Thomas: 00:00
Hello, my name's Thomas, and you're listening to the Pensions and AI Podcast, the show where we explore the way that generative AI is being used across the pensions industry. In every episode, we talk to people in pensions who are working at the frontier of generative AI. You'll hear from super users, builders, advisors, and big thinkers, the people who are shaping the future of how our industry could look and how it can deliver better outcomes for members. This is not a show for people to come and sell you things. Far from it. We're here to share, to have open and honest conversations, to talk about what's really going on, what actually works, and what lessons we're all learning along the way. If we're doing a good job, then you'll leave each episode feeling more informed and confident with generative AI, and better equipped to use the technology both yourself and to shape how it's used in your organization. In this episode, we're going to be exploring how pensions lawyers are thinking about and using generative AI. And to help bring that to life, I'm joined by two brilliant brains from the pensions team at the law firm Burges Salmon. Our first guest is Chris Brown, a partner in the team who advises trustees and companies on all pensions legal issues, and is also a co-host of the Burges Salmon Pensions Pod. And we're also joined by Callum Duckmanton, a solicitor who advises on a wide range of pensions matters and also pensions tech, in particular AI and cyber security. Chris and Callum, welcome to the pod.

Callum: 01:22
Thank you very much, Tom. Yeah, really excited to be here.

Chris: 01:25
Yeah, thanks, Thomas. Thanks ever so much for having us on. And really nice of you to shout out the Burges Salmon pensions pod. Actually, it's really exciting being on another industry podcast and really looking forward to the conversation today.

Thomas: 01:38
Awesome. So for people that haven't met you or haven't had the chance to work with you before, tell us a bit about you and about the team at Burges Salmon.

Callum: 01:46
Yep, so I'm Callum, solicitor in the Burges Salmon pensions team. I've been in the team two or three years now, and like Thomas said, advise on a wide range of pensions law matters, in particular most recently with a focus on pensions tech, AI, and cybersecurity.

Chris: 02:08
My name's Chris Brown. I'm a partner in the Burges Salmon pensions team. And like you said, I advise on all aspects of pensions law, mostly in the defined benefit space, but in the defined contribution space too, for trustees, but also employers, and advising employers on risk transfers and and big projects is a large part of my work. But as Callum says, increasingly we are seeing trustees and other stakeholders in the industry interested in the interaction between AI and pensions, and that's taking up a lot of my work recently.

Callum: 02:43
Yeah, and I think just in the two or so years I've been a qualified solicitor, the amount of pensions AI work that we've been doing has kind of skyrocketed from a point where you know it was one, two percent of the work I'm I'm doing to now probably 20 plus percent of the the work I'm doing. And we're speaking to different clients, different bodies every single week. So yeah, it's a it's a really active space at the moment.

Thomas: 03:09
Brilliant. I'm really excited to hear about the work you're doing with clients. I want to start first, though, with how you're doing the work. So we've had tools like ChatGPT, Copilot for about three years or so now. I'm curious, in terms of you actually doing your jobs, what's different now compared to a couple of years ago?

Callum: 03:27
Yeah, I wouldn't say it's massively different yet. I wouldn't say it's been transformative for us yet. For me, anyway, I normally use AI as the first kind of port of call when I've got a bit of legal research to do, rather than going to Google or one of the legal databases that we normally use.

Chris: 03:46
Yeah, absolutely. I'd echo that. I think it's really useful for research. First port of call, as Callum says, but also the last port of call. So maybe rephrasing something we've created or something like that. But I think also in the way Burges Salmon uses AI in our client work, there has to be that human element in all of our advice as well.

Thomas: 04:11
So, in terms of what you're using, does that mean you're asking legal questions to something like Copilot, or are you using something else?

Callum: 04:18
So both. So we use Copilot, but then we also have a legal specific tool called Harvey. So Harvey has been trained specifically to be more like a lawyer. So when you ask Harvey a technical legal pensions question, you will see it be a bit more technical in its thought. You'll see the kind of workflow and thought process, and it does typically provide a better answer than Copilot. But Copilot does still have its uses for us. So, like Chris mentioned, something like where you're trying to rephrase something, it feels to me like Copilot is still probably the best tool for when you're trying to make something a bit more human, a bit more concise, whereas Harvey can be a bit more technical than sometimes you need it to be.

