How will AI reshape culture?

[Music] Welcome to the CCAB Ethical Leadership Podcast. I'm Tom Parker, and in the second episode of a two-part special on AI, we'll discuss the impact of AI on organizational culture. AI has the potential to be both a positive and negative force on culture and could impact multiple aspects of how we work together. We're going to weigh up the pros and cons and discuss how we can get the best from AI, while avoiding the worst. Joining me in the studio today for this episode are James Barber, Director of Policy at IKAS. Good to be here. Thanks very much, Tom. David Leiford Tilly, Head of Standards and Technical at SIPFA. Thanks for having me. Laura Hough, Director of Trust and Ethics at IECAEW. Great to be here. And social and organizational psychologist, Vika Shulton, from BR Insights. Thank you, Tom. Hello, everybody. Before we get into the AI part of this puzzle, I'd actually like to start with you, Vika. How does organizational culture shape the behavior and outcomes of an organization for better or worse? I think that answer starts with talking about what organizational culture is. So I would define that for the use of this conversation as the sum of how we do things here in an organization and why. So this is about behavioral patterns and drivers. How do we do things in an organization? And there are always aspects of that that help you deliver good outcomes. And there are always aspects of how we do things here that may unintentionally often lead to something that nobody wants. So I think that is a helpful sort of perhaps start of talking about culture that makes it more practical. Because how we do things here is about how you make decisions, how you communicate, how you respond to things that go wrong, and not so much about values and intent and where we want to be. Well, David, how transformative is AI potentially then when it comes to organizational culture? Potentially quite a lot, I think, because it changes the shape of the discussion around what people do. It's a very much a technology, I think, that interacts a lot with trust. I'm, as opposed to I'm an accountant, so I see pretty much everything in terms of trust. But it is about both does the organization trust the individuals that are working there to use it and to use it appropriately? Do the people that work at the organization trust that their bosses aren't going to kick them all out and replace them? Do they have a kind of open conversation about AI and its use in its risk? Or is it more of an open secret or even worse, a closed secret where people think, "Oh, no, we're very good on AI. We don't use it here because that's our policy." Whereas actually, in reality, it's being used and perhaps not used very well. Well, that looks at some of the negatives. I actually want to start looking at some of the positives. So Laura, what can sort of greater efficiency, I think, as a lot of what AI has seen is that it gives efficiency to organizations. So what can greater efficiency that AI can provide us and how does that improve culture? So I was at a conference yesterday and a lady mentioned something called relational offsetting, which is using AI tools to do the kind of heavy lifting of an activity that doesn't require a human input and allowing the human to focus very much on that human interaction. So the example she gave was a conversation between a probation officer and the person on probation and how the individual could really focus in that discussion and the AI could take the minutes, for example, rather than tapping away on a phone while trying to have this in-depth conversation with somebody. So I can see that could create efficiencies but without taking the human out of that situation. And also, I can see in my kind of job, there's so much information I have to read today. I had to read yesterday and I've had to read the whole time I've had to roll and having a tool that could pull together those pieces of information that are really relevant for a particular subject matter would be really helpful for me and be more efficient in my kind of job. And those efficiencies work across lots of different parts of say an organization and an accountancy and financing. There must be lots of different ways that that can give you some data and spare you some time, I assume, as well. Yeah, I think absolutely would save time. But I don't see it as necessarily reducing people's roles but allowing them to focus more on the things where they're adding the value rather than a task that the AI could add the value and actually be better at than a human. I always see that there's something's humans are very good at and some things AI tools are very good at. And we should get that balance right and the mix right and we can make the best use of it and make our organizations as efficient as possible. I think it's also about, this is what I was thinking in terms of the culture, it's also having to have that conversation about what are the things that AI is good at? What are the things that the humans are better at? Because if you don't have a shared organizational understanding of that, I think it's very easy for people to go and try stuff out that maybe isn't the best way to go about it or haven't thought about some of the implications of it. Whereas if I think if you have that open trusting two-way conversation about how you want to use AI in your organization, you both can increase the trusty organization also and make sure you're using it in the best way. Yeah, and if you can free up the mechanical tasks, that should allow us to more actually interact with humans within the organization. And that's where real value, you know, what I was highlighting, the real value can be added by that. And we've been getting the third