization, but I think that's a completely wrong starting point.
00:37:15.360 --> 00:37:44.320
I think AI, really the way I look at it as an employer, you know, manager is I'm looking at how can AI enable me, first of all, and ultimately teams, actually become more effective in what they're doing, you know, and you know, personally, I think it's increased my output by about 20%, which has enabled me to redeploy that time into areas I always wanted to be able to think, you know, strategically, more strategically about work, you new ideas, etc.
00:37:44.639 --> 00:37:51.280
But where a lot of my time was being, you know, taken up with more transactional, sort of mundane type work, because hey, you've got to do it, right?
00:37:51.440 --> 00:37:51.920
You have to do it.
00:37:52.079 --> 00:37:58.000
Whereas I'm finding, and I that's where I get excited about the the art of the possible in terms of bringing that together.
00:37:58.159 --> 00:38:06.239
Um, but yeah, I don't think the starting point should be, and I do sadly see some businesses thinking, oh wow, I can reduce my headcount by 50% by deploying these technologies.
00:38:06.320 --> 00:38:11.440
And some of them are finding out the hard way that but they're not quite just there yet, even to do that.
00:38:11.519 --> 00:38:13.679
And also, I don't even think it's the right starting point.
00:38:13.760 --> 00:38:22.639
I think it's like, does that enable me to maybe redeploy some of these uh folks into areas that we've always known we needed to address, but we just couldn't, you know, afford?
00:38:22.800 --> 00:38:37.119
So I really do feel that that is a a good way to come at it is really start with how can we use AI to better enhance our employees' abilities to create value for their customers and ultimately for the business.
00:38:37.280 --> 00:38:38.000
Does that make sense?
00:38:38.559 --> 00:38:39.119
Total sense.
00:38:39.360 --> 00:38:39.599
I love it.
00:38:39.679 --> 00:38:41.360
And it's a very people-first approach.
00:38:41.519 --> 00:38:52.159
And actually, you know, unless you're a startup with a team of like one or two and your whole thing is building AI agents, your whole, you know, I mean, that's also gonna have its own challenges.
00:38:52.400 --> 00:38:53.920
Organizations are made up of people.
00:38:54.639 --> 00:39:02.719
And if I understand that people have to, you know, when people have to cut costs, um, but there are implications also to organizational culture.
00:39:02.960 --> 00:39:18.960
If you're creating an environment in which you're making very quick decisions, like the think sort of think later, deploy now, think later sort of thing, where suddenly people are uncertain that they're going to have a job the next week, um, their motivation's likely going to drop.
00:39:19.119 --> 00:39:27.039
There are examples where people have sabotaged um programs and platforms, they're not performing as well.
00:39:27.119 --> 00:39:30.800
So there's also like that flip side of you know, what kind of culture are you creating?
00:39:30.960 --> 00:39:35.360
And to your point, which I love is you know, you've hired these people for a reason.
00:39:35.599 --> 00:39:43.280
You've put lots of money, time and effort and cultural and emotional investment into making sure that they're a great fit.
00:39:43.440 --> 00:39:47.840
Find other ways to make them valuable, to retain them, to reskill them.
00:39:48.000 --> 00:39:53.920
Um, because there are so many blind spots that we still have around AI.
00:39:54.000 --> 00:40:05.039
You know, there's model anthropomorphization where we go, oh, this model must understand me, therefore it's got the human capacities of understanding the zeitgeist or coming up with something novel, which is not the case.
00:40:05.280 --> 00:40:08.320
Or, you know, the fact that it fabricates with conviction.
00:40:08.480 --> 00:40:24.000
We know from studies that when we look to human counterparts, politicians, for instance, or leaders, the most charismatic ones are the ones that we are the most credulous of, the most confident in, even if what they're doing is actually not matching up to their promises.
00:40:24.239 --> 00:40:26.079
And the same is true with models.
00:40:26.480 --> 00:40:35.519
Um, other things that humans can adjust better for between and among ourselves, and we can fact-check or like give feedback for, sick of fancy and positivity bias.
00:40:35.599 --> 00:40:51.440
You know, if you're in a cultural organization or an organization where the culture has been refined to give really clear, useful, valuable feedback of okay, there's something about what you're doing with this particular piece that doesn't work, let's look at it together and let's optimize.
00:40:51.679 --> 00:40:55.360
You can't say that to a Gen AR because it wants to please you.
00:40:55.440 --> 00:41:00.079
So you can't say to an AI, make it less uh consistently be more critical.
00:41:00.159 --> 00:41:02.960
Because that, you know, people have tried this and it's very hard to do.
00:41:03.199 --> 00:41:04.719
Confirmation bias is another thing.
