Graeme Cox & Sarah Clarke: AI Governance as the Strategic Enabler
Summary In this episode of the AI Pathfinder for Private Equity podcast, host Steve Budd discusses the critical role of AI governance with experts Graeme Cox and Sarah Clarke from Attercop. They explore the misconceptions surrounding AI governance, emphasising its importance as a value enabler rather than a hindrance to innovation. The conversation covers various themes, including the risks associated with AI liability, the parallels between AI governance and cybersecurity, and the necessity...
Summary
In this episode of the AI Pathfinder for Private Equity podcast, host Steve Budd discusses the critical role of AI governance with experts Graeme Cox and Sarah Clarke from Attercop. They explore the misconceptions surrounding AI governance, emphasising its importance as a value enabler rather than a hindrance to innovation. The conversation covers various themes, including the risks associated with AI liability, the parallels between AI governance and cybersecurity, and the necessity for private equity firms to understand and manage AI risks within their portfolio companies. The guests highlight the importance of visibility, education, and a structured approach to governance to leverage AI effectively while mitigating risks.
Takeaways
- AI governance is often misunderstood as a cost rather than a value enabler
- AI governance is becoming essential, akin to cybersecurity hygiene
- AI governance frameworks are still developing and need to be prioritised
- Understanding AI risks requires a focus on use cases rather than tools
- Effective governance can lead to competitive advantages
Chapters
00:00 Introduction to AI Pathfinder and Governance
03:34 Understanding AI Liability and Governance Risks
08:39 AI Governance as Cybersecurity Hygiene
11:09 The Evolution of AI Governance
14:41 Visibility and Management of AI Risks
21:33 The Role of Private Equity in AI Governance
23:50 Linking AI Use Cases to ROI and Risk
26:40 AI Governance as a Competitive Differentiator
31:20 First Steps in AI Governance Implementation
Links
Attercop website
Graeme Cox on LinkedIn and Sarah Clarke on LinkedIn
AI Pathfinder Private Equity AI Strategy Meet-ups:
https://www.aipathfinder.co/private-equity-ai-events
Have a question or would like to suggest improvements, please contact steve@aipathfinder.co
Hello and welcome to the AI Pathfinder for Private Equity podcast. AI Pathfinder helps private equity firms make sense of AI and make it work. It's an expert network built on insight, experience, and connection. We bring the right people and ideas together to turn AI from something you're curious about into something that delivers real results. If you'd like to attend one of my regular AI strategy briefings in London, Manchester, and Sioux, New York, please check out the show notes for details. Now today we're taking a deeper look at one of the most important and often misunderstood enablers of AI success, governance. Governance is often seen as the brake pedal in innovation, but my guests today argue that done well is actually a value enabler rather than a cost. To unpack this, I'm joined by Graeme Cox, founder of Atacop, and Sarah Clark, Atacop's governance expert. We're going to talk through four big themes. Why you can't just disclaim AI liability, how AI governance is becoming the new cybersecurity hygiene factor, why visibility is power when it comes to managing AI concentration risk, and finally the commercial upside. How governance can become a genuine competitive differentiator. Graham Sarah, welcome to the podcast.
SPEAKER_00Hello.
SPEAKER_01Thank you. Lovely to be here. Well, we've got a fantastic topic here to get our teeth into. But first of all, I just wondered whether you could give us a brief introduction to yourselves and Atacop. Thanks, Steve. Yes.
SPEAKER_02I'm Graeme Cox. I'm the founder and CEO of Atacop. I created the company on the back of my career in AI. I have a degree in artificial intelligence from Edinburgh University from the early 1990s, and I have been a practitioner in some form or another across my career, alternately working for enterprise businesses in global consultancies and building my own companies. Atacot came out of several years of work that I did in private equity, helping initially primarily on the buy side with AI and Data Tech DD, and then increasingly over time starting to get involved in value creation work, AI strategy, and practical roadmap that organizations could execute to improve their data and AI play and drive new value for the business. Just under two years ago now, I realized that there was a lot more to go at than I could possibly do myself, and so decided to found another company and hence Atacop was formed.
