Latest / The DEX Show: A Show for IT Change Makers / Shadow AI and the Coming Workplace Reckoning (w/ Kay Firth-Butterfield)

Transcript

Shadow AI and the Coming Workplace Reckoning (w/ Kay Firth-Butterfield) · The DEX Show: A Show for IT Change Makers

• 0:02
  Speaker 1: Hey, it change makers and welcome to the.
• 0:05
  Speaker 2: Deck Show with Tomac Grath. Let's get into it.
• 0:10
  Speaker 1: Hey for Butterfield, welcome to the Deck Show. Absolutely wonderful
• 0:13
  Speaker 1: to have you one. How are you.
• 0:15
  Speaker 2: I'm fine. I have a slight cold, so forgive me
• 0:19
  Speaker 2: for that.
• 0:20
  Speaker 1: Sorry to hear that. So do you know, okay? I
• 0:22
  Speaker 1: was familiar with your your familiar with your writing for
• 0:25
  Speaker 1: some time. We've wanted you on the show for some time,
• 0:28
  Speaker 1: but I never knew you were English. I always assumed
• 0:30
  Speaker 1: you were American. And I'm at the moment we're on this,
• 0:33
  Speaker 1: We've got this, We've got this streak of English guests,
• 0:35
  Speaker 1: and everyone's going to think we've become like an Anglo podcast.
• 0:38
  Speaker 1: I was, But you are based in America? Is that right?
• 0:41
  Speaker 2: I am made in America and I have been since
• 0:44
  Speaker 2: two thousand and eight, so quite a long time. But
• 0:47
  Speaker 2: the accent still.
• 0:48
  Speaker 1: Here, absolutely absolutely whereabouts the whereabouts in the states?
• 0:52
  Speaker 2: You I live in the Northeast, so I have snow
• 0:56
  Speaker 2: on the ground.
• 0:57
  Speaker 1: Fantastic. And another thing, so I was refreshing my memory
• 1:02
  Speaker 1: on your on your bio right as they say, and
• 1:06
  Speaker 1: I was kind of already had a note that you
• 1:08
  Speaker 1: were you were the first Chief AI Ethics officer. But
• 1:12
  Speaker 1: when I was refreshing these notes, I saw it was
• 1:14
  Speaker 1: back in twenty fourteen. K Yeah, nobody, how did that happen?
• 1:19
  Speaker 1: What was that? What was that? Well?
• 1:21
  Speaker 2: I think that was just that I had. You know,
• 1:25
  Speaker 2: I'm a human rights barrister by background. I still co
• 1:29
  Speaker 2: chair the work at my chambers in Doughty Street, London
• 1:34
  Speaker 2: on AI and human rights. And I had been working
• 1:40
  Speaker 2: a great deal in human trafficking and this is actually
• 1:44
  Speaker 2: before Westworld, and so they.
• 1:50
  Speaker 3: I had.
• 1:52
  Speaker 2: Started reading about artificial intelligence. My husband's a computer scientist,
• 1:59
  Speaker 2: and I've been thinking about, you know, if we created
• 2:02
  Speaker 2: humanoid robots, would we want to traffic humans anymore? And
• 2:08
  Speaker 2: so I got myself into this big thought process about okay,
• 2:13
  Speaker 2: so we might have super intelligence and what would that
• 2:17
  Speaker 2: mean for humans? And how would AI help us and
• 2:21
  Speaker 2: how might it hinderers? And that was back in twenty
• 2:24
  Speaker 2: ten or so.
• 2:27
  Speaker 1: The last.
• 2:30
  Speaker 2: Chapter of my book from that period, which was on
• 2:33
  Speaker 2: human rights, actually had this thought experiment in it. And
• 2:40
  Speaker 2: then in twenty thirteen, I Stephen Hawkin said in the
• 2:47
  Speaker 2: Times that AI might be the best thing that humans
• 2:51
  Speaker 2: ever do, or it might be the last thing that
• 2:52
  Speaker 2: we ever do. And I happened to be on a
• 2:56
  Speaker 2: plane ten hours from London to Austin, Texas. Well we
• 3:00
  Speaker 2: lived at the time, and I sat next to the
• 3:03
  Speaker 2: CEO of an AI startup and I was perhaps one
• 3:08
  Speaker 2: of I don't know ten of ten people in the
• 3:10
  Speaker 2: world who could have talked to him about this issue,
• 3:14
  Speaker 2: and he offered me that job, which of course didn't
• 3:17
  Speaker 2: exist at the time, so we had to come up
• 3:19
  Speaker 2: with a title. As we were coming into land in Austin.
• 3:25
  Speaker 1: Well, that's an amazing way. Indeed, there's an amazing entrance
• 3:28
  Speaker 1: to discuss your very unique career history right as you say,
• 3:31
  Speaker 1: human rights barrister right seeking into AI conceptual specialist. And
• 3:40
  Speaker 1: if I can, if I can just ask you so
• 3:42
  Speaker 1: I want to hone in on what you just said.
• 3:44
  Speaker 1: So you were specializing in the issue of human trafficking.
• 3:49
  Speaker 2: I was human rights in human trafficking. That's what I
• 3:52
  Speaker 2: was teaching at the time at the university in Austin.
• 3:57
  Speaker 1: And just to clarify for my ownunderstanding, you saw in
• 4:02
  Speaker 1: a I a potential technology that could replace the need,
• 4:06
  Speaker 1: as it were, for human trafficking. Do I have had
• 4:08
  Speaker 1: approximaty yes, that's right.