component of this, we've got the AI. So we can have humans interact with the AI. One of the best uses of the AI I have is just trying to break the blank page syndrome. So it doesn't need to be perfect, but it just gives you a starting point. And it's far easier to criticize something once it's there than actually starting it. So I think it's a lot of value to be added. Yes, it can have an impact in culture, but overall we need to make sure that it doesn't over-react and push into place we don't want to be. The culture has to be set at the top of the organization, not by machine, by humans. They then have to make sure that's replicated through the organization. To the extent that AI seems helping us, and what we do, we can never override that human factor. Yeah, there's a great point there about leadership. So Vika, how important is leadership when it comes to shaping that organizational culture, especially around AI? Yeah, so I think there are different ways to look at that. So to your point, James, Tom from the top, have what the, let's say, the executive team or the leadership, senior leadership teams send out in terms of cultural direction, and that is, of course, important, because it helps people to understand this is where we want to go. However, what we also know from behavioral sciences, where culture really shapes itself, is at that shop floor, call-face level, where the daily operational reality, so that leadership, so direct line management that you receive, what people see in what we do together, and how a direct line manager, for example, contextualizes, why we use this AI tool in this case and why not, and having that initiating that discussion for direct line management is very important. It's almost like, I don't know if you know the movie Silence of the Lams, with Hannibal Lecter, and he has this quote, "You learn to love what you see every day." Hopefully only a little bit like Hannibal Lecter. I think there's a point as well, isn't there, about how the AI will be used in the way that other things are already used in the organisation. So if you've got an open culture where people discuss things, challenge things, say, "Oh, I've got this new innovative idea, shall we use it?" That's what will happen. But if the culture is, keep your head down, do your work, and go home as quickly as you can, then that's what will happen with the AI tool, so when they're used. It just can be a force multiplier. It exaggerates, perhaps, what's already there. There are some parts of it, I think, which maybe cause changes, but a lot of it is just about doing the things we've always done more quickly, and more, perhaps more efficiently, and more. So if you've got perhaps some cultural issues, it can blow those up. If you've got some cultural strengths, I think it can reinforce those as well. I think I want to build out, on James's point, which is around it, it frees up more human-to-human interaction. What's the benefit of having more human-to-human interaction when it comes to, say, creativity, or is it the water-cuna chats or anything like that? How important is that for organisational culture? Very, because we're social creatures, so the way we behave, and if culture shows itself in our behaviour at work, then the way we behave is very much shaped by our direct social environment as well. So those water-cuna conversations are essential in shaping culture in that daily operation reality. And I think what an interesting point is perhaps to think about as well is we talk about here at the table, open culture. I have an open communication culture. So if you really have a culture where people speak out and speak up and share what they think, that's great. And most people want to do that, actually. We know that also. The people want to do the right thing, and people want to belong where they work and be open about what they have in terms of internal dialogue. However, there are always situations where they don't. And I think that is maybe, instead of saying, do we have an open culture, saying, where is it most challenging, actually, to express that you are not sure about what you're currently doing with AI, for example? That's absolutely key. And for me, if people speak up early and feel comfortable doing it, that saves disasters at the end of the line. Let's solve the problems at an early stage, and it's all for the benefit of the organisation that people and investors, whoever it may be. But if we do it, then these things tend to fester. It leads to bad, organisational culture might mean a lot of people leave, good people, good talent, and you don't want that. So if we can get that in place, proper speak-up, listening-up systems, and those who are tasked with listening up have to listen. It doesn't mean that the end of the day someone is disciplined, but they have to investigate and see, look into, what has been put before them so that people trust the system. And if you get that trust and you get that culture installed within an organization, I think it can be really, really beneficial. Yeah, I want to look at a little bit about upskilling here as well, because there have been quotes banded around about the fact that AI isn't coming to take your job, but it will come for those that haven't used AI and don't understand AI. And I wonder what this means for upskilling and professional development within organizations, Laura. How can AI potentially help that upskilling, but also how important is upskilling to using AI and becoming part of your day to day? So I think AI could be a very powerful tool in terms of training people, tailoring, training to them, providing them access to the knowledge that other people already have in a way. You know, recordings of training sessions or recordings of other people's meetings. They