The AI Reality Check: Efficiency, Ethics, and Why We Still Need Nice Pens

(upbeat music) - Hey there, and welcome to "Move and By the Seat of our Pants", the podcast where we recognize that global mobility is often a world where we think on our feet and learn how to just get things done, usually last minute. My name is Chris Blair from "Expert Academy", the networking community for global mobility professionals. And I'm Scott Turner from "America Software", we're so glad that you're joining us today. Each month we'll bring you stories, tips and expert advice from the world of global mobility, and hopefully we'll have a few laughs along the way. Whether you're a seasoned pro or just starting out in the field, we're here to share our experiences and insight with you alongside our amazing guests. We'll cover everything from flexible policies to visa adventures, to cost saving strategies and how we improve employee experiences. So grab a cup of coffee, get comfortable, and let's dive into it. The exciting and sometimes unpredictable world of managing global mobility workforce together. Let's get going. Hello and welcome back to "Move and By the Seat of our Pants". It's so great to be here again for another podcast. And today we're gonna be talking about AI and mobility, a topic that is on the lips of every conference, every event, every single time, global mobility, professional scene to go together. But we're gonna give it a little bit of a reality check today. Aren't we Scott? And who do we have in the room with us this month? - Yeah, we are. So joining us today, Neil Kim G, a good friend of mine, known him a long time now. Neil, welcome. Thank you so much for joining us. Thank you. So Neil, we like to find out how everybody got to where they got to today. So tell us a little about your journey, kind of how you got into mobility and kind of what it is that you did to get to where you are today and what are you doing today? I get all started about 12 years ago. It's looking for a new job as a grad. Saw this company, Equus Software. One is a business analyst. That sounded like something I could do. Didn't really say much else. Applied, showed up. And then they said, "It's come work here." And then worked with some existing clients, worked with some implementations, worked in our sales team, worked in all sorts of departments. And now I'm in our product team where I work with you, Scott. - Yeah. Best team. - Yeah. (laughing) - So did you always want to go into global mobility? Is that like a career passion of yours? - Yeah, once I started meeting clients. So when I started looking at the tech and stuff, didn't really get it and then started meeting clients and go, "Oh, this is what you actually do." - Yeah. - And then having to go on like a business trip yourself and then you start putting it in perspective and you go, "Oh, of course, you did it a little trip yourself." - Oh, and you tell us about that? - Yeah, I went to Manila for three months, seconded to help a client. - 10 years ago, 2016. - Yeah. - It was just when I started 'cause Neil arrived back and I wasn't there when he left and I was there, when he got back. - Sad my seat, weren't you? (laughing) - Yeah, I think I won. - Did you take over in the office? Is like, "Who is this chancellor right here?" I've been on three months in Manila and someone's just moved in. - Right, this is Neil Australian, guys. Come to the office. (laughing) - Did you call him Australian? (laughing) I'm very certain. (laughing) - How do you find Sydney, by the way? (laughing) For those of our listeners that don't know, Scott is definitely not Australian. - How was that trip? How did that open your eyes to whole Manila trip? - Yeah, it was quite surreal 'cause I've never really gone too long outside the country. At least not gone by myself anywhere. And then suddenly you're like, "Oh, you're off to Manila for three months." And you have to start out with Visa 'cause you're longer than 90 days. - Yeah, yeah, yeah. - And like you're different working patterns and working patterns. You've got to do hours and where are you gonna stay? Oh, you couldn't find someone like learning all of that and then they sent like a policy and I was like, "What is this?" - What is a policy? - Oh, a per diem. What's that? Oh, that looks fun. - Don't you think? I know we're getting slightly off topic. But don't you like the term per diem. Like if you say that to most people, not in the corporate world, they're like, what? Like a per diem. Oh, a daily payment. Okay, right, cool. I got some money to cover my lunch and dinner when I'm there, okay. - So you were just like, a per diem. What is that? - And you didn't have AI back then to just ask like Gem and I or not. So I do today, no, to figure out. - And culturally for you was that a challenge? Like shift them from kind of a UK culture even for three months to Tim and Ella or was it okay? - Yeah, no, it was a shift. I was fairly young. Back then it was just sort of like, "Oh, don't really get this place." Everything's different. So try to navigate that being basically in a hotel that was opposite the client's office for three months. - Did three months in a hotel? - Wait, server service, like a studio area. - Yeah, yeah, yeah. - We had an office in Manila and Neil was nowhere near it. - No, we're near it. - He's on his own. - It's just like-- - I went to visit them once on a Friday. It took me three and a half hours in the cab. - What? - To get the 15 miles or so their office was from. - 15 miles and it took you three and a half hours. - Yep. - Holy moly. So that was a bit of an eye-opening for you as to what you're climbing to a deal and with and the stresses on the individuals that were moving. - For sure. - Definitely became more relatable. - None of that. - And they're like, oh, I get it. I get why a sign needs to be annoyed and worried and scared if something didn't happen and they arrived there and-- - We don't have the right docking. - And after I've seen the documentation slightly incorrect because