• 4:10
  Speaker 2: Yeah, yes, I mean people choose all robots, as we
• 4:15
  Speaker 2: saw when the West World series came out, you know,
• 4:19
  Speaker 2: maybe we're going to abuse them and not human beings.
• 4:24
  Speaker 2: I think I've since come to the view that maybe
• 4:26
  Speaker 2: as a species we should just grow up and not
• 4:29
  Speaker 2: abuse anybody. But it would certainly be better not to
• 4:34
  Speaker 2: abuse people if I have to make a choice.
• 4:37
  Speaker 1: It reminds me a little bit. There was a Christopher
• 4:40
  Speaker 1: Hitchin's quote I remember reading years and years ago about
• 4:45
  Speaker 1: it was about if it was about the issue of
• 4:49
  Speaker 1: animal cruelty. Yeah, he said that obviously that only exists
• 4:53
  Speaker 1: because people suppose that animals experience something, because you know,
• 4:57
  Speaker 1: they experienced something like consciousness, because otherwise they might well
• 5:00
  Speaker 1: I'm taken out on a toaster. Yeah.
• 5:03
  Speaker 2: Absolutely. Well, interestingly, you know, I sit on the board
• 5:08
  Speaker 2: of a small nonprofit called a Species Project, and what
• 5:14
  Speaker 2: we are doing is we're using machine learning to discover
• 5:21
  Speaker 2: a large language model to discover animal language. And our
• 5:26
  Speaker 2: purpose in discovering animal language is to see whether we
• 5:30
  Speaker 2: can actually find out what they're saying to one another
• 5:34
  Speaker 2: and whether we could communicate with them. And so, you know,
• 5:39
  Speaker 2: there's a lot of talk about in AI about Oh
• 5:43
  Speaker 2: it might be sentient or becoming sentient, But we actually
• 5:48
  Speaker 2: do have a lot of sentient beings on our planet
• 5:55
  Speaker 2: already and we're not talking to them. So that's what
• 5:57
  Speaker 2: we're trying to do.
• 5:58
  Speaker 1: Yeah, before trying to invent some want to talk to,
• 6:00
  Speaker 1: we might try talking to the creatures already around us. Well,
• 6:04
  Speaker 1: that would be a very awkward moment, considering our history
• 6:08
  Speaker 1: of behavior towards animals.
• 6:11
  Speaker 2: Absolutely, absolutely, and I think that it raises as well,
• 6:16
  Speaker 2: you know, responsible AI questions, because supposing we do get
• 6:21
  Speaker 2: to talk to whales, do they then say, well, we'd
• 6:25
  Speaker 2: rather not be talked to and and we don't much
• 6:29
  Speaker 2: like your species after anyway, because look at what you've
• 6:33
  Speaker 2: been doing to you. But you know, if we get
• 6:38
  Speaker 2: it right, it will fundamentally change the conversation that we
• 6:43
  Speaker 2: have about the planet.
• 6:45
  Speaker 1: Indeed, it would. And so I'm curious, Okay, given that
• 6:50
  Speaker 1: rather deep background in thinking about and indeed working directly
• 6:53
  Speaker 1: with AI or what we I guess at this point
• 6:56
  Speaker 1: maybe think of as like early nascent forms of AI
• 7:00
  Speaker 1: compared to the acceleration we've seen in the last what
• 7:02
  Speaker 1: three four years? In particular, what has that experience been
• 7:06
  Speaker 1: like for you, I mean where we Where were you
• 7:08
  Speaker 1: when the LLLM was first unleashed and what did that
• 7:11
  Speaker 1: mean for you? Was this were you like, oh, finally
• 7:13
  Speaker 1: exactly what I've been anticipating all this time? Or was
• 7:16
  Speaker 1: it arresting and striking? Was what was that? Oh?
• 7:19
  Speaker 2: I mean I was working for the World Economic Forum
• 7:22
  Speaker 2: at the time. I was leading AI and then eventually
• 7:26
  Speaker 2: Quantum at the World Economic Forum, and I you know,
• 7:31
  Speaker 2: I think the first thing the it's exciting and to
• 7:40
  Speaker 2: see AI get better. Maybe we could have conversations about
• 7:45
  Speaker 2: whether it's getting better or not. And I am a
• 7:50
  Speaker 2: great believer that you know, if we get AI right,
• 7:54
  Speaker 2: it will fundamentally change and help humans and the planet.
• 7:59
  Speaker 2: But if we don't get it right, and we don't
• 8:02
  Speaker 2: put in the right guardrails, both for our home, the
• 8:06
  Speaker 2: way we use it at home and engine our lives,
• 8:10
  Speaker 2: and the way we use it in business, then that
• 8:12
  Speaker 2: might not be the impact. And so you know, I
• 8:18
  Speaker 2: was at Davos in twenty twenty four and everybody wanted
• 8:26
  Speaker 2: to talk about LMS. Whether they knew anything about it
• 8:31
  Speaker 2: or not, they wanted to talk about it.
• 8:35
  Speaker 1: You had a full schedule at DeVos.