can easily access those now, which didn't used to be so easy. I think we'll need to find a way to train the future workforce to critically assess what AI produces for them. So maybe the best way in the future will be to get, I don't know, an AI tool to prepare the financial statements for you, but what's your role then as an accountant critically assessing those, finding out whether weaknesses are validating them, just as sort of different lens, I think. And that's something that people say other times almost at least, say human in the loop, you do have somebody involved in that, but that's got to be somebody meaningfully in the loop. So there has to be somebody who's actually trained and understands what the AI is doing and what it's good and bad at, who's empowered to actually challenge it and override it if they think it's made a bad decision and who's had the organizational support to do that. It can't be a human in the loop in the sense of well, the AI says to press the red button, the person presses the red button. There has to be some actual real two-way empowerment. And that is, I think, of case we're upskilling and that's exactly the kind of area that I think would be really valuable is to, if you're going to have that human involvement in those processes, the challenge I think, you know, for many of us here speaking on behalf of professional accounting organizations is, okay, our current workforce probably has the knowledge and the expertise to challenge those AI's doing those things because they've gained that experience in a pre-AI world. 10 years from now, how do we get people who have only ever worked in an AI world to get to that level of skill if the things that we cut our teeth on now, nobody's doing because AI does them. So that's going to be a challenge for us for the future, I think. Yeah, how do we solve that? I mean, if we're looking at the path of education, I mean, we still teach kids in schools in a traditional sort of examination way of retaining knowledge and doing maths in your head and things like that. When we've got a technology that does it all for us, where's our empathy being taught throughout educational formats and then into, you know, the next generation of workers? Maybe we can look at that a bit. We need to look at the aviation industry. The pilot doesn't fly the plane. The copilot does automatic copilot, but the pilot could still fly the plane and we'll land and take the plane off. So they use visual simulators and so forth to train. And we only to look at more of that, how can we create some of the environments artificially that will allow that sort of training to be embedded. And, you know, a lot of consideration we need to be given to that. And the whole case study type scenario can come alive because when you think about ethics and what can go wrong, it's a disaster stories that leave, you know, the real sort of visions. And that's what you want from people is the visions. Because if you think of something, have you always had people thinking before they act, how could I appear before my peers, before a parliamentary committee or on use night to defend my actions? Have you always had that at the front of mind? It's usually a good way of getting them to do the right thing. We touched a little bit on this on the first episode of "Around Trust" and around this accuracy of AI as well. I mean, if you've got a human in the loop and the experience of that human is going to make a decision based on the data that the AI has given you, how do we make sure that we are having the right data, the clean, the correct data that is fueling that system, and then the right human to accurately tell whether that information is correct? I'm thinking about outside of the accountancy industry about medicine. You have pixel AI, which looks at melanoma. And unfortunately, the data that it's been built on over years and years has been fundamentally about Caucasian images. And we don't have the pictures to train the AI models on. So you need an experienced doctor to then say, well, yes, even though that has a greater accuracy of 97% compared to the human of 96%, I'm actually still missing out a large part of the population that could tell whether it's melanoma or not. How do we equate that to accountancy and finance? How do we equate that to the human experience versus the efficiency and credibility of an accuracy of AI open to the floor? So I think with AI Trust so much of this is about getting the right data to get the right input to get to what we want. We have to make sure that the organisations that we're working with have the data that we need in a format that can be used. But it's also about what the AI systems that we're using have been trained on. Large language models and the way that they work now are very probabilistic in how they work, which means that they're really, really good at things like generating speech, where it's a lot more open to pattern interpretation, but aren't so good at things with a definitive answer, mathematical things, where there's a specific right answer, because it will give a probabilistic answer. It'll do, you know, if you ask it to generate a random number, it gives us seven more often than the other options, because humans gives heavens more often than the other options from one to 10. So it's just replicating that pattern that it's seen. And while the systems are getting better at that, they're not there yet, and we still have some work to do on getting to that right point. So understanding things like, well, what data does this system have access to? In what way was it trained? What things is it missing? What errors and omissions, like you were saying, with having non-white skin pictures to work on for medical information? What's the