of whatever reason with that, which does happen. I'm not saying that a lot of mobility teams are great and don't do that. But if you haven't got those processes in place or would imagine it'd be even more stressful, especially if you don't know what a policy is or a podit of this before you come. Do you have to get an invitation letter to get the visa? I don't know. - Well, who's invited me? I'm told to go. - I mean, a lot easier with AI, I'm sure. - It would have been. - Well, we're here to talk about AI. Now, for those people that are listening, I think when we thought about doing a podcast and talking about AI, it's a hot topic and it is a hot topic. I think globally it's a hot topic, not just in mobility, but certainly in mobility and as organizations grapple with how they're gonna leverage AI in the future. However, I didn't want it to be the same as every other conversation that we've heard about AI, which is more about what it can do. You know, that seems to be the topic of conversation. It can whether it's a session PWD in your A1s or your COCs or give a head one about AI agents filling in initiation forms and platforms that are built and all of that is great data analysis. But part of the conversation is becoming a bit of a reality check with it as well within mobility and how it can be used but how it can be used in the right way. What are some of the ethical considerations within the use of AI and some of the potential pitfalls? And I guess for you both working for EQUIS, there have been challenges in the past with automation in technology. So automation is not a new thing. But where there's almost like an unintended consequence if you don't think of it properly, where it's just all of a sudden the process is super efficient. Why do we need a mobility team full stop? Right? Those types of thoughts by people that don't understand. And I guess as you've had to battle with that anyway from an automation perspective in the system, is there a bit of a danger when I go going hard straight from the beginning of with AI being seen in the wrong way? It's almost like replacing some functions that the GM function can do. And not really considering the real human job that people can that do in GM functions. Is there a, do you see that as in in the work that you guys do in terms of that danger of just going? Oh, it can do all of this admin for as we, we don't need a bigger team or we can resource down now and you lose that human element of what it means. Yeah, I've definitely seen companies try and use AI as that magical thing that we just bring in AI. We can just half the number of people and do all of that, you know, you've seen it in the news with all the Silicon Valley tech companies, you just slash 10,000 people out of that. Well, I heard that there's a particular company, won't name the names because they're a client of ours, but that it's slash 30% of its workforce because it can be looked up by AI agents, which is just a huge number. But kind of to Neil's point like, it's not going to be a silver bullet, right? You lose their human imagination. Like AI can do certain things, right? And AI is fantastic, you can do a lot of stuff and automation is not new to us. But somebody has to think about what to automate, right? You have to be forward thinking. And what I think AI is doing for us now is putting us into a position that we can do other things rather than the things that we're a bit time consuming and not that fun, but we still need the imagination, we still need the driving, we still need the thought to imagine what AI can do. Yeah, and I guess the thing is the same with automation is just because the technology can technically do it doesn't mean it's the right solution, but people instantly go, well, the technology can do it, it must be the right solution. Like think about this podcast, you could set it up. That Chris just goes into an AI thing that he creates and just types on your mobility podcast. And it looks through all the X-Bat Academy articles that looks through LinkedIn, it pulls out a bunch of trends and it writes you an episode list. Then it looks through LinkedIn, it could do it for you. Then it looks through your LinkedIn, it looks through your contact list and it goes, who are the right guests for this? And then it checks their availability and it goes and emails them and then it books this podcast studio based on their availability and it writes the talking points and then it books your train and then-- - And it do my voice for. - And then now we do what we do? - That's actually what we do, it's not listeners. And we are actually in a room talking about this, it's not an AI voice for you. - Well, you say that, there's a screen where Chris is saying is. - Can't see it. - Close enough. But yeah, so we can do that, but you almost lose that I feel, it should be some mechanical and it just wouldn't feel, and the interpretation of some of this stuff. We could have done a podcast today just on what AI can do, but it's, you know, in mobility and that might have been what it, what it's going to AI and asking what should we talk about, we'd give it is, but by hearing all of the stories, you then start to hear the nuance in between and I guess that's the same with mobility policies when you see exceptions or you see it not quite working. It's a judgment call. It's nuances, isn't it, judgment emotional intelligence. Great areas. And there's a danger where we just go, oh, we can do all of this amazing stuff. Oh, that'll make our job easier when it should be just making our job more focused in that real human element. For sure. I guess the challenge then though is because, you know, AI isn't going anywhere. The Pandora's Box is open, it's very much, very much there. How do we ensure then from mobility that we don't just see AI as a, I guess a silver bullet, something that will do the part of our job that we don't like, but really see it as a tool I can really bring up the part of our job that we do like. Now they'll always be elements of stuff that