Thomas: 05:04
Do you think you get the best out of a tool like Harvey from being a lawyer? Or is that the kind of thing that a trustee could pick up and use to ask legal questions to?

Callum: 05:12
I think I think it it's definitely been produced to focus on law. And then there'll be the the licensing requirements, which I suspect will probably mean it is best suited, best suited for lawyers. Yeah.

Thomas: 05:25
Because I imagine in the kinds of questions you're asking and the way that you're thinking, the things you're prompting and the things you're exploring, I guess you're coming at those from a specific angle in a certain way as lawyers, right?

Callum: 05:37
Yeah, yeah, definitely. Most of the questions we get asked, you kind of go through the same thought process of can they do it under the rules, can they do it under the statute, and then should they when you're looking at it from a fiduciary point of view?

Thomas: 05:50
So using AI to help you with research, to explore client questions, using it to rephrase and kind of repackage what you're trying to say. When you look at the uses of AI across your team and the way that people are using it, what are some of the best things that you're seeing them use it for?

Callum: 06:07
Yeah, so in our team, we have weekly meetings where everyone attends, and at each of those meetings we have something called prompt of the week or just generally use case of the week. So we share all the favourite use cases from the team and bring them together. One of my favourite ones from that is kind of bringing together an executive summary of an advice note. So you draft an advice note, might be a few pages long, and then at the top you have the executive summary, and the AI is quite good at bringing all of it together into maybe a few sentences, a paragraph. The other one is quite a technical one. So sometimes Acts of Parliament will refer to as prescribed by regulations, but then there's no reference to where those regulations are, what regulation within the regulations there are. Whereas AI is actually quite good at finding what is being referred to there. So that really can improve how efficient we are with a piece of work.

Thomas: 07:08
Yeah, and I love that. And that frees up even more time and brain power for you to be better lawyers.

Chris: 07:13
Yes. And and we might come on to it in a bit, Thomas, but sort of thinking about what is AI good at to help lawyers. Certainly what it doesn't do is the sort of human element, the strategic element. And yeah, if you can free up time for us to do more of that, all the better.

Thomas: 07:31
So if you were going to take your job and just kind of break it down into the different tasks that like you as a lawyer would do over the course of the week, what are the big buckets of things that you do?

Chris: 07:40
So we will advise clients, we'll meet clients. A lot of my role is as a partner is management. So that might be internal management, but it might be relationship management with our clients and internal management will encompass all sorts of different things around the people, the finances, the way our business works. I'm sure I'm forgetting a million and one different things that I do in my job.

Thomas: 08:08
Yeah.

Thomas: 08:09
Callum, how about you?

Callum: 08:10
Yeah, I would I would I mean from my perspective, it's probably quite different to Chris doing the kind of first drafts of things. So a lot of it is advisory. So the the the the scheme wants to do something is is asking whether something that it's done is is correct and whether it needs to take any rectification. So a lot of advice in that way. Some transactional, so contractual stuff. So if there's a risk transfer project, buy-in, buy out, things like that would be would be a big thing as well. And then the last part is probably if there's a bit of a dispute, whether that's between the member and the pension scheme or the pension scheme and potentially a service provider. So yeah, the the the whole range.

Chris: 08:53
Where where I use AI at the moment is just because of the nature of my role, is not so much with producing the first draft of a piece of written advice. But I find it really useful where I know the answer to something, but I might not know the precise regulation it's in, or I might not know the precise requirements for, you know, say a notice period or whatever it is. And actually I've found that Copilot not so much, but Harvey in particular has been useful for that. Where I've known the answer, but it's just saved me time.

Thomas: 09:21
So this this just gets me thinking about where is the line between human and machine in your work? Where do you see the line right now?