• 8:38
  Speaker 2: Absolutely and I think it's interesting to note that, you know,
• 8:42
  Speaker 2: I was there this last month, month before, and one
• 8:47
  Speaker 2: of the things that's happened is that, you know, we
• 8:50
  Speaker 2: went from people were talking about it, you didn't really
• 8:52
  Speaker 2: understand it a lot of height over two years, and
• 8:57
  Speaker 2: then this year there is much much more. Everybody's not
• 9:01
  Speaker 2: got AI all over the houses. They've got we can
• 9:05
  Speaker 2: help you reach your ROI with AI on the houses.
• 9:10
  Speaker 2: And I think that that's showing us sort of maturation
• 9:13
  Speaker 2: of how we're seeing AI develop across business.
• 9:18
  Speaker 1: Well, your your your recent book where it coexisting with
• 9:21
  Speaker 1: AI work glove and playing and changing world obviously is
• 9:24
  Speaker 1: quite a holistic narrative. So interested before we kind of
• 9:28
  Speaker 1: drill into your insights around AI and the workplace specifically
• 9:33
  Speaker 1: looking at it a little bit more holistically and thinking
• 9:35
  Speaker 1: about those guardrails as you just articulated it. Yeah, sure,
• 9:41
  Speaker 1: tider for workplace, what what where? Where the guard rail
• 9:46
  Speaker 1: has become most important in your view before we go
• 9:48
  Speaker 1: into the workplace.
• 9:50
  Speaker 2: Yeah, certainly. So. You know the reason that that the
• 9:54
  Speaker 2: way that I the reason I wrote the book is
• 9:57
  Speaker 2: that I believe that there are many people who are
• 10:01
  Speaker 2: struggling in the workplace because they're not getting the proper
• 10:04
  Speaker 2: training to use AI. They're worried that they're going to
• 10:08
  Speaker 2: lose their job to it. They don't understand it, and
• 10:12
  Speaker 2: so therefore they don't understand that it can help them
• 10:15
  Speaker 2: do their job and they won't necessarily lose their job
• 10:18
  Speaker 2: because of it. And so I wanted really to ensure
• 10:24
  Speaker 2: that everybody could pick up this book and learn something
• 10:28
  Speaker 2: and have a conversation and really have a say in
• 10:32
  Speaker 2: how we humans go forward with AI. Because of the
• 10:37
  Speaker 2: at and also because I'm a crazy Shakespeare fan, I
• 10:41
  Speaker 2: base the book on the Jake Bay's speech and as
• 10:46
  Speaker 2: you like it are the seven stages of man, and
• 10:52
  Speaker 2: so I'm and so I start by looking by obviously
• 10:57
  Speaker 2: looking at what the technology is, but that, but then
• 11:02
  Speaker 2: I go into looking at infants and AI, schoolboy, lover, soldier,
• 11:11
  Speaker 2: business person, justice, and society, and then into old age
• 11:16
  Speaker 2: and healthcare. And so I just some just some things
• 11:26
  Speaker 2: to pick out of that. We're being asked as parents
• 11:32
  Speaker 2: and grandparents to buy toys for the under sixes. The
• 11:38
  Speaker 2: under sixes are in that stage where they're creating their beliefs,
• 11:43
  Speaker 2: values and attitudes, and you know, do how do we
• 11:48
  Speaker 2: think about ensuring that they are able to communicate and
• 11:55
  Speaker 2: work in teams with other human beings. If actual they
• 12:01
  Speaker 2: are that they have a toy that is always sycophantic
• 12:07
  Speaker 2: and loves them all the time and doesn't ever make
• 12:12
  Speaker 2: them cry, and then they have to go and talk
• 12:14
  Speaker 2: to humans. You know. So when we're thinking about the
• 12:19
  Speaker 2: future work, future of work, and future workforce, it actually
• 12:25
  Speaker 2: is starting at that point of what do we how
• 12:28
  Speaker 2: do we think about educating our very youngest children, And
• 12:33
  Speaker 2: then if we go into the student portion, what we're
• 12:38
  Speaker 2: seeing is that what we're seeing is that AI is
• 12:46
  Speaker 2: everywhere in the curriculum. There is no good evidence that
• 12:50
  Speaker 2: it is actually helping us learn. And what we do know,
• 12:57
  Speaker 2: because there is now some evidence, is that if we
• 13:01
  Speaker 2: write things down we learn it better than if we
• 13:05
  Speaker 2: type things. And if we spend too much time outsourcing
• 13:11
  Speaker 2: our homework to AI, then we're not learning anything. And
• 13:18
  Speaker 2: I was at a major university giving a speech two
• 13:21
  Speaker 2: weeks ago, and the a professor of law and a
• 13:27
  Speaker 2: professor of AI, both said, we have problems amongst our
• 13:31
  Speaker 2: seniors because we don't actually think that they know enough
• 13:36
  Speaker 2: to get their degree, and yet they have passed all
• 13:42
  Speaker 2: the exams along the way and say, and you know,
• 13:46
  Speaker 2: one particular, the AI professor said, one of my seniors
• 13:52
  Speaker 2: brought me code the other day and asked me to
• 13:54
  Speaker 2: explain it to him. And I said, well, you know,
• 13:57
  Speaker 2: shouldn't you be explaining it to me? And he couldn't
• 14:01
  Speaker 2: because he had always used an al alam code for him. M.
• 14:07
  Speaker 2: So these sort of things that then feed into how
• 14:12
  Speaker 2: we think about employment because you know, in the past,
• 14:16
  Speaker 2: we've always said, oh, you've got an undergraduate degree, you've
• 14:19
  Speaker 2: got some critical thinking skills, you've got some analytical skills.