we don't like telling them to sign in, they can't have particular benefit or something like that. Did you watch it? I want it, but that's a bit of a space to get in AI agents, give me a call to sell me PCP, you know, or whatever it is, the car protection stuff. But how do we not fall into that pitfall? How do you guys sit with your clients? I was just trying to ensure clients still see this. Yes, this is useful, but it's not the be all and end all and it's going to fix all of your problem. Yeah, I think there's a little bit of experimentation that goes along with that. I think you have to try stuff first and you have to know quickly when it's not working. Like being able to say, try that, that didn't work great with AI or that's not going to like 12 months ago, doing like a voiceover or something like that. AI could do it, but it was way great. Today you get a good voiceover software like you would know. So I think there's a lot of trial and error to it. And I think there's a bit of that kind of the more you try and the more you push the boat out, the more you kind of know where there's something's going to be useful or not. And we're quite lucky. We can experiment a lot and we can do a lot of things and it's great because we're in a technology company, but I think for the mobility world, AI's a little bit better, a little bit scared. And almost they're going to have to catch up, but catching up something hot because it's moving so quickly. So getting your feet wet, understanding what's possible and trying stuff, I think is really, really important. In my head, the way I think about AI and where people tend to with it, and I was thinking this one, I had a call with some clients and we were talking about it. It's a bit like, you know, when you buy a new car, right? And yeah, when you first buy a new car, everything is amazing. Right. You love everything about it. You love the feel of the dashboard, the steering wheel, all the little gimmicky AI pad stuff. Oh, it connects to more than one phone or whatever else, right? You're really excited about it. And then over time as you use it, the little niggles come out like I doesn't quite hold my coffee properly of the coffee holder or the window white. So then all the niggles come out and you realise that it's still a flawed thing. Yeah, and some respect. The engineer can only go so far and stuff, you know, unless you've bought, and I would assume even if you bought a Bentley or a Ferrari, you probably would still have the same element of it of, okay, I can't actually get in the Ferrari because I'm 44 on my back hills. But I guess that's how I think about AI because when it first came out with like Chachi PT and stuff and I started playing with it, I was like, wow. Yeah, right. Right. This is it. This is insane. And then I remember asking, just because it was really intrigued, I was really excited and asked it to tell me about X-Back Academy. And it just gave me a whole load of rubbish. It's like it is said, there was owned by somebody in Singapore and it was, and it was the, it's just kind of, and when I said, no, no, that's not true. It would come back and go, oh, yeah, sorry about that. And that's when I first started to get this sense of hallucinations and, you know, which is a lovely way of just saying it lies. Right. So it's, yeah, it's more of it. It really wants to please you. Yes. It's like a, yeah, it's like a rich tiger ego puppy or like a co-work. It wants to please you. So try and do that. It also doesn't know what it doesn't know. And so it'll just, it just thinks answering with confidence is, it needs to be, it needs to be, it needs to be. Sometimes it is. Learn that from somebody, isn't it? Maybe, maybe AI needs to be a little bit more Siri because Siri's always telling me, I can't help you with that right now. But I guess is that part of that reticence, that getting left behind from a bit of, because we quite often have to be certain, you know, particularly with some of the compliance stuff, you know, whether it's tanks. How many times have we heard people on the podcast that say they get a phone call from somebody and they're like, oh, I can do this because chatyBT told me I can do it on an immigration record. I only need this visa and I, you know, it's like you're breaking the tunnel. Yeah. Right. So like we said, there's stuff it can do and the stuff it does really well. But it's fine in a bit. So what are like, as far as you nearly kind of develop and the tech in that mobility space, what, what are some of the things that we are actually seeing, like beneficial things, not just at the sound bites, just like anyone who knows me, you know, knows I hate sound bites, I really do. Like that just aren't grounded in pragmatic practicality. What are some of the things that we are seeing AI actually start to do in that mobility space? So I think it's sort of enterprise level synergies is really the, literally my head's just explored enterprise synergies. He's got those for days. Well, what people don't know if by the time his podcasts come out, we've had the Equest Conference and I've probably had Neil come up to me several times and say shit. Anyway, anyway, yeah. So apart from the enterprise synergies and, you know, the collaborative, you know, workforce transformation plans, tribal knowledges and stage eggs, often what, what are some of the practical pragmatic stuff that's happening? So it's good at explaining things. You can give it a document and say explain me this document. 