00:41:04.880 --> 00:41:19.280
Like if I go in and I say, oh yeah, I'm convinced that all Gen Z feel X about, you know, this particular product, and I go and test that in my, you know, whatever the LLM is that I'm using, it wants to make me feel um positive about it.
00:41:19.360 --> 00:41:22.000
So it's going to give me all the information to confirm that perspective.
00:41:22.239 --> 00:41:28.639
If I were to go to my team and one of them says, Well, actually, one of my kids is Gen Z, and among their friends, they hate this stuff.
00:41:28.960 --> 00:41:31.519
So it's like, you know, relating it back to the real world.
00:41:31.599 --> 00:41:33.679
We don't live in a virtual world.
00:41:34.000 --> 00:41:42.639
We live in a complex, dynamic, messy environment where we need one another to really gauge what we're doing, where we're going, how to relate.
00:41:42.880 --> 00:41:44.159
I think that's an important point.
00:41:44.239 --> 00:41:45.119
I it was interesting.
00:41:45.280 --> 00:41:49.920
Remember, uh are you I can't remember where you're based, but I won't say just in case we're actually gonna go.
00:41:50.079 --> 00:41:50.719
I'm in Barcelona.
00:41:51.440 --> 00:41:55.440
Actually, yeah, Spain and on the uh Portugal lost power, right?
00:41:55.679 --> 00:41:57.519
For for a day or so.
00:41:57.599 --> 00:41:58.320
What was it?
00:41:58.559 --> 00:41:59.920
And uh exactly, yeah.
00:42:00.239 --> 00:42:18.559
I remember listening to some folks being interviewed about the actual joy they got out of reconnecting with folks, you know, because obviously there was no power, phones were down, everyone was down, and people just went out and sat one another and broke bread and you know, or used the time to to to disconnect, right?
00:42:18.719 --> 00:42:20.000
So it is interesting.
00:42:20.159 --> 00:42:22.239
I do think there's a growing sort of movement.
00:42:22.320 --> 00:42:24.000
And again, it's not about one or the other.
00:42:24.079 --> 00:42:27.679
I do, but you know, I think you can over-index on technology.
00:42:27.760 --> 00:42:29.599
Um, it's not the answers to everything.
00:42:29.679 --> 00:42:39.599
Um, and I do think that human connection back to my earlier point, you know, if I think about some of the best work we do, my team's does is when we connect and talk and push one another around, right?
00:42:39.679 --> 00:42:41.199
In terms of the ideas, et cetera.
00:42:41.360 --> 00:42:45.199
So I think that's really important part of the mix going forward.
00:42:46.000 --> 00:42:46.880
Yeah, completely.
00:42:47.039 --> 00:42:52.800
And I think that thing about like, you know, pushing, pushing against one another, like these are some of the things that help us to grow.
00:42:52.880 --> 00:42:55.679
And as humans, there is such a hundred, yeah.
00:42:55.840 --> 00:43:07.440
And you know, if you're lucky to work in a business, in an organization or in a team where you get to have some of that juicy, healthy friction where you're learning and growing as a person, everyone benefits.
00:43:07.599 --> 00:43:26.239
And I think one of my largest concerns with over relying on AI platforms is that so much of the friction is stripped out, and it's such a kind of potentially easy interaction that we just become quite uh we we we lose our strength.
00:43:26.480 --> 00:43:32.320
It's like a tree that is having to face headwinds grows more strongly.
00:43:32.559 --> 00:43:37.840
And a tree that doesn't face those headwinds, when it is caught by a brisk wind, it will snap.
00:43:38.159 --> 00:43:46.320
And I'm not saying that humans necessarily will snap, but I do believe that resilience comes through grit and growth and pushing and challenging, and it's just more stimulating.
00:43:47.199 --> 00:43:50.000
Um yeah, and the ideas that we generate are just so much better.
00:43:50.239 --> 00:43:50.800
A hundred percent.
00:43:51.039 --> 00:43:51.599
So, yeah.
00:43:52.159 --> 00:43:54.960
So as we look forward, I'm I'm keeping one eye on the time.
00:43:55.039 --> 00:43:59.760
I know I knew I knew that we would uh have no trouble uh talking for an hour.
00:44:02.159 --> 00:44:04.800
I mean, we've sort of taught people and process.
00:44:05.119 --> 00:44:10.800
I'd like to sort of just close off with sort of a few thoughts from you on the technology side.
00:44:11.119 --> 00:44:21.679
And although we have been touching on technology through that, but specifically in terms of the use of technology, there's often a fine line between persuasive technology and manipulation.
00:44:21.840 --> 00:44:28.400
You know, how can organizations ensure they're empowering their customers and not, quite frankly, exploiting them?
00:44:28.719 --> 00:44:29.360
Yeah.