• 14:23
  Speaker 2: You know, you've learned something. You bring these these skills
• 14:27
  Speaker 2: into the workplace. But maybe people don't bring those skills
• 14:31
  Speaker 2: into the workplace. They don't bring the ability to work
• 14:34
  Speaker 2: in teams, the things we value in human in human workforce,
• 14:39
  Speaker 2: and so maybe we have to think very differently about
• 14:43
  Speaker 2: how we train our new workers.
• 14:47
  Speaker 1: You know, I'm desperate to get into the workplace with
• 14:49
  Speaker 1: you a little further. But you just snagged my attention
• 14:52
  Speaker 1: with this mention of the university system, because obviously one
• 14:56
  Speaker 1: of the one of the differences there is they can
• 14:59
  Speaker 1: put you know, if you if a din at university,
• 15:00
  Speaker 1: if you're a university. You can impose whatever learning regime
• 15:04
  Speaker 1: you like. You know, you can you can limit the
• 15:07
  Speaker 1: way people use technology, and you can and you can
• 15:09
  Speaker 1: think about the way people are tested. For instance, k
• 15:13
  Speaker 1: I've always thought that the the oral exam, right, the
• 15:16
  Speaker 1: which was the traditional you know approach, right, people would
• 15:19
  Speaker 1: have their what it's even called, you know, your vivor
• 15:21
  Speaker 1: or something. You would stand there and you have the
• 15:23
  Speaker 1: field questions. Obviously PhD students are still the case, I think,
• 15:27
  Speaker 1: isn't it. But maybe maybe that more oral approach to
• 15:31
  Speaker 1: examination is going to come back in for undergraduates or something.
• 15:34
  Speaker 1: I mean, well, just from your experience of talking to
• 15:36
  Speaker 1: professors and thinking about it obviously so deeply, is that
• 15:39
  Speaker 1: one of the directions you think is inevitable and any other?
• 15:43
  Speaker 2: Yeah, I mean I think that what a lot of
• 15:46
  Speaker 2: people are doing, having noticed this, is saying we can't
• 15:51
  Speaker 2: stop using it because you're going to use it. What
• 15:55
  Speaker 2: we can do is we can say we can tell
• 15:57
  Speaker 2: you that it'll get to your B grade, but it
• 15:59
  Speaker 2: won't get you the next level up because that requires
• 16:03
  Speaker 2: some extra thinking at the moment. But we could restructure
• 16:10
  Speaker 2: the way that classes are This will work in smaller universities.
• 16:16
  Speaker 2: But it won't work in some of the big American
• 16:18
  Speaker 2: universities where you have one hundred people in a class
• 16:21
  Speaker 2: or more than one hundred people in a class. But
• 16:24
  Speaker 2: if you've got smaller classes, you can get the homework
• 16:30
  Speaker 2: done and then you can make them explain what they
• 16:33
  Speaker 2: are doing, what they learned, and defend what they said
• 16:40
  Speaker 2: in their homework in class. And that is certainly a
• 16:43
  Speaker 2: way to begin to think about this. And you know,
• 16:50
  Speaker 2: I certainly, I certainly when I was at university, there
• 16:54
  Speaker 2: were seven people in my in my tutorial, and so
• 17:01
  Speaker 2: then you there was nowhere to hide. But obviously for
• 17:05
  Speaker 2: the big universities, it's it's a different it's a different thing.
• 17:09
  Speaker 2: And so then maybe it's rethinking how you examine.
• 17:15
  Speaker 1: Yeah, makes a lot of sense. And okay, so let
• 17:17
  Speaker 1: let's let's go into the workpost. And we already mentioned
• 17:19
  Speaker 1: but one thing, one of the implications here is that
• 17:23
  Speaker 1: new employees or graduates in particular, maybe you don't necessarily
• 17:27
  Speaker 1: have the luxury of assuming they're going to bring critical
• 17:30
  Speaker 1: thinking skills or any solid knowledge base. And so how
• 17:35
  Speaker 1: do we address that? First of all, I guess is
• 17:37
  Speaker 1: at the gate, what's the first thing bring.
• 17:39
  Speaker 2: Out of the gate. We we are not doing a
• 17:42
  Speaker 2: good job generally of teaching AI one oh one two
• 17:47
  Speaker 2: people and say we need to be doing that immediately.
• 17:53
  Speaker 2: But we also need to be assessing our new employees
• 17:58
  Speaker 2: when they come in or what they know about AI
• 18:02
  Speaker 2: and how they use it. So you know, twenty five
• 18:05
  Speaker 2: percent of American men are said to be in an
• 18:08
  Speaker 2: intimate relationship with a chatbot. Seventy two percent of teens
• 18:17
  Speaker 2: are in some form of relationship with a chatbot. That's
• 18:20
  Speaker 2: maybe as a companion or as an intimate companion, and
• 18:27
  Speaker 2: or they know somebody who is and I guess that
• 18:30
  Speaker 2: might be a parent for example. But what does that
• 18:36
  Speaker 2: do you know when you when you bring that into
• 18:40
  Speaker 2: the workplace, how does it how you thinking about the
• 18:44
  Speaker 2: AI tools that you're using in the workplace. It's it's
• 18:48
  Speaker 2: very different from the way that you or I might
• 18:50
  Speaker 2: think about using AI in the workplace because we don't
• 18:54
  Speaker 2: have that intimate connection with it. So I think that
• 18:57
  Speaker 2: that's something that we have to really think about out
• 19:00
  Speaker 2: as we are taking young people into our companies at
• 19:04
  Speaker 2: the moment and then retraining them to remember that this
• 19:09
  Speaker 2: is this is a machine and and you should be
• 19:14
  Speaker 2: using it to help you, but not but not to
• 19:20
  Speaker 2: do all of your work, because otherwise, you know, we're
• 19:24
  Speaker 2: we're seeing workslop from that, we see hallucinations from that,
• 19:29
  Speaker 2: and and all of those are really bad for business
• 19:32
  Speaker 2: and bad for your bottom line.