Chris: 09:29
So being a lawyer is not just about dry written law. It's actually about relationships and helping people solve problems and meet their objectives. That's the bit that excites me about the job, and that's the bit that computers and AI can't do. And so thinking about strategy, thinking about creative thinking, that's that's where the line occurs for me. And actually, there was another Clara transaction announced yesterday, the fifth one, I think. So we've done one of the five, and the one we were involved with for the Church Mission Society was the first one with a connected covenant with the employer. So that's unusual, never been done before, and that's not something that AI would come up with because that is a novel transaction structure. And so all the sorts of things we're doing, thinking about different pension products, different alternatives to end game solutions, that's something that needs a human mind.

Callum: 10:29
Yeah, completely agreed. Because I haven't been in the pension industry too long, but from what I understand, the end goal, the gold standard has always been insurance, get into buyout. And now we're at a point where most schemes are fully funded, it feels like the industry is starting to question that a bit. So we're going into partially you know, uncharted territory, and it feels like AI won't be the best place to comment on that because everything it finds online will probably be about buyout being the gold standard. Whereas now we are seeing different things being explored, such as Clara, such as capital-backed investments. Exploring the new things maybe AI won't be as good at because it just doesn't have the materials to advise on that.

Chris: 11:18
And also one thing that really strikes me when I use AI tools is to do with tone and the way things are presented. So often there is no answer, but there's a range of legal answers, and so AI has a habit of presenting something as the answer, and you know that that may not be the case. You might come up with a perfectly suitable solution for your client, but that isn't the one. The other thing that springs to mind is it might be the opposite, and it might be that everybody knows what the answer is, but where the legal value comes in is how you get there. And that is something that I think requires a strategic mind and expertise, and that's what lawyers can bring to supplement all the help we'll get from these new AI tools.

Callum: 12:09
There's also a partially philosophical kind of question in looking at the development of juniors, and when I think about a few of the the pension schemes that I work the most on, I know how those like trustee board you know functions, what their priorities are, how they view their fiduciary duties, things like that. But if I was reliant on AI, I wouldn't know all of that information. And whilst now I could probably prompt the AI to you know explain what the client tends to prefer, down the line that may be more difficult if we start to rely on AI too much.

Thomas: 12:53
Yeah, this does get me thinking about how your clients might be using AI. Because I I know from my personal experience that AI is brilliant at things that I don't know too much about. So if I was going to spin up a quick draft of a website or a prototype or do some quick data analysis, you know, I can now do that really quickly and look at an output which feels really compelling. I know at the same time though, if I use AI for things that I know lots about, I look at the results and go, it's not it's not really that good.

Callum: 13:25
Yeah, absolutely. So pension industry generally, I think there's a lot of usage at the moment. So especially from the governance point of view for trustee boards, so member communications, can we use AI to produce maybe avatar speaking, videos, maybe videos in a more efficient way than you would be able to at the moment to improve member engagement. The other one's probably recording trustee meeting minutes. So what can be quite an expensive and long process if it's a several hour-long meeting, you can really have it be more efficient if AI takes the first draft of those, if it transcribes, if it records that. Although there are quite a lot of risks involved with using AI in that way as well. So this is something that we've been speaking to trustee boards a lot about. So for example, there's a DSAR risk, so data subject access request. So if a member was to submit a DSAR and their name's been mentioned in the AI transcript, then they may need to be provided with that transcript. Potentially, it may be able to be redacted in places, but you may still have to disclose that to them. If anything was discussed about a litigation that was going on, you may have to preserve the transcript for that purpose. And then also if any legal advice was given during the meeting, you could potentially risk waiving that legal privilege over the advice. So there's a lot of risk because of AI taking everything word for word, whereas generally we'll see meeting minutes being taken from a more high-level perspective.

Thomas: 15:11
Gotcha. That's so interesting because it feels like almost every meeting we walk into nowadays is recorded and transcribed, and Teams wants to do that by default out of the box. I never would have thought of that.

Callum: 15:20
Yeah, so the one of the worst kind of scenarios when you're kind of war gaming this is what if an ill-health early retirement case is being discussed by the trustees. If a human was transcribing, recording what was happening, they may exercise an element of discretion and use their judgment from years and years of taking meeting minutes. Whereas the AI will just take it word for word or will record the entire thing. If that member then has their request rejected, they may put in a data subject access request or maybe submit an IDRP internal dispute resolution procedure, and the trustees may have to disclose that entire transcript, which then may result in that case, which could have been dealt with quite simply, being a bit more complex, potentially going to the Ombudsman. And you know, that's one of the things that that we think trustees really need to be considering.