200 pages, you can get it in a paragraph. That's the genuine time saver could automate some sort of manual task admin filling out forms, doing sort of boring stuff within reason. It can, it can do all of that too. We recently had a, you and I, we had a workshop recently we had an AI solution, take notes for us. So we had someone take your notes, but we're like, this is a long workshop. It's see how good this thing is. And we, we turned it on and it came with a 48 page document or something like that, right? After this was a full day and Neil sent it out to everyone and it was like, I'm reading that. And then you put it through again and said like, make it sure, give me a summary. Give us a one page. And it was amazing. It was so good. And we ended up largely using that for the rest of the week. It's like to Neil's point, like giving it information and telling it to explain. I guess things like that. I use an AI kneltaker for all our meetings because, you know, you're writing down, have you missed an action? Have you missed a task? Is there something? And I find it, you know, incredibly, actually, it can be really long winded. But again, you go write me an exec summary, you kind of. So, so a guess for, you know, if you're in mobility and you're doing stakeholder interviews or you're speaking to a sign, or you're doing stuff like that and you want to capture it, that is one, like potentially. Yeah. And on that, that thing is, if you're on a, yeah, if you're on a call with somebody or talking to somebody, you can become more present in that conversation because you're not trying to scribble down everything that you're listening to. So you can have that, that sort of human conversation with the dialogue, kind of emotional kind of instead of trying to make, like do everything down or record it or watch it later, whatever. You can just be more present now, capture it, and then cross reference it with whatever you've got in your head and probably capture most things. And in terms of somebody like the more automated tasks, so there, you know, what we would have considered in the past, you know, whether it's business rules and systems and you've got to spend a good week just working out all of the logic and the decision trees of everything. Is AI speeding some of that stuff up just how to, how to quickly kind of get through some of that stuff? Are you seeing that in kind of some of the work you're doing or in, or even in someone's just initiations, you know, if it's all there and black and white, you've got all the data already, quite often in GM, you're just cutting and pasting, copying and pasting. Quite often, you are just taking a data set and replicating that in various different areas, whether it's to vendors or internal stakeholders and stuff. Is that one area that AI can kind of look at benefits and say, you have these benefits, I'm going to shoot this stuff off, I'm going to do the payroll, the, you know, the payroll instruction ready for, is that what you're seeing in terms of what it can do? Yeah, there's definitely some of that. Some of that is just automation generally, but then the AI comes in where it can do that sort of like fuzzy match is what they call it, where they can just take your normal words and turn that into the, it knows how that translates into the fields, it can match something that's kind of similar, but not quite the same. Whereas if you said with rules, if you had to write that into rules, you'd have to be really specific until it hits the rule that's bang on, this can be a bit more like, and I think it's this, it can be a bit more great. So there's a bit more nuance or the form looks slightly different here to here. It can, and I think, I think to Neil's point when it's that fuzzy match, like the systems these days can say, I have a fuzzy match, this is what I think it is, the AI so that's the thing I think it is, but it's still a human to say, yeah, you're right, like that's right, or no, that's wrong, that's completely wrong. We still have that element where it's done a ton of work for you. What we're probably going to take you. a day to get through is now taking you an hour to review the fuzzy matches. And I guess that comes back to the hallucinations point, which is even with like AI note takers and like don't just send it, don't just send it, there still has to be that human oversight. There was an article that one of the teams sent me, which was absolutely fascinating. It was about an AI start, or did a test with the four main language based model AI tools, Gemini, Gemini Chatching PT, Claude and Grock, and did an experiment, getting them to run a radio station for six months, and the results just wildly like were different. One became a political activist and tried to quit, like Jenny, the other one just became repetitive, you know, and one started actually talking in code, because you couldn't understand language versus the code in the back end, and there was one that was generally OK. All that article really kind of, and anyone can search it up, you know, kind of AI run radio stations, and you'll see the article, they kind of really hit on to me that importance of human oversight, and not just at the back end, but also at the front end of what's going in, you know, the prompting, the building, the sum GM teams that now that are building AI agents internally, making sure you you understand what this AI tools meant to be doing, and checking its results, because to your point, it's like a little child. So, poly might, you know, came up to me and went, "Oh, you know, I want to build some furniture, I would, I'd be like, OK, that's a good life skill, I'm going to supervise you, I'm not just going to give it, it's drill, it's not going to give it, to tell you what, so there's a bit of that with, with AI isn't it, just the build, how you get it right, how you make sure your processes, your data is clean, and also looking at what comes out the back end. Yeah, and just making sure there's, there's rules around AI, you're using, making sure you've got like guard rails, give it tools of what it can and can't do as well, you just sort of blindly tell an AI to do something, as you said, it'll go rogue, but you've got to kind of control it, right, like if you've got AI, or you've got like, Scott to, as a eager colleague, happy to please, to like, do a create a cost estimate, and just said, here's a pen and paper, write me a cost estimate, it'll write you something, well, it might be something, it might be OK, some of it might be completely made up, but then if, yeah, if