equivalent of data that's missing that it doesn't know about? Yeah, if we're done with the rate data, how are we going to get the rate output? So it's the old garbage in, garbage out. And so it's absolutely essential that you know what data you have. Is it the right data for what you're trying to achieve? And that to an extent will fall upon those in the accountancy function within an organization? Because if you look at sustainability and emissions, no, those data coming through from that, but is the data accurate? And that will evolve, but AI tools will be there to assist as well. But it really comes back to accountants can take on this role of governing the data, because data guardians, an essential role, I think, has been moved forward. And it won't just be in sustainability. There'll be more and more comes on the role of CFOs accountants, but the underlying skills that they have in the train then will serve them very, very well. And that comes back to trust and the fundamental ethics that is installed in all accountants. I think there's a question isn't there as well about what we're willing to accept as our risk tolerance. You know, we're thinking of driverless cars earlier. And even if they have fewer accidents than human drivers do, we are not prepared to go with any accidents as being a tolerable level of risk in that situation. And that'll be different for every different situation, different type of data, different activity. So we need to get people involved very early on, I think, in designing these things from a range of different backgrounds, not just programmers, but people that do jobs like us, or different aspects to bring in on those design of the tools before they become products. That understanding of risk I think is really important, because this also goes to that same organizational, cultural discussion about the use of AI is to understand, well, what is an acceptable use case and what is a really high risk case where actually the AI should not be involved or should be involved with a very thorough review, very involved oversight, because it's really, really important and we can't afford to be even a little bit wrong. We're perfectly segueing into the negatives of positives. So thanks everyone for that. So Vika, I want to ask about some of these behavioral risks when it comes to the adoption of AI. I think behavioral risk is actually it's a term to talk about what are poor outcomes that can be driven by aspects of the way we do things here. And I think introducing something new, let's say to certain teams that have not worked with AI and start doing that, that it's indeed also important to look at the risks from a behavioral angle. So where are our vulnerabilities in when we come to decisions, or are we, for example, under high commercial pressure that creates a certain, let's say, incentive to take a shortcut in some way. And not intentionally, again, have we said earlier, often unethical acts in at work are not because there's lots of male intent, but it's just factors in people's daily working environment that encourage something that is to be seen unethical. And I think introduction of something new like AI, we tend to focus on all those positives, but we simply just have to balance that with, okay, and where does that may not work. So if we have indeed open communication, where are we, where do we not speak out? Or yeah, we always weigh the risks, but where do we not? And why? And maybe commercial pressure could be one of them, but there are a range of factors, I think. Yeah, it also I think goes to what the incentives in the culture are because so for we were talking earlier about trust of do I trust that the the management level is going to leave my job intact? Because I've seen this happen when people say, I've written this computer argument, or now I've set up this AI thing that automated a big part of my job. How do I stop my boss finding out? Because I think if they know that I've done that, am I going, they're going to just fire me? And if they that shows, I think of a poor culture, or they should be saying, oh, I'm really pleased to tell this because my boss is going to be super happy with me and promote me and give me into other things. So it's at that difference of how people react to that situation and what they expect to happen if they say, oh, I've automated a big chunk of my job, is a good example of what perhaps the culture is this thing. Because I wonder whether it then moves a little bit more towards that the longer we are in this process of of a interaction between artificial and human. Because I think maybe a year or so ago, if you talked about chat GPT having written a part of your bit, then bosses may look and say, do we trust that technology yet? And also I've paying you for eight or 10 hours a day, whatever it might be. And you've done it in three and then gone and walked the dog and then gone to the bar or whatever, that does look bad. But if it frees up more time for you to be better at your job, then I think that's where we are now, right in this conversation. - Yeah, but going back to the list, there's a list of automating bias. The people just accept what's coming out of the eye without properly checking. We've seen hallucinations appearing in reports and well, publicized stories there. And so we have to be careful in that. And it goes back a bit to me to the culture points we're making earlier, but they speak up. If someone within an organization has a sense that the AI is not working appropriately, they should feel safe to go and report that. Regardless of how much has been spent and developing that particular tool. Because if they don't, there could be severe repercussions further down the lines. It goes back to that organizational culture, I think