• 19:34
  Speaker 1: Absolutely, So let's start with the the HR element I get.
• 19:40
  Speaker 1: So you have these employees where where as you put it,
• 19:44
  Speaker 1: intimate relationships with AI is becoming normalized. And I'm trying
• 19:47
  Speaker 1: to picture what the challenges that that causes for a
• 19:51
  Speaker 1: workplace might be. Is my assumption that there's like the
• 19:56
  Speaker 1: danger of excessive fluidity between a work tool and a
• 20:00
  Speaker 1: personal tool. Okay for instance, you know, and this kind
• 20:04
  Speaker 1: of which withdraws attention to the kind of notion of
• 20:07
  Speaker 1: like the the workplace l M is against the out
• 20:12
  Speaker 1: of work. Talk to us about that fluidity and how
• 20:15
  Speaker 1: leaders should be aware of it and should be thinking
• 20:17
  Speaker 1: about it and preparing to encourage better discrimination between the two.
• 20:22
  Speaker 2: Perhaps I don't know, Yeah, I mean I think if
• 20:25
  Speaker 2: you if you come into using AI in the workplace
• 20:29
  Speaker 2: from a point where you're where it's your best friend
• 20:33
  Speaker 2: or it's your intimate partner, then you have forgotten that
• 20:39
  Speaker 2: it is a tool that you need to check on
• 20:42
  Speaker 2: and that Halley snakes and that doesn't always give you
• 20:47
  Speaker 2: the right the right answer, And and so you you've
• 20:58
  Speaker 2: you've lost your ability to interact with with it wisely
• 21:05
  Speaker 2: in in in the workplace, because you have suspended all
• 21:11
  Speaker 2: of the things that you should know about how to
• 21:13
  Speaker 2: use AI, including the fact that it's a machine and
• 21:18
  Speaker 2: it's not your best friend or your lover, or which
• 21:21
  Speaker 2: perhaps shouldn't be your best friend or your lover. I
• 21:26
  Speaker 2: was at a conference in January where the CEO of
• 21:30
  Speaker 2: the Kinsey Institute said, if we keep going on like this,
• 21:33
  Speaker 2: you know, it will be an existential crisis for human beings.
• 21:38
  Speaker 1: Yes, no, that that about it, and the allusion to
• 21:43
  Speaker 1: workslove as well as a result of its regular listener
• 21:46
  Speaker 1: as well know, I'm often quite preoccupied with what I
• 21:49
  Speaker 1: suspect is going to be a crisis in quality control
• 21:51
  Speaker 1: in the workplace. Yeah, and that one of the you know,
• 21:56
  Speaker 1: AI delivers quantity often to the deadment of quality, and
• 22:03
  Speaker 1: your thoughts on that theme very welcome. And also also
• 22:07
  Speaker 1: what does that mean? What does that imply for leaders?
• 22:09
  Speaker 1: What does that imply for organizational structures?
• 22:12
  Speaker 2: Yeah, well, I think I think you know the HBr
• 22:16
  Speaker 2: article says that you know it takes up to one
• 22:21
  Speaker 2: hour and fifty minutes to put right some of the stuff.
• 22:25
  Speaker 2: And so what I don't think that companies are doing
• 22:29
  Speaker 2: well is looking at the cost benefit of AI and
• 22:35
  Speaker 2: what they think about efficiency and resilience. So they might
• 22:39
  Speaker 2: be being efficient using AI, but are they actually resilient
• 22:45
  Speaker 2: using AI? Where you've got workslop, where you've got people
• 22:49
  Speaker 2: who haven't been educated properly about how to use the tools,
• 22:54
  Speaker 2: and what's the cost benefit? You know, it's it may
• 22:59
  Speaker 2: be that AI is really good as a tool in
• 23:04
  Speaker 2: your organization for writing emails, and you don't care if
• 23:09
  Speaker 2: there are hallucinations in the emails, but if there are
• 23:13
  Speaker 2: hallucinations in documents in your finance department, you probably do
• 23:18
  Speaker 2: care about that. But I'm not seeing companies doing that assessment,
• 23:25
  Speaker 2: either correctly or at all. So and that then feeds
• 23:32
  Speaker 2: into this workslop thing. But also I think it affects
• 23:38
  Speaker 2: your bottom line and it perhaps affects your brand value
• 23:42
  Speaker 2: as well, And it's going to be worse if people
• 23:48
  Speaker 2: are not thinking about how they constrain their agents as
• 23:52
  Speaker 2: they are using them. And we all know that agents
• 23:57
  Speaker 2: can if you don't constrain them properly, and you don't
• 24:02
  Speaker 2: and you don't allow and retrain them, then you end
• 24:11
  Speaker 2: up with this terrible problem of them going off peacet
• 24:19
  Speaker 2: And there's a recent IDC report that suggests that by
• 24:25
  Speaker 2: twenty thirty, twenty percent of the leading companies of the
• 24:29
  Speaker 2: world will have been sued because of their agents.