SPEAKER_01Fantastic. Thank you. Sarah.
SPEAKER_00Yeah, I'm Sarah Clark. I've uh known Graham for about 20 years. We worked together before in a cybersecurity-focused firm back in the day. But no, I I've come through um sort of IT, then cybersecurity, then data protection, and um for the last few years focusing on sort of AI and novel tech governance. So I'm absolutely delighted with Whoop to work with Atacop. I've worked in regulated sectors, I've worked in a range of other sectors as well, and you know, very, very much always been focused on right-sizing governance for whatever kind of organisation we're working with. And uh yeah, it's been quite a journey with how that's developed around AI.
SPEAKER_01Yeah, fantastic. And and you joined us at the AI strategy day in September and and talked about this topic of of um you know uh AI governance being the new source of cyber security. Let's start with the idea that AI liability might now be, let's start with the negative, might be now be one of the biggest uncapped financial risks companies face, and that you simply can't disclaim your way out of this. I wonder if you could just unpack this a little bit for us and in practical terms what that means for P back businesses.
SPEAKER_00I mean, I'll I'll kick off with a little bit of a backstory to that. So there's been a couple of very famous cases. One of them was Air Canada and one of them was Chevrolet, where chatbots had said things about people getting refunds and about honouring deals that there was court cases that tried to argue that because it was AI, the vendor had some accountability for that, that the company could disclaim it was not the outcome of those cases. There was liability held for the company who'd put it on the website. Um, that pattern has just continued to roll. And one of the challenges is that we're seeing an insurance market, which is uh often a way to transfer that liability risk that's still maturing. There are a lot of silent coverage in cybersecurity policies and that kind of thing. But more generally, it's just really a call to pay attention to do you understand that supply chain that you're buying into when you are inserting AI into your business?
SPEAKER_02It's it it's critical to critical to understand risk, obviously, when a private equity firm is in investing or thinking about investing in a business. And being able to see how that risk is managed is the critical piece. It's the whole point of our discussion today to talk about governance. It's not about not investing in companies that have risk. The companies that are that are working in potentially risky propositions are likely to be the ones that uh offer the greatest potential upside. It's about understanding whether that risk is governed effectively, whether it is recognized, whether it is talked about and managed at the C-suite level and across the organization. And that has both a that is both a management reporting and just basic governance and governance requirement, but it also delivers down into technical controls, ways of working, observability platforms for AI solutions and other components that show a mature approach to recognizing and managing risk in as it as it might exist across any artificial intelligence or machine learning solution. I think the other thing I'd like to just put in as the as a starting point to this discussion when we're talking about these genuine potential financial and legal risks that might exist in organizations, is that it is, it should all be always be seen by use case rather than by rather than by tool. It's not around, for example, the particular use of a particular foundational model, and I should we use Grok versus Google Gemini or or whatever. It is around what you are intending to do with them and specifically the information that you are intending to put through them that matters. So managing managing AI risk and implementing governance across use cases is is absolutely required to get this right.
SPEAKER_01And we're in a uh situation where the models themselves, the large models, are um almost void of of uh governance. We can't push the liability to to them, that that's for sure. So uh and and nor can we uh necessarily rely on sort of broad disclaimers. Why can't you know companies r rely on that? Are we in a new, a very news uh world right now compared to what what we uh what we've what we've been used to?
SPEAKER_00I I don't think so, really. I mean, we've been in this situation with cloud vendors, um, we've been in this situation with SaaS vendors. I think the journey we're going on is understanding where the edges of those things are, and just to Graham's point, exactly that we need to understand that intersection of the tool you're trying to use, the vendor behind it, the supply chain behind that, and then apply it to your last mile, what you're trying to do with it. Because for one use case, something that you perhaps can't get the visibility and you can't get the vendor governance at the level you'd like it. Most businesses can't sort of say to Microsoft, let's just have a wander around your data center and have a look at how good your controls are. We know that doesn't happen, and it never has. So vendors are trying to govern at a level for general use cases, and that's never going to translate into your last mile. So your specific use is really where your responsibility starts. To the extent you can get information for about models and other solutions that satisfies you, it may be exactly the right tool for the job, that might be a manageable risk, but you can't govern what you can't see. So you do need to do that, go on that journey to understand how much assurance can they provide, how much do their baked-in governance tools in things like Microsoft Assure provide you, and what do you then have to do to make sure that it's providing what you need for the right outcome?