being really, really important. Yes, we want to get the benefits, but we need to make sure appropriate safe guards are in place. - Yeah, so looking at those safeguards, 'cause around data security around bias is, how, with such a fast evolving technology, is changing all the time, regulations are coming after things have happened. How do we feel about creating this secure ideology within an organization that the tools we are using are essentially our friend, they are gonna help us, but we do still just have to have a bit of checks and balances on the information that's coming out. - Yeah, I mean, I, again, Tyler, it comes back to that balance point. It's all a bit balanced. We can only really explore the opportunities as far as I'm concerned, if we have the appropriate guardrails in place and people feel safe. For example, they're using enterprise systems so the data's not gonna be inadvertently leaked and cause major issues there. So the use of AI should be encouraged, but within certain parameters. And, you know, as was mentioned earlier about the more risky cases than that really has to be subject to greater controls and so forth. So I'm not saying good on the route of the AI act and a sense they're already coming back, but to me a lot of this is back to personal responsibility and accountability. People also need to play their part in this 'cause that really is important, as well as having the governance and the controls, professional candidates are taught, you know, this output is yours. It's not the AI. If you're gonna sign off in this ultimate you have to take responsibility. And I think that's really important. But I also think that's fair, also that is socially driven. So of work that has been conducted in accountancy firms where every individual accountant has that professional integrity and pride. Then when we work together in that daily working life there can still be social factors that let's say distract us from that sort of internal moral compass. And what we do know is that the group moral compass over rules individual moral compass even though we think that's not the case and it would be great if it wouldn't be because then it's a lot easier because then we can just tell ourselves this is how we want to behave and then we'll do that. But we know at work it doesn't work that way. So that means I think you need a continuous management of those behavioral risks to always be curious about and where does the desired culture that we have in terms of for example feeling safe. Where are the situations in our organization where people do not feel safe because they will always be there and be curious about that instead of just saying this is what we want to be. - The way I think about this is that we've gone through industrial revolutions before and the workforce has changed. Farmers to factories and things like that and everyone thinks that there's gonna be negatives and there are negatives that come with it. But essentially you retrain and you remove the organization going back again to how quickly AI has come into it. In fact, how quickly technology has changed. I'm thinking pre-pandemic to post-pandemic. I mean, we are working in a hybrid way before it was everyone in the office all interacting together, lots of team interactions that took away time. Where are we now I think can I ask everyone compared to we were in the last couple of years and then maybe where we're gonna be in the next two years 'cause everything is moving so quickly. - Sometimes people phrase this as being in the posts singularity world. The pace of how quickly technology changes at some point passes is the ability of human beings to keep up understanding what's happening and the changes in it. And I don't think we're at that point yet but at some point it becomes that you can't necessarily always know what the latest technology is and is capable of and it does change so quickly. I think we've seen this huge shift, as you say, I think the remote hybrid change over the pandemic this sort of natural experiment that caused everybody having to try doing that 'cause there was no choice and actually finding actually in many cases that it worked well for people and that many organizations have not sprung all the way back to where they were beforehand. They've reached some sort of new balance. But we still are figuring out the kind of cultural implications of that. They still have, I think managing and training of new people and team building that practice is based on a very, very long time of working together in the same location most of the time that we are still figuring out how to make that work in a world where that happens some of the time and for some people doesn't really happen much at all. My own team in my organization is spread across the UK, we've got one in all the line, got one in Oslo, they've spread about a bit. So we have to think about these questions and I think that's pretty typical now and I think has some real benefits but it does require some extra thought as well. I think I worry most about people who are new to the workforce. So the first job out of university or the first job out of a training contract or even during that training experience as an accountant those three years are very formative and doing all of that work on your own in your bedroom at home is not really the same as interacting with colleagues of making friends, learning from older colleagues as well. So I think we have to really give that proper thought on consideration. James Fika, anything to add? I think we really need to work out what is best for the organisation and but I mean, there are a bit of certain people that's better working from home most of the time, certain people better than the office, depends on the role. And I think