• 24:33
  Speaker 1: Fascinating they think about CIOs and head of EEM and
• 24:39
  Speaker 1: head of IT, people in charge of distributing this technology
• 24:44
  Speaker 1: and hopefully managing adoption and monitoring safety and such like.
• 24:50
  Speaker 1: What would you say of a key government governance decisions
• 24:55
  Speaker 1: around employee AI? But will mattter most in the next
• 24:58
  Speaker 1: twelve to twenty four months.
• 25:00
  Speaker 2: Yeah, I would say certainly.
• 25:04
  Speaker 3: The the what you need is not CIOs or CTOs
• 25:16
  Speaker 3: being in charge.
• 25:18
  Speaker 2: What you need is a good internal committee that includes
• 25:23
  Speaker 2: everybody across all the silos who can really get a
• 25:28
  Speaker 2: grip on where AI is being used, why it's being used,
• 25:33
  Speaker 2: whether it's being used effectively, where you need training. All
• 25:37
  Speaker 2: of these sort of things are incredibly important for a company,
• 25:45
  Speaker 2: and leaving it to the CTO or the CEI CIO
• 25:50
  Speaker 2: what you're doing is actually sort of leaving it in
• 25:53
  Speaker 2: the hands of the people who are being told to
• 25:56
  Speaker 2: deploy it, so you need to give them some support
• 26:01
  Speaker 2: because they're being told to deploy it, and yet there
• 26:04
  Speaker 2: are problems with that deployment, and so so really making
• 26:09
  Speaker 2: sure that there is a cross company committee working on it.
• 26:16
  Speaker 2: The best the best deployments of AI that I see
• 26:20
  Speaker 2: have that, and the best deployments of AI that I
• 26:26
  Speaker 2: see come in companies that are already regulated. They understand
• 26:32
  Speaker 2: the regulatory procedures, so you see, you know, banks and
• 26:36
  Speaker 2: insurance companies, they've really got this on the whole because
• 26:41
  Speaker 2: they already understand what problems can be from the from
• 26:47
  Speaker 2: the regulation. And I think and I you know, to
• 26:52
  Speaker 2: give you an example of training and that fluidity that
• 26:56
  Speaker 2: you talked about earlier, shadow AI. That's where the employee
• 27:06
  Speaker 2: brings their own AI into the company and is using
• 27:11
  Speaker 2: that instead. And already you know, I have advised a
• 27:15
  Speaker 2: company on whether they will be sued successfully because an
• 27:22
  Speaker 2: employee used a shadow AI. And just to give you
• 27:24
  Speaker 2: and give you that example, it was a safety critic
• 27:28
  Speaker 2: company making safety critical systems and the employee used their
• 27:36
  Speaker 2: own AI version of AI on their phone to decide
• 27:42
  Speaker 2: how big the valve should be and when it was
• 27:47
  Speaker 2: all installed, the valve blew up and it blew noxious
• 27:53
  Speaker 2: fumes and content all over people and the site. So
• 27:58
  Speaker 2: you know what they said to me, well, you know,
• 28:01
  Speaker 2: are we going to get sued? And if we're sued,
• 28:04
  Speaker 2: are we going to lose? And I said, well, what
• 28:06
  Speaker 2: rules have you got about shadow ai about your employee
• 28:11
  Speaker 2: using this? And of course they had no rules at all,
• 28:14
  Speaker 2: so of course again you get sued. And yet it
• 28:17
  Speaker 2: would be really simple to have a rule about shadow AI.
• 28:21
  Speaker 1: Is there okay, so what kind of rules should you have?
• 28:24
  Speaker 1: And you know, because you can fall out the left
• 28:27
  Speaker 1: side of the right side of a boat, caren't you? You
• 28:28
  Speaker 1: know there can be a excess and you know as
• 28:33
  Speaker 1: much as there can be, yes.
• 28:34
  Speaker 2: I mean for that, you could just simply say you
• 28:38
  Speaker 2: are only allowed to use the tools that we give you.
• 28:41
  Speaker 2: You are not allowed to use any AI tools that
• 28:44
  Speaker 2: are not supplied by the company. And you might because
• 28:48
  Speaker 2: it's quite critical if people do, you might want to
• 28:53
  Speaker 2: add that it's a sackable offense. It's extremely simple to
• 28:58
  Speaker 2: put that into your into your handbook Rachel handbook.
• 29:03
  Speaker 1: Yeah, that's fascinating. And I guess then visibility though it
• 29:06
  Speaker 1: becomes hugely important for organizations. I mean, do you think
• 29:10
  Speaker 1: that in most companies at the moment, they don't really
• 29:12
  Speaker 1: know who's doing or using what that's part of the problem,
• 29:16
  Speaker 1: part of your overall picture.
• 29:18
  Speaker 2: Yes, I do. I do think that that's part of
• 29:21
  Speaker 2: the problem. And that goes back to training. We're not
• 29:24
  Speaker 2: training people well enough to understand the problems that they
• 29:29
  Speaker 2: might be encountering, and we are not in many In
• 29:35
  Speaker 2: many companies there are no guidelines at all, and having
• 29:41
  Speaker 2: some guidelines and some training would actually help most of this.