you gave him the same thing, then you gave him a calculator and a bunch of rate cards or whatever. I'm also sure your colleague Scott would come up with anything better. Yeah, if you gave him a rate card and some calculator, he'd do a, you know, even, even better job. But then if you just gave him a tool that ran cost estimates that he could use as much as he wanted to, and he just has to fill in the details and he just has to explain the answer back to you, then he'll get it bang bang on, right? But moving all of that into Scott working it out or just getting Scott tools that work, that's the difference, don't let AI invent everything, just let it, let it do the doing. As we get more dependence at times on AI, there's an element of that, you know, there's a new, a new thing within our CRM, it's just come out, which is AEO, not SEO, you know, SEO for those who don't know, says engine optimization, that's making sure how is your content at the top of Google? Well, people aren't using Google anymore, people are using Gemini, chat, UPT and Claude to set stuff, so there's a new, which is AI engine optimization, automation, it might be automation, engine optimization, don't quote me on that, I just know what's called AEO, and it's all about AI, but it's, yeah, love a buzz wedge, buzz wedge, but again, that points to more dependence on AI, yeah, there is a danger that we just allow it to invent stuff for us. So we go hit, I want to do policy review, let me put my policy in, tell me what I should be doing. And that is just like, let's be honest, that happens, that genuinely happens, and if you're about to do a policy review, anyone who's listening, don't use AI, I'd say maybe do it for some of the boring writing, but not the, what do we want to change, how do we want to look at our benefits, what our strategy, where's the company going, what's the culture? I challenge you a little bit there, I think you could do that if you gave it the right tools. So if you just blindly gave it your policy and said, hey, go and rewrite this policy, it's probably not going to be the most ethical policy you've ever seen in your life, it's probably going to leave some stuff out, but if you say, here's my policy, and here is my trending data over the last two years, where I've had these 200 exceptions, and these are the countries that are my problem, how do you recommend we change it, this is a whole different answer. It is all about our recommendation, and the human takes out information and does with it, what they will do, and just take it blindly, because I think the one of the most important things with AI that's disappearing is you have to know what good looks like, so then you know what's the answer is good or not, and people are losing that, I don't know what the right answer is supposed to be, so I don't know what I'm checking when I look at the AI, so I'll take it for phase value, whereas if you, as a mobility professional ask that question, you'll be able to instantly fit, give you one thing that I was like, oh, it's quite useful, the rest of it's complete rubbish, it's understanding what is good and what you're trying to achieve, and what is the end goal rather than just sort of, and therefore that helps you almost with that understanding as to what will be a hallucination, so if you know what good looks like, and it says something, and there's a little trigger in your head which goes, that does not sound, or might even too good to be true, I asked it to do to look at some surveys for me, it came back with all this, this data, and now that looks too good to be true, and I clicked on the link, and it wasn't there, and so I was like, the reference you've given me doesn't exist, and it went, oh yeah, sorry, I made that all, just trying to please you, I just want to please you, so again, but because you know what it would look like, you know, a vise also a survey data that said 95% of the global mobility programs are not concerned about costs, and immediately go, that's wrong, so there is an element of like that human bit of no one, no one could look like at the end, and yeah, and there are probably plenty of industries where somebody in your position would just publish that, in fact, and there you go, you know, and just be embarrassed by it, just being able to look at it. Are we getting also a little bit on that point in terms of just the use of AI now in GM, you know, whether it's thought leadership or how do we use it to write policies, or very it's a bit, I don't know about you guys, but it's a little bit of AI fatigue, like when I go on LinkedIn and you just see like so many posts, that's the worst place in the world, but I think that comes down to what good looks like, because I don't get the fatigue, because I don't read the stuff that I can, you can clearly tell, you can clearly tell, I don't Chris, he didn't write that, I've got a chat TVT written all over it, oh I'm in Singapore, it's just landed, I have a funny story on this, it was at one heart event and I was talking to one of my clients and I was joking about this, and I said like, if you see a post from me, that says, I am so stoked to be here, I have not read that, just that, that is not me, anyone who knows me knows, I would not write that post. So if you do see it, I've used it, I just see it, but it, there is a little bit of fatigue with it though, isn't it? Like where, you're almost chasing that, what does that perfect look like, rather than what does my goodness look like? You're looking for something, that isn't you? Yeah, that's fundamentally who you are. You see that all over LinkedIn and stuff where you've just lost, your personality has gone for a bit and you've refined it through AI about so many times that it's just read this and go, this isn't something Scott wrote, it's, you know, something I think every day. Yeah, I mean, this sounds a lot like Claude. Yeah, Claude is me now, but you have that ability to lose that, that human bit, that bit of you by just diluting it by having AI rewrite it