organisations need to accept that rather than force one particular way of working. Now there might be certain businesses have to work in one particular way but I do think it really is trying to get the best for the organisation sitting down and saying this is how we can best come together and I think it will be different for every organisation. I don't think there's one answer. So then perhaps one final point from everyone is if you had some advice to give to a leadership of an organisation on how to best implement AI, maintain good company culture, organisational culture, what would it be? I think it would be a combination of and setting the standard in terms of this is how we want to use AI and this is where we don't because of these reasons, so that contextualisation and setting direction and pairing that with a continuous curiosity about daily situations where that is hard to do. So where it's challenging, where we feel to do it in the right way, where things pop up that we said we would never do. It's that type of curiosity about those daily situations and learning from that. I'm always a big advocate that people should have some sort of an AI use policy, something accessible, non-technological, something that's useful for everybody because I think otherwise you have the default do whatever you like, policy that you start out with and that has to be also evolving but you can also set some bright lines. You might say, okay, for example, we don't think that using AI generated art instead of commissioning artists is an ethical thing to do so that we are not going to do that. That's the sort of line that you could set for your organisation and explaining why that, you've made that decision, I think helps to both set the line and build the culture. I think it's important to take everyone on the journey of using those tools with you right from the bottom of the organisation, explain to them why you're bringing in certain things and not others because if you just try to set something from the top, it's very hard to filter it all the way through to everybody. I believe the use of AI tools will follow the values and culture of the organisation that are already there so in a sense you almost need to sort those out first and then build the AI in afterwards. For me, it's build the strong governance controls, ethics foundations and then let's explore and not just about efficiency savings. What is there that we could potentially do whatever they've done before? So explore real business opportunities and to do that I think it's good to get different age groups together within an organisation and really explore and challenge why have we never done this before? And I think if you do that, you can open up potential new sort of areas that things could be developed for the better of the business. It gets people again together, culture wise, very positive. So I think it is less of a safe environment and then let's explore the universe. And for me, that's where we're going. Well, one last very quick question because I'm fascinated about this. If we use AI more and more and more and it makes us more efficient and it makes us better at being humans, are we going to get down to a three day working week? Is it a two day working week? Are we still going to be this industrial nine to five five days of the week? What do you think? There's a question about productivity that isn't there rather than just efficiency. So perhaps by being more efficient on these transactional tasks for one to a better phrase, we may become more productive. So maybe not fewer working hours, but more productive working time. Yeah, perhaps we will also have more time to think about quality because that's a thing with efficiency and quality, right, in terms of trade off. And I think I can't imagine that it will be a three day. That's my conclusion. I mean, I suppose full disclosure, I work for an organization has a four day working week already. But I do think that there's a point that, you know, work expands to fill whatever space we give for it. So I think there's a room for us to do more, if we're more productive, you know, as you say, more efficient. But then maybe that means that, you know, so much of what we do, I think tends to be deadline driven and in the acting to the immediate, and then it perhaps gives us some time and some space to try to think about the longer term and accomplish more as well. And I think you always have to consider, we spend a lot of time at work, and I don't always see work as a negative. There's a social environment, you like the opportunity to think as well. I've seen a lot of people who retire and can't cope. We couldn't get used to this having all this time. So yes, we really need to focus on wellbeing, and if we can reduce our lives great. But there's also I think we have to just make that workplace as good as possible. So it's an environment people get in every day, I'm not dreading it. But I say this is an environment I want to gain, and we can do a lot of things to make society better. Well, another fascinating conversation, we are at the end of the episode though, so I'd like to thank all of our guests. Thank you to James Barber. Thank you very much Thomas, we're in a pleasure. Thank you to David Lightford Tilly. Thank you to Laura Hough. Thank you so much. And thank you to Fika Shulton. Thank you. Well, that's all for this special on AI. If you've found these podcasts useful, let us know by getting in touch or by following this podcast on your favourite app. We may well make more of these if you ask for it. There are links to a number of useful resources in the show notes that will help you determine how to approach AI from a cultural perspective. So be sure to check those out. Thanks for So, listening, bye for now.