00:44:29.679 --> 00:44:45.360
My answer to this is actually has consistently been fairly the same, which is ask your customers, you know, do some qualitative research, sit down with them, have a conversation, find out, you know, is this actually something that serves you?
00:44:45.440 --> 00:44:46.400
What are your priorities?
00:44:46.559 --> 00:44:53.119
And of course, it's not going to be a blanket everyone, you know, 100% of people say, well, yes, this is perfect and the rest of it.
00:44:53.280 --> 00:45:04.559
But it might be, especially with our ability to hyper-segment, it might be that there are people who are very, very happy to give you more data in exchange for a more personalized tailored experience.
00:45:04.719 --> 00:45:08.400
Other folks might really place a premium on privacy.
00:45:08.559 --> 00:45:13.599
And so the question then always comes back to asking your customers, don't make assumptions.
00:45:13.840 --> 00:45:18.320
Just because you can doesn't mean you should use everyone's data all the time.
00:45:18.559 --> 00:45:20.079
Um, great examples of this.
00:45:20.159 --> 00:45:22.800
I always like the idea of um this particular case study.
00:45:22.880 --> 00:45:29.199
So Bloom and Wild years ago invited people to opt out of specific marketing messages if they didn't want to receive them.
00:45:29.440 --> 00:45:45.599
And they had a huge positive response in the media where people did opt out of, in this instance, it was marketing for Mother's Day messages, and the brand reputation shot sky high because people felt heard, respected, they were being listened to.
00:45:45.679 --> 00:45:49.360
And of course, loyalty increased it, increased, as did customer spending.
00:45:49.440 --> 00:45:57.679
So I think that basic extension of respect, asking people, um, and then that the user buy-in, the customer buy-in is going to be so much higher.
00:45:57.760 --> 00:46:00.400
So it everyone stands to win in that scenario.
00:46:00.719 --> 00:46:01.679
Yeah, I love that.
00:46:01.840 --> 00:46:14.400
Um I often talk about the importance of building brand utility, and this is really at the heart of that, is really coming at it with a service-oriented mind um mindset as an organization, right?
00:46:14.639 --> 00:46:24.079
So there's loads of studies out there that show if people feel that you're creating a utility and how you use the data they're sharing with you.
00:46:24.159 --> 00:46:27.360
Um, and this is where AI can be incredibly powerful, right?
00:46:27.440 --> 00:46:41.519
It can actually, you know, as you said, use the Ford example earlier, can really help provide a genuine assistance in ensuring that um a utility is delivered to you and enhancing the service that's delivered.
00:46:41.679 --> 00:46:53.679
But again, it comes, uh it's it's important that we understand that uh that the the way we use that data is going to be absolutely critical in in establishing and maintaining that trust, right?
00:46:53.760 --> 00:46:56.800
And I think that's an important point for us to consider.
00:46:57.199 --> 00:47:03.360
Maybe, maybe just sort of wrap up on this section, and uh I want to then kick to my final sort of question.
00:47:03.599 --> 00:47:09.920
Are there any specific examples of where AI, in your observations, has strengthened or weakened customer trust?
00:47:10.000 --> 00:47:13.679
Um, and what lessons can other brands maybe learn from those cases?
00:47:14.719 --> 00:47:18.559
So the ones that come to mind that are the strongest ones, there's there's a couple.
00:47:18.639 --> 00:47:30.079
So there's the, I mean several of them, but like certain organizations, um, some of which are actually NGOs using chatbots to respond to customer service inquiries and that going terribly wrong.
00:47:30.239 --> 00:47:41.760
And especially, you know, if you're an NGO, the last thing you want to do is for kind of create conditions in which people who are donating you money feel fobbed off for giving you their hard-earned cash.
00:47:42.000 --> 00:48:01.599
So customer chatbot examples, um some of the other ones that I can think of without wanting to name names, when you've got um AI-generated influences, AI generated music that then top your charts, um, AI generated books.
00:48:01.840 --> 00:48:15.360
So I'm thinking particularly in quote-unquote content or products that people believe are created by people, yeah, that then it becomes apparent that they're not, that undermines trust massively.
00:48:15.440 --> 00:48:23.039
And then I think it kind of points towards the power of, especially in a kind of a deep fake world, the power of authenticity.
00:48:23.199 --> 00:48:48.800
Um, and just kind of with a nod to the future, I think that the companies, inferences, organizations that are going to be the most successful might end up being the ones that include things like mess-ups or, you know, a bit of shonky spelling, or, you know, if they if they swear or they express their distinctiveness in a different way, because that's going to be one of the ways, at least in the short term until AI catches up, that people can prove that they're human.
00:48:48.960 --> 00:48:49.280
Yeah.
00:48:49.440 --> 00:48:55.280
Um, so those are just some of the areas where I've seen examples of the use of AI backfire.
00:48:55.599 --> 00:48:56.639
Um, yeah.