Chris: 16:21
So it sounds Callum, some people might be listening and saying, well, are we saying then that trustees should not record their meetings at all? But I don't think that is quite what we're saying, is it?

Callum: 16:34
No, completely agreed. With the correct risk mitigation procedures in place, we think it's it's reasonable for trustees to be using it for that purpose. So, for example, being aware of those risks and therefore considering turning off AI if legal advice is being given, if a sensitive member case is being discussed, making sure you switch it off for those parts. And then after the meeting, making sure that it's reviewed as quickly as possible and deleted as thoroughly as possible as well.

Chris: 17:07
We've seen some trustee boards interested in a well, AI policies generally. So use of AI by the trustee board, but specifically an AI policy on sort of governance in meetings and recording meetings.

Thomas: 17:21
Yeah. Is that the most common use that you're seeing from trustees and trustee boards kind of using for recording and governance? Anything else you're seeing?

Chris: 17:28
Yes, at the moment, some trustees are speaking with service providers about innovative sort of member experience tools and things like that. But most of the discussions that I would say I'm having are around trustees' own use for governance purposes. And their service providers' use for governance purposes, so the admin team recording minutes, for example. Yeah.

Thomas: 17:49
And are you finding that your clients are able to spot and preempt the legal risks? Or are you finding it's kind of the opposite that you're hearing about things and going, ah, actually, there's something you need to be looking out for here?

Callum: 18:02
Yeah, I I think in a lot of areas you typically see trustees knowing at least the basis of the legal risk. I think in AI, because it is such a new area, I think it it is often the opposite. So using AI to record trustee meeting minutes is probably the prime example because those are such specialist areas of law when you're talking about litigation privilege, when you're talking about DSARs. So I think a lot of the time when we go to trustee boards and explain these risks, they they they weren't necessarily anticipated. Similarly, sometimes we'll go to a trustee board and they'll say we don't use AI as trustees, therefore, you know, we we it's not really a concern of ours. And then we have to explain the member perspective as well. So if members are receiving communications from the scheme, they may be uploading that to kind of open source or public AI, which poses a cybersecurity risk to them. It may also be producing misinformation because pensions is such a complex topic and is also specific to that scheme, specific to that member. So a member could be getting faulty guidance, could be getting faulty advice from AI and then relying on it. And often we see trustee boards not always anticipating that risk.

Thomas: 19:30
I'm really interested as to where that risk kind of actually, who that risk sits within practice. So in the the last episode we did on this podcast, I was talking to the team from Quietroom. And what so Quietroom's a pensions communications consultancy. In the work and testing they've done, they find a correlation between like well-written communication and the ability of tools like ChatGPT, Copilot, Gemini, to generate accurate, useful, usable answers. Basically, you know, if the communication is easy to understand and use by a human, it's more likely that if that's then fed into AI, that a more accurate answer is going to come out the other side. Now, of course, we know there's a big spread of quality in pension communications, let's put it that way. Where does that risk sit? Does the risk kind of sit with the person that wrote the communication in the first place? Does it sit with the AI lab that made the model? Does it sit with someone else?

Callum: 20:20
Yeah, so I think legally it's not a path that's been trodden yet. We we we can't say for sure. I think from our perspective, if the source information, so what is on the scheme's website, what is in the member booklet that the AI has used, as long as that source information is correct, then we think the scheme would probably not be liable for that. And that's especially if it has the appropriate wording in place, for example, all the benefits and rights of members is dependent on the trust deed and rules, that would further help the trustees. I think the other end of that is even if legally the scheme isn't liable, you still have a member that's potentially acted in reliance on that misinformation. And then they may submit an IDRP, so internal dispute resolution procedure. The scheme then has to deal with that potentially one or two stages, and it could further go to the pensions ombudsman, the amount of expense and time that's incurred in doing that, when you could have potentially mitigated that risk by auditing your scheme information online or providing a warning to members.