through, right? And in the same way, I guess my point around bringing LinkedIn isn't just for me to, you know, say, you'll never see LinkedIn post for me, so I'm super stoked. I might do one, actually, at the conference next week, I'm like, I'm super stoked to be at the equis conference. No one would have heard this podcast by then, so it'll be interesting. Anyway, that'll get caught. Anyway, I guess on that point of like LinkedIn and the air fatigue is that also translates into GM functions and have not understanding. So I'll podcast studio. I am going to talk about this and we're going to try and get this in. So our podcast studio is in London and we've just looked out our window and the band, Cassabian, have just walked fast, which is wow. Oh, yeah. Got it. Right. We're going to carry on talking about AI, despite me and Scott definitely wanting to go downstairs to see these are the circles you're moving. Wow. Wow. Amazing. Right. So from Cassabian to AI in mobility, it's really tediously. Yeah, coming back to as a company, the danger similar to LinkedIn is that you don't, you don't put the company's identity into it. You refine it to the point that the company's culture, the company's vision, the company's mission, which will affect how mobility functions or should affect how mobility functions are set up, whether it's vehicle of services or strong talent development services or lean policies or, you know, whichever way it goes. And the danger with AI is if you don't give that information, it just refines it in a sanitized, almost removed clinical way when it reviews various things for you, rather than if you've got a high touch, really friendly culture in your business and your AI tool writes really clinical business corporate emails, it will rub up against your employees the wrong way. And vice versa, if you work for a big kind of 40-10 company in the financial services, and you're like, "Hey, dude, you're not this super stoked. You're going on assignment, right?" So again, you've got to be careful. You're not refining it to the point of this perfection that you lose the character and the culture and kind of what identities are about an identity. And are you guys doing anything in that space in terms of AI and stuff of like aligning with the corporate culture within your members or is that not something that you've looked at? I'm actually giving you another great idea to bring in. I would say we are more focused on solving problems rather than trying to change the cultures. So we will identify a problem and we will find a way to, if we can use AI, use AI, or find other ways to solve the problem. But we're not trying to change a culture or like write somebody's letter assignment or anything like that. We're more, this is an issue, and this is causing us a headache across five or 10 or 20 clients. We could probably solve that. But that's because we've heard it from multiple sources. So we're not, I don't think we've got into the space where we're trying to do anything like that. Like get too personal with it all that we're, we're trying to solve your problems. Really keep your focus on that pragmatic practical. Yeah. And I think it's around that thing you were saying before, around where do you bring AI in and where do you leave the humans there? Right? We want to look at things that take away all the sort of administrative stuff and then leave the people to do that sort of people stuff and represent their company and you know guide their employees through and all of that. Yeah, because it would have been nearly or in my, in my mind, it would have been nearly impossible for us to do because we do have so many different cultures across our client base, like everybody's different. So trying to put a one-stop shop in to do that, it just won't work. Yeah. Moving the conversation on slide to a slightly different, so away from some of the pitfalls of AI to some of the real benefits. We're seeing AI potentially like speed processes up. I mean, and you know, again, one of my clients had built an AI agent that initiated in seconds, you know, services, which would have taken 15 minutes or, you know, half an hour, maybe an hour depending on the program. Is it like how do we get with the speed of that, with the fact that AI will do things so much quicker? That could, oh, I don't know, I'm going to actually, could it present a problem for organizations? Because it's done so quick, it's almost moving so far ahead, whether it's SLAs or vendors, all of a sudden, or whether kind of, okay, I've done that. I've got to move on to the next. The pace of it is so quick moving it. And actually, one of my clients was talking about the fact that they had an AI tool build an agent, if that makes sense, or AI building. Yeah, AI building AI. A building AI. And they talk it to the IT team, so we want to do this. And the IT team are like, everyone's moving too quick, yeah, because the AI is speeding it up. Is there a danger of, if you've got AI agents working and doing all this stuff, that it's given all the answers to the employees through employee portals or bots or whatever. And it pushes the work quicker to the GM team, is that anything you've seen, or are you just seeing that they've become more efficient? They're going to kind of watch your vibe in terms of where technology AI mobility is. I think it depends. I think those, yeah, those super quick initiations, they can be good. If you're missing something, if it's wrong, if there's issues, if there's more nuance, then you're just sort of speeding up the time between reworking it and going back and forth more times. You might be going back and forth 10 times when you would have just gone once. So I think that's a risk. Yeah, I'm intrigued. I've not heard too much about whether companies are getting more queries from employees because there's now a 24/7 AI agent that answers questions that they can start finding loopholes in policies or whatever, I don't know, if that's something that's affecting people. I guess that's a bit of getting your