00:48:57.039 --> 00:48:58.000
Oh, that's brilliant.
00:48:58.079 --> 00:49:05.360
And look, I end I usually end these episodes by asking our guests um what resources they would recommend.
00:49:05.519 --> 00:49:11.119
Um, but you've already uh graciously shared uh all of these resources on your page.
00:49:11.280 --> 00:49:20.960
Um so we've included a link actually uh within um this session for folks to go and dive in and geek out on the content that's influenced you or does influence you.
00:49:21.039 --> 00:49:25.360
So I really encourage you, uh the audience, to check that out.
00:49:25.519 --> 00:49:31.039
Um so instead of that, I uh maybe I'll I I've just sort of will tweak the my closing question.
00:49:31.199 --> 00:49:40.480
You know, considering you know technology is advancing so quickly and we and you know we're seeing significant shifts in sort of cultural norms.
00:49:40.639 --> 00:49:46.800
What do you believe will define trust between organizations and individuals over the next decade?
00:49:47.840 --> 00:49:57.760
So I suspect we are going to see um a return to experiential interactions in person, real life.
00:49:58.239 --> 00:50:07.679
Um I'm thinking about, for instance, I'm doing a talk later in the year for Alianz, and they're booking out, as they they do most years, I believe, a tent at Oktoberfest in Munich.
00:50:07.840 --> 00:50:09.360
It's actually at the end of September.
00:50:09.599 --> 00:50:26.480
Things like that, where you get to do something silly with people in real life and have some fun, but there's something about the inherent humanness and messiness and unpredictability of it that people will miss as everything becomes increasingly homogenized online.
00:50:26.719 --> 00:50:31.119
And I also think, you know, we're seeing with younger, with younger folks counter trends starting to appear.
00:50:31.199 --> 00:50:45.840
So offline clubs, people moving off dating apps and going running instead, or you know, the boom in craft um kind of hobbies like painting or woodworking or uh ceramic.
00:50:45.920 --> 00:51:08.719
So we're seeing a huge emphasis on people who perhaps have been raised from year dot in an AI-enabled context or in um a completely online digital nomadic uh digital native context who are not rediscovering a love for in-person relationships and hiking and the natural world, but they're discovering it for the first time.
00:51:08.960 --> 00:51:11.360
Like spending time offline on a digital detox.
00:51:11.599 --> 00:51:18.880
There was a study that came out fairly recently that that was looking at what happens when people come off social media for six whole weeks.
00:51:19.119 --> 00:51:25.039
And one of the lines in the article was that it was tantamount to having therapy, except obviously much cheaper.
00:51:25.280 --> 00:51:43.599
Because all of these stresses on our nervous systems, on our psychological and emotional systems, on our ability to self-regulate, to be able to form relationships with those in our locality, all of that gets impacted when we spend more and more of our time online thinking we have to catch up, consume, perform.
00:51:43.760 --> 00:52:02.960
And so I think the brands that are able to really get ahead of that curve, use AI intelligently in ways that help support those more fundamental, creative, person-to-person relationships and building, um, those are the brands that are going to gain the most in the years to come.
00:52:03.119 --> 00:52:04.960
But it requires some risk taking and investment.
00:52:05.039 --> 00:52:06.559
And so not all brands will do it.
00:52:06.960 --> 00:52:10.079
Uh, I'm very excited to see who does get ahead of the curve and lead the way.
00:52:10.400 --> 00:52:11.920
No, I I couldn't agree more.
00:52:12.000 --> 00:52:18.400
And uh and it'll be a combination of how they interact with their consumers, but also their employees doing the same, right?
00:52:18.559 --> 00:52:20.000
Um, couldn't agree more on that.
00:52:20.159 --> 00:52:27.679
Look, Natalie, um I totally enjoyed interviewing you last year at uh signal.
00:52:27.760 --> 00:52:34.239
And again, today I really am genuinely see you as a a real key influencer in this space, someone I learn a lot from.
00:52:34.320 --> 00:52:37.280
And again, I'm walking away enriched from from this session.
00:52:37.360 --> 00:52:40.800
So, again, thank you for sharing uh those insights.
00:52:40.880 --> 00:52:44.480
I'm sure the audience that's dialed in today um will agree with me.
00:52:44.880 --> 00:52:46.639
There was some real great nuggets in there.
00:52:46.800 --> 00:52:47.840
So, thank you again.
00:52:48.079 --> 00:52:50.400
Uh, your brilliant questions.
00:52:50.559 --> 00:52:51.360
No, I look forward to it.
00:52:51.519 --> 00:52:52.559
You always go straight for the heart.
00:52:52.719 --> 00:52:55.360
It's so good to have that kind of like interview style.
00:52:55.440 --> 00:52:56.800
You're absolutely wonderful.