Chris: 21:33
And to the question of whether there is a duty for trustees to think about how members might use AI in deciphering scheme information, there's sort of nothing in legislation that requires that, you know, naturally because AI is relatively new. And if you think about it, there's you know, there's there's nothing in pensions legislation around how members might use information that they receive from a scheme. So, on the assumption that information you're providing is accurate and clearly and concisely written, the most sort of relevant references that we've seen in the regulator's code are in the record keeping module, which simply says that trustees, governing bodies might consider it good practice to take into account developments in tech that may be available to the scheme. So that's about the scheme using AI or other tech as well. And then in the communications module, that you know, trustees should ensure comms are accurate, clear, concise, relevant, and in plain English. Query whether you can do all of those things at once, but you know, I'm sure you can. And then when deciding on the format of comms and the information to be published, trustees should consider any technology that may be available to them and available for their members. So we would suggest that trustees ought to be thinking about the fact that their members might be going off and using public AI tools to decipher information that trustees are sending about the pension scheme. And yeah, we we know Quietroom very well, actually, Thomas. We've had the pleasure of co-hosting an afternoon in September last year, a sort of a conference on AI and pensions, and also recently did a breakfast round table with them, as well as a number of other sort of smaller and more directed events as well. So yeah, good friends at Quietroom, and and we like what they say about how you can change your copy to make it easier for members to get good results if they are to use AI.

Callum: 23:38
Yeah, agreed completely. So you've from our point of view, you've got two main ways that you can mitigate that that member risk. So you've got auditing the online information, which is probably more so the Quietroom angle, and making sure everything's AI-ready for an AI world. And then you've got sending out a warning to members explaining the cyber risk of it, explaining the information risk, explaining that if the AI tool produces something that's incorrect, it's very unlikely that the AI tool is going to accept any liability for that. So we think setting out those risks to members can hopefully prompt them not to use AI in that way. And if they are going to do it, then to do it very, very carefully.

Chris: 24:27
And and and alert them to the administrators and the you know the proper sort of you know place to ask questions. That was something that Callum, we were mentioning well, over a year ago now, but we were pleased to see that that was one of the recommendations in what, as far as I'm aware, is the only real industry publication about AI and pensions, which is the PASA guidance on use of AI and pensions administration. And that little thought there was in that guidance about you know, sending warnings to your members.

Thomas: 24:58
And it just seems increasingly more common that these tools are becoming part of our daily lives, of members' daily lives. There was a really interesting kind of tidbit from OpenAI, the makers of ChatGPT, who were saying, like, you know, the most sensitive things we talk about include things like our health. And some of the most common conversations that people are having with those tools are things about medical symptoms, diagnoses, treatments, and so on. I hear pockets of chatter in the world of pensions saying, oh, members, members aren't going to be doing this stuff. But I think we're seeing from testing that they are. I'm even seeing I'm hearing this in really interesting ways. I was talking to a scheme a month or so ago who said that they ran a consultation exercise and they saw far more consultation responses than they were expecting. And when they went back and looked through those responses, they could see the hallmarks of AI-generated writing. In a sense, I think it's kind of lowered the barrier to members engaging.

Chris: 25:49
Yeah, we've heard the same anecdotally as well.

Callum: 25:52
Yeah, and I've I've heard the same anecdotally from the pensions perspective, in particular on IDRPs. So several schemes have said to me they think they've seen complaints from members that have been clearly produced by AI. And there isn't too much schemes can do about that, apart from making sure that they monitor the numbers of IDRPs that they're receiving. And if they do see that number increase, then maybe they should consider whether they need a policy in place to deal with potentially AI-produced IDRPs.

Thomas: 26:30
Yeah. Does it change how one of your clients should consider a response or a complaint if it looks like AI's had a hand in writing it?