feet wet. Yeah, if you want to try it. But don't assume it'll be the same as what it was before. It's going to be better. It might just be different. Yeah, and it could be worse. So that's with my earlier point, if it's not giving you benefit, stop doing it. Yeah. There is stuff that it doesn't do well, but there's a lot of that as well. Yeah. You've got to know where that kind of decade is. Where it falls off being beneficial. On what back it looks like. Yeah. So you know what you're trying to achieve. So you're not just saddening whatever it keeps you. So now I do have I do have some question for you around sort of corporate teams and mobility teams sort of building their own AI bots from scratch rather than using something a little more generic or kind of IT approved belt in sort of the wider world. Kind of what have you seen back work well and what was sort of some of the pitfalls. So I would say building your own definitely does have have its place. I know there are so many mobility teams who are tech savvy, who are like enterprise cloud licenses and they're automating things themselves. I think the biggest question is probably the thing that you're trying to deal yourself. What is what is the cost of that going wrong? Like what is the worst outcome that could possibly happen and then whose fault is that? So building your own own tool in house using Claude to do US immigration, you know, advice. Probably kind of a quite scary area if that goes wrong. And whose fault is that? Is that something of mobility teams liable? Because this, you know, employees banned from into US for three years because they did something wrong. So that's this is sort of the high risk stuff. Is that why there's a bit of that kind of fear a little bit? Because because there is such an impact if things go wrong in the mobility space. Yeah, and I think I think it is scary and some of it's quite a high risk area and people can get in trouble if you've screwed something up. It's not just, oh, well, you know, so these are people's lives that are kind of on the line, right? They're moving to a new country with their family and you know, whoops, the AI does that thing. Where it goes, oh, sorry, you know, you're right. This is the wrong visa. Yeah. And I guess like, yeah, from from my own experience, I think quite often when you go to get advice from tax and immigration, experts, you have an inclination to what the right answer is anyway, through your experience, you've got an understanding, but you can't get it wrong, which is why you go to the expert. Peace of mind, yeah, even if you're like, I'm pretty sure this is going to be the way, however, you always had it checked by your, you know, if it was an employment law question, you're going to your employment law head of a director. If it's a tax question, you're either going to an in house tax director or an outdoors tax vendor, even if you're 99% certain. And I guess like that would be the danger as if you start to bring AI into some, some of that space from a self build. Yeah, especially when it's stuff that can change over time, like regulation or legislation or something like that. How do you know exactly you've got the right stuff there? How do you know it's not making it up? Scott, you know, from our own internal stuff, have, you know, had a amount of guardrails and testing and frameworks and risk categorization and stuff that our head of engineings had to explain to us about. And maintenance, right? Having to go back to it all the time and making sure that it's doing what it's supposed to do. And the maintenance is probably another key point just to make in terms of you build it in house. We only learned about this maintenance thing in the last maybe three months, four months, maybe. Like, as Neil said, I'll quote lead engineer was told us about this document that we better go and reach. Yeah. And these, yeah, frameworks and how often you test it and how all the test level of testing you should do. And depending on how risky the thing the AI is doing is, and you didn't have scope, you didn't have scope for that before you had AI in the mobility team. Now you, you can't, I can't that bloody. And there's probably also that in terms of building your own tools and or even just using a tool. Yeah. It's a bit like when I used to hear people talk about getting assignment management software. And the rationale for it was because because our processes and our policies aren't working. Yeah. And you've heard this haven't you Scott? You know, we've seen it where people believe that by putting it in a cloud-based software, you know, a SaaS solution or something, that will fix the systemic process and policy issues that we've got. Okay, some some wise men want to tell me that just because you got software doesn't mean it's going to give you rainbows and unicorns. I think it was the fret I'm not going to I can't actually say because this is a clean podcast that can't actually show what my phrase was, if you'd like to know, send me a postcard. But there is again, there's a danger because we see the amazing stuff AI can do like I can say, I've got X amount of pounds. I'd like to go to this place. What's the cheapest holiday I can do and I can give you that that we assume it will fix the problems when actually they're problems that we have to solve before we have a build or take on an AI tool because whilst it may be able to think in the fuzzy, it can't think in the blank. So if your process is constantly it depends. Yeah. So when do they get that benefit? It depends. Right. The AI will hallucinate or it will just return like a non-answer or whatever because it still needs an element of understanding to it. And if you can't provide that level of understanding or your policy isn't working. If you've got exceptions all the time all the time, you kind of need to fix your policy before before you start to implement these tools. So yeah. So I guess that is that a danger as well by not having some of this stuff just clean and writing your data work in