SPEAKER_01And you've mentioned, I mentioned at the top the the session you ran in September about AI governance as the new cybersecurity. What other patterns have you have you seen there between the two?
SPEAKER_00I think from my point of view, I've seen patterns coming through the development of the governance frameworks that are being mentioned by most people. So you can't get away without mentioning the EU AI Act. We're not in the EU anymore from a regulatory point of view. But if you are servicing uh EU companies in any way, you've still got to pay attention to that. You've got the ISO suite of standards, ISO 4201. Now that is almost a direct parallel, that AI management system standard to the ISO 27,001 standard for cybersecurity that a lot of people have focused on with the cyber hygiene, along with things like cyberessentials. The big theme I've spotted is that those things are coalescing. There's about 70 to 80% crossover in early required steps around those. But more generally, in terms of threat and risk, we are building up our knowledge base around that.
SPEAKER_02Yeah, if I can, if I can pick up on that, I think the parallel between the development of cybersecurity as an industry, which it actually is these days, and AI governance is really interesting. So as Sarah mentioned, we both worked together in a cybersecurity company uh around the turn of the century when cybersecurity was really just starting to find its feet. We we entered the world of cyber at a place where there was no such thing as a chief security officer. There was no such thing as security engineers being embedded in organizations. Basic security systems were employed. It was seen as a technical control landscape first and foremost. And we have, of course, developed over the last 25 years into a really sophisticated multi-layered approach and a very strongly implemented philosophy of managing security built around the 2701 standard, which applies policy, process, and technical controls under those in order to provide a visible, observable, measurable, and manageable approach to governing risk as it appears on computer systems and computer networks. We're starting that same process with AI in that we are just seeing we're seeing it as a set of generally technical controls that you would implement to manage AI use case risk at the moment. There are no or very few AI governance professionals who exist out there in the market and very few people who embedded in organizations that genuinely understand and take ownership of AI risk. But as AI use cases, as AI systems and features of production systems start to grow rapidly within organizations, customer-facing systems particularly, the need to understand the new multi-layered risks and what the right policies, processes, and controls for AI that are necessary to manage that risk, we're starting to see that growth in sophistication. And I fully expect that 15 years from now, probably a lot less given the pace of the AI industry, let's say five to ten years from now, we will see AI governance professionals being embedded in organizations. Whether it becomes the remit and role of the cybersecurity expert or it is an addendum expertise, I think the latter is to be seen. But I think that's there's a if you've been around long enough to see that growth and maturity developing in cybersecurity, we can absolutely see the same requirement in AI. One more thing to say on that, if I may, which is that there are obviously differences between computer and network security and AI security and governance. The risks are different. That is multi-level, and Sarah, I'm sure, could talk about that for hours. But but uh but what one thing I would like to just express again to your audience around the difference in AI from traditional software and computer systems is that AI is a stochastic medium. AI is not a deterministic medium. In software, if something doesn't work correctly, it's a bug. And you put effort into fixing the bug or finding a workaround for the bug. In AI, when something hallucinates to use the commonly used term or fails to come up with an answer, it's a feature. It's not a bug. Even though, you know, the AI is now a talked-about subject in every organization. I still spend time educating people on that fundamental difference between AI and traditional software and what that really means. Like, how do you build a business around something that isn't always right? It sounds to the average CEO when you express it that way, that it's horrific and you should immediately walk away from AI because the risks are absolutely massive. But actually, we have always built businesses around systems that