Chris: 26:40
There was probably a trend a few years ago of IDRP responses, sorry, IDRPs, so the process by which you respond to a complaint, switching to single-stage processes for simplicity. And maybe we might see a trend where trustees go back to using a two-stage process so that the first stage process can simply weed out any AI-generated nonsense complaints. Thomas, Callum mentioned two good things there that trustees could be doing to mitigate risk. Just to add to his list a little bit, we have been speaking with lots of trustee boards about AI. And I think training on how AI works, on legal risks and mitigation, particularly in the pensions world, will help mitigate risk of challenge. And with that, also speak with your service providers. How are they using AI? What are they doing for it? And particularly your administrators, to what extent are they using AI in the way that members can, you know, can get in touch with them?

Callum: 27:45
Yeah, I think the one other step I would recommend for trustees is just putting AI on your risk register. Yeah. So risk registers, I'm conscious that they're getting longer and longer with everything that needs to be included in them nowadays, but we do think it's worth having it there as a as a prompt for whether you review it every six months, every year. After that period, you might think there has been a lot of improvements in AI. Maybe it it triggers the thought of, ah, maybe that IDRP was produced by AI. Let's have a think, let's you know, discuss with our advisors whether we need to do anything more on the topic.

Thomas: 28:25
What I'm taking from this is it's it's not really a viable option for trustees or a scheme to go, look, we're we're not going to engage with AI because you know the reality is that members, the people who our industry were set up to serve are going to be using it. Their service providers are going to be using it too. With that said, the trustee hopper is already full and spilling out over the sides. It seems so daunting now that kind of the next thing on the list might be, okay, I've got to learn about it, I've got to figure out how to use it, I've got to use the use cases, I've got to bring suppliers in, I've got to audit my supply chain to kind of figure out how it's being used in and amongst all the people who kind of do services for me and for members. That does feel incredibly daunting, right?

Chris: 29:04
You said at the top of this podcast that we weren't going to sell, and so I'm not, but that is somewhere we can help with training and helping people understand their duties. What I would say is yes, there's a lot in the hopper for a trustee. It is a big role and a big responsibility being a fiduciary. And there have always been lots for trustees to need to work through and address. And this is something else in that mix, but it's always been a challenge for trustees to balance all the different things they need to do.

Callum: 29:38
And I I guess on that note, it'll be interesting to see whether the pensions regulator provides any guidance on what trustees need to be doing, whether it becomes part of one of the existing general code modules, whether it becomes its own general code module, that'll be an interesting one to watch.

Thomas: 29:57
Let's let's flip this on its head then, because there's one way of looking at this, which is like, oh, there's another thing on my list of things to do. There's another where you go, look at all the opportunities for us to improve the way we run, how we deliver the experience for members. What do you think if your client came to you and said, we want to be like right at the frontier of what's possible, we want to embrace the technology to the fullest extent that we can and maybe push that line even further, what are the kind of things that you would be talking to them about?

Callum: 30:22
I think the most exciting opportunity is the member engagement perspective. So when you think about and talk about pensions with family members, with people that aren't involved in the industry, there is a real information gap. There is an information deficit in the general public. And when you think about how AI can help with that with member comms and making those communications more accessible, more likely to be read, more likely to be watched. It feels like it is a really exciting time for trustees and for members in the in the industry.

Chris: 31:02
Absolutely. I'd say that's great. Definitely, trustees should be thinking about AI and how they use it and how they mitigate risks of it and say, great that you want to do that. I would also add a note of caution about whether it's appropriate for trustees in their fiduciary role to be at the frontier of driving what it is possible for AI to do. And so I would say work with service providers and advisors, understand the risks, but whether trustees should be sort of at the cutting edge of where those risks might be, I don't know, different trustees will take a different view to that.

Thomas: 31:36
If you were going to go on balance and give it a percentage, like of the split of who you advise, like what percentage is conservative, what's in the middle, what's chasing the frontier, how would you assess that split?

Chris: 31:47
That now that's an interesting question. And some of our clients are doing very innovative things with sort of structuring of schemes or ways to perhaps new structures for providing pension benefits. But thinking about AI in particular, I would say that most of the clients that I speak to are keen to understand what's possible, but maybe not to and and there have been some really innovative jumps forward. So on a session that we did with Quietroom, there was an example of a member chatbot by a large pension scheme sponsored by the Ford Motor Group. And that was fantastic to hear about. And that was a giant leap forward, but I would say that most trustee clients are exploring what might be possible, but not wanting to be right at the frontier. And I'd say that's probably the right place to be.