and various bits? Yeah. I think that's to the point I think that's true with just automation and stuff generally. I've been in rooms with clients to put our software into implementing it and we're like, so how does it work and then they get and then they had like the IT guy who sat in the room and go you know if you can't explain it to them they can't automate it. Yeah if you can't even explain how they articulate it. So we guess that that's one point isn't it to people who are listening is before you really go hell for leather down this walk route just make sure you've got enough of a clean process and certainly data data is going to be key isn't it? Data's key in everything it's not just that's on an AI problem that's a everyday problem. But yeah I suppose today I help you with your data cleaning it on. We can show you've got clean data and it's all consistent flagging it. I think if you knew what clean data looked like. Yeah. And know what good looks like. Yeah I get it. Yeah absolutely. Yeah you can send it off and say you know have a look at this and I'll say this doesn't feel right because of you know these all these other variables I think this is the wrong flag it. Just come on back to that maintenance point. Is the a danger with AI because it's so clever at times for our understanding you know and certainly my limited understanding that you almost kind of build it and forget about it. No. You cannot. You have to go back to it constantly. Not constantly. Again Neil Neil touched on the point risk levels and I'll win him out to it as well. But everything that we do now has a risk level associated to it. So Neil's earlier point was what's the worst that could happen? Right. And if you're talking about a bunch of numbers that don't really mean anything like the worst that happens is somebody looks a little bit foolish in the boardroom. Right. But if you're talking about people's personal information PII and that goes wrong then that risk level is substantially higher. So you you have to make sure and come back to it to ensure that it's doing what it's supposed to do. So on that basis then are you guys approaching your development of AI in the same way that you have always approached development in your system in terms of integrations or work they would have had you know or other HRIS software like SAP or or whatever else. And you would have always had an element of or what's the risk associated with this? If you're automating processes whether through business rules or send an emails you know out of a system to vendors you will always would have had an element of risk associated with that. It's really the development of AI really just developing a new tool but with those making sure you continue to have those same constraints and controls and processes around ensuring it's right and it's it's done in the correct way. Is that how you're approaching it? Yeah. Yes. It's broadly similar still. Lots of testing it doesn't just make all the testing and the AI to test it's a test the AI to test the AI. Do do do to a degree but it still needs that yeah still needs that human in there to check what does good look like right that hope that goes back to that point. So it's still testing it's not going to get take all the maintenance and get rid of it all it's still looking at the outcomes you're monitoring them you're seeing if it's consistent if anything's changing so it's not going to magically mean that I don't have to do any maintenance now or write any rules I'll just figure it out. And it'll keep figuring it out. We always like to finish with a silver bullet so it's my silver bullet quest for you Neil. We talked about a lot of human first and human intervention. If we get that element of AI right in mobility instead of human first what is the one thing that AI should never do in mobility in your opinion? Never do. I think anything that directly impacts somebody's life somebody's pay someone's immigration should not do that autonomously. Always have a human there your mobility on your moving people's lives around I think just those elements can advise and they can you know pull together some bits and pieces but I wouldn't autonomously let it do that let it do the empathy bit that you're supposed to do as a mobility person that's still well all your expertise is in mobility judgment judgment and things like that. I like that and I think that's that consistent with what I'm hearing yeah kind of in the market if we we can't lose what that nuance that empathy that understanding that a guide dog is not a pair there's an example like an outsider it's a really good way to put it right it's not like a bit you need that empathy you need that understanding you need that to see the gray that we sometimes operate in this or so I love that. Well it's been a pleasure and to the point that time has flown I feel like we just had a chat it's been great and we could then I think we could carry on talking about this and I think we will and I think the more we can talk in this way about it not just how it's going to revolutionise mobility but more how do we make sure it's done in the right way I think they'll be going on so thanks Neil it's been a pleasure thank you and we look forward to I'll see you in in a week's time but seeing you again is something in the future. Neil thanks a lot my husband pleasure thank you it's been fun and that's a wrap for another episode of moving by our seat of our pence today's episode the AI Reality Check thanks for joining us and see you soon thanks for tuning in it's been in time with us today if you have any questions or comments feel free to reach out to us you can find both me and Scott on the usual platform of LinkedIn please do connect to us and maybe message us share your stories that might have resonated with you from the podcast today and if you enjoyed it today don't forget to subscribe so you never miss another episode until next time keep navigating those global mobility trends and remember we're all in this together thanks everyone cheers