• 29:47
  Speaker 1: For sure. And you know, there's got to be one
• 29:50
  Speaker 1: thing I've spoken to sort of think as an AI,
• 29:54
  Speaker 1: they often emphasize the need for freedom of experimentation, particularly
• 29:59
  Speaker 1: in organizations, you know, the dissolution of traditional silos and
• 30:02
  Speaker 1: professional silos and just wanting all your employees to feel
• 30:07
  Speaker 1: that freedom of you know, the AI gives you of
• 30:10
  Speaker 1: being able to do in many ways, almost anything. And
• 30:13
  Speaker 1: so if there's some tension of sorts between you know,
• 30:17
  Speaker 1: that need to give people that kind of autonomy and
• 30:20
  Speaker 1: the need to make sure they don't they don't violate
• 30:24
  Speaker 1: certain important rules and boundaries, and how do you strike
• 30:27
  Speaker 1: that balance?
• 30:28
  Speaker 2: Do you think, Yeah, absolutely there are. And I think
• 30:34
  Speaker 2: that that goes back to my cost, my risk and benefit,
• 30:40
  Speaker 2: resilience and efficiency comments. You know, you do you need
• 30:47
  Speaker 2: everybody to be using AI all over the organization all
• 30:51
  Speaker 2: the time. If you do, then you need to put
• 30:54
  Speaker 2: in some governance unless you say, I have done my
• 30:58
  Speaker 2: cost risk as now as this, and it doesn't matter
• 31:02
  Speaker 2: to me if there are lots of hallucinations in my
• 31:07
  Speaker 2: finance finance reports, or there are hallucinations in reports that
• 31:13
  Speaker 2: I give out to my clients. And you know, we
• 31:18
  Speaker 2: have seen where companies have passed on to clients hallucinated
• 31:26
  Speaker 2: material that is not good for your business, and it's
• 31:29
  Speaker 2: certainly not good for your bottom line. And also I
• 31:34
  Speaker 2: think that one of the things that people are not
• 31:37
  Speaker 2: thinking about sufficiently is if you have hallucinations in your
• 31:44
  Speaker 2: internal material, those are actually going into your proprietary data
• 31:51
  Speaker 2: and can be called upon by any other employee and
• 31:55
  Speaker 2: say then you get double hallucinations or triple hallucinations. Those
• 32:00
  Speaker 2: are the things that I think that companies are not
• 32:03
  Speaker 2: dealing with or are not dealing with well. And so
• 32:08
  Speaker 2: how would you how would you allow employees to innovate
• 32:13
  Speaker 2: as well as as well as having some rules. I
• 32:19
  Speaker 2: think you have to have some basic rules. You have
• 32:22
  Speaker 2: to make a decision as a company about what rule,
• 32:26
  Speaker 2: what are your red lines, and then you have to
• 32:31
  Speaker 2: think about how you train everybody so that they understand
• 32:35
  Speaker 2: the red lines and then can experiment understanding the red lines.
• 32:42
  Speaker 2: There are also ways that I've I've seen companies do this.
• 32:47
  Speaker 2: One would be putting a training and an AI ambassador
• 32:52
  Speaker 2: in each part of the company, and so that person
• 32:57
  Speaker 2: knows how AI might be really well used in, for example,
• 33:03
  Speaker 2: finance or and so then they can help other employees
• 33:10
  Speaker 2: in that department to think about innovating with AI. Another
• 33:17
  Speaker 2: way that I've seen that works well is taking a
• 33:21
  Speaker 2: group of employees, often from different departments, and saying here's
• 33:26
  Speaker 2: an AI tool, use it, see what you can do
• 33:29
  Speaker 2: with it, make yourself into a relational group, a support group,
• 33:35
  Speaker 2: and then from there it means that that person can
• 33:39
  Speaker 2: go on to be that ambassador in that in that department.
• 33:44
  Speaker 1: Mmmm, I mean interested? This is This is maybe a
• 33:49
  Speaker 1: curveball for you, but I think it'll be right, Okay, great.
• 33:52
  Speaker 1: If we were to fast forward like five years in
• 33:55
  Speaker 1: the workplace, specifically, what do you think would surprise us
• 34:01
  Speaker 1: not in terms of what AI is making possible, but
• 34:03
  Speaker 1: in terms of what those guardrails might look like when
• 34:07
  Speaker 1: our workplaces have been our economies, if you will have
• 34:11
  Speaker 1: been through crises coming from its technology, have learnt the
• 34:15
  Speaker 1: hardest lessons have in some areas AI has been contracted,
• 34:21
  Speaker 1: and you know it's not just it's not just extrapolated everywhere.
• 34:25
  Speaker 1: You know, what would there be elements that would would
• 34:29
  Speaker 1: surprise us about caution and reversion in AI as much
• 34:34
  Speaker 1: as its potentialities.
• 34:36
  Speaker 2: Yeah, I think. I think if we see some big failures,
• 34:42
  Speaker 2: then people are going to be are going to pull back.