Callum: 32:42
Yeah, yeah, I'd I'd agree with that completely. I think especially some of the smaller kind of sub 100 million pound schemes that we're speaking to, they're focused more so on the risk mitigation and the governance. Yeah, risk mitigation and governance and considering how how to get the best out of the governance kind of efficiency gains. And then when speaking to the bigger billion pound plus the LGPS funds of the world, you're seeing them try and be a bit more proactive. What what can we do? But also definitely focusing on those risk mitigation measures as well.

Thomas: 33:20
I've got one eye on time, so I'm gonna bring us to a close with some quick fire questions. So here's my first question for you. Favourite Gen AI tool that you use at work?

Chris: 33:31
Of course it's Harvey, which Burges Salmon have partnered with.

Thomas: 33:35
What's Harvey named after?

Callum: 33:38 I'm not sure, but one of the faces of it is Harvey from Suits, which I think Harvey from Suits.

Thomas: 33:42 That's what I was gonna take a guess, yeah. 

Callum: 33:45 I think I don't know whether it was completely intentional, but he's he's definitely definitely the face of it. And I think one of the other legal AI tools has now got Jude Law as their face of the company. So there's there's a real influencer war going on at the moment in the legal AI space. 

Thomas: 34:00 That's brilliant. Favourite tool for personal use? 

Chris: 34:10 So I use ChatGPT. 

Callum: 34:13 I would I would say mine is Copilot just because I'm used to using it at work. So the prompting and the the language I get from it, I feel more used to it from a personal use point of view.

Thomas: 34:23
Now I want you to imagine, let's say in like six months' time, we have a massive leap forwards in AI capability. And I was able to give you an AI agent that could do any part of your job really well. Let's say it could do it as well as you could. What's the stuff that you would still choose to keep to yourself because you as the human really love it and want to keep doing it?

Chris: 34:42
Oh, that's a good question. Callum, do you want to do you want to go first? I've got an answer. Do you want to go first?

Callum: 34:46
Yeah, happy to. I'd say it's probably the the training. So especially the the the AI training that I do, even if rather fittingly an AI tool could produce the AI training, I would definitely still want to continue doing that. That's the part of my job where I have the most contact with with clients and and with other people in the industry and have the conversations both you know before after the meeting and also deal with the questions and you know, host some of the debate about this new technology in the pension industry. Yeah, I find that really interesting. And yeah, I definitely wouldn't want that to be kind of replaced.

Chris: 35:28
Look, this job is all about people. It's all about relationships with people and our clients' relationships with the people that they deal with in their business. And AI can't replace that. That's the bit of the job I like. And so if AI came about trying to do that job as good as me, then I would still want to do that bit.

Thomas: 35:47
And my last question for you then. So we're all still on an AI journey together, we're all at different points on that journey, but there are some people who are just still getting started, getting their feet under the table today. If I could give you the world's worst time machine that you could only use for one thing, and that's to travel back in time to the launch of ChatGPT. Right. Do a bit of work with your team and kind of go, okay, here's how we're going to adopt it now, get to learn to use the technology based on all the experience we've had so far, kind of going from day one. What would you go back and change? What would you do differently?

Chris: 36:18
I think it's just a case of getting stuck in and accepting that there's a bit of investment time needed to learn new skills.

Callum: 36:27
Yeah, I would I would agree completely on those same lines, just talking to people about it, whether they use AI more than you or less than you, understanding how they're using it, I think can really help you and and also help them in their kind of AI journey.

Thomas: 36:43
Thank you so much for joining. This has been a really good conversation. I've been so interested. I could, if I could lock the door and keep you in here for another couple of hours, I would. But I know I've got to I've got to let you both escape. But look, thank you for your thank you for your time, your energy, your thoughts. You've both been brilliant guests. I really appreciate it.

Callum: 36:58
Thank you very much. It's been a pleasure. Thank you very much.