• 34:48
  Speaker 2: The problem at the moment is everybody has invested so
• 34:52
  Speaker 2: much money in AI that it's hard to think about
• 34:57
  Speaker 2: pulling back, and you know, there's a there's this sort
• 35:00
  Speaker 2: of push. So I was asked by our company the
• 35:03
  Speaker 2: other day to talk to them about being on an
• 35:08
  Speaker 2: advisory board, and they said, we'd like to pick up
• 35:12
  Speaker 2: this conversation with you after we've deployed agents. And I said, well,
• 35:18
  Speaker 2: you know, probably we should have this conversation before we've.
• 35:21
  Speaker 1: Done, or the AI horse has bolted, I'm afraid.
• 35:26
  Speaker 2: And so I do feel that there is a in
• 35:31
  Speaker 2: that case if people begin to have problems with it.
• 35:35
  Speaker 2: What what we're going to see is much more training,
• 35:39
  Speaker 2: it's much more thoughtful. It would be much more thoughtful
• 35:43
  Speaker 2: about where you're using it and how you're using it,
• 35:46
  Speaker 2: So those things about efficiency and resilience would come into play,
• 35:51
  Speaker 2: I think at that stage.
• 35:53
  Speaker 1: And what do you think organizations are perhaps underestimating about
• 35:58
  Speaker 1: the human distinctly human consequences of AI adoption, but beyond
• 36:02
  Speaker 1: these issues of productivity and efficiency.
• 36:05
  Speaker 2: Yeah, I think it's interesting because you know, we are
• 36:08
  Speaker 2: beginning to see people call out what are uniquely human skills,
• 36:18
  Speaker 2: and you see CEOs saying, you know, human attitude, human
• 36:26
  Speaker 2: really the ability to create relationships and working teams is
• 36:32
  Speaker 2: important for as an important human skill. The I think
• 36:39
  Speaker 2: we should add to that the ability to know and
• 36:45
  Speaker 2: be able to check from hallucinations. And then just recently,
• 36:50
  Speaker 2: Daniella Amandai said that one of the things that she
• 36:54
  Speaker 2: felt we should all be doing is reading humanities rather
• 36:57
  Speaker 2: than science and that that you know, sea change and
• 37:07
  Speaker 2: that that would enable us to work more successfully with AI.
• 37:13
  Speaker 1: That's so interesting. Yeah, that's fascinating speaking of the humanity's
• 37:18
  Speaker 1: k just that there's no other way to drop it
• 37:20
  Speaker 1: in but have you read. Have you read or seen
• 37:22
  Speaker 1: the movies Doom, do you?
• 37:27
  Speaker 2: Yeah, Yeah, yeah, I've seen I've read Frank Her quite Yes,
• 37:33
  Speaker 2: I read them, read them many years ago, and I've
• 37:36
  Speaker 2: seen both movies, yes.
• 37:38
  Speaker 1: And the the the the workplace Butlery and You had
• 37:44
  Speaker 1: is approximately maybe a partial one, not a complete one,
• 37:48
  Speaker 1: but it presents it just occasionally. What you were saying
• 37:51
  Speaker 1: makes me think we are we approaching some partial Butler
• 37:55
  Speaker 1: and you had at least in certain areas of life
• 37:58
  Speaker 1: over the next few years.
• 37:59
  Speaker 2: Well, I think you know, the one that we probably
• 38:02
  Speaker 2: should be thinking about, if we're thinking about humanity is
• 38:06
  Speaker 2: the is to use another to use another sci Fi U,
• 38:11
  Speaker 2: the emergence, the emergency medical hologram from Star Trek, And
• 38:17
  Speaker 2: you know, do we want what jobs do we want
• 38:23
  Speaker 2: to keep as human jobs? And so you know, where
• 38:28
  Speaker 2: in medicine, for example, do you deploy an AI and
• 38:33
  Speaker 2: where do you keep the person? And I think that
• 38:38
  Speaker 2: those are the sort of conversations we're really not having.
• 38:41
  Speaker 2: We're sort of blundering ahead with AI without saying akay,
• 38:47
  Speaker 2: so I really like a human in this circumstance. There
• 38:51
  Speaker 2: was a wonderful article in the New York times. I
• 38:57
  Speaker 2: don't know last year where a urgent had said, look,
• 39:01
  Speaker 2: you know, I could get AI to tell people that
• 39:03
  Speaker 2: they're going to die. All I learn is a script
• 39:07
  Speaker 2: and it will probably be more empathic delivering this script
• 39:12
  Speaker 2: than I am. But is that what we want for our.
• 39:16
  Speaker 1: Future personal questions? Kay? Absolutely delighted you were able to
• 39:21
  Speaker 1: finally come on. It's been fascinating talking to you. Where
• 39:24
  Speaker 1: can people find you, your work, your book, et cetera.
• 39:27
  Speaker 1: I know it's so obvious nowadays LinkedIn and.
• 39:29
  Speaker 2: I'm on LinkedIn and the website is just Kfirth Butterfield
• 39:34
  Speaker 2: dot com, so it's really easy. And the book, of
• 39:36
  Speaker 2: course is on Amazon and Bonds, Noble and Waterstern.
• 39:41
  Speaker 1: Anyway, I wrote all yourself, no doubt, Okay, I.
• 39:45
  Speaker 2: Wrote every line it was. It was actually in my
• 39:49
  Speaker 2: contract to that idea. That's it for this week's Deck show.
• 39:57
  Speaker 1: We'll be back next week and if you don't write,
• 40:00
  Speaker 1: be sure to follow the show so you don't miss
• 40:02
  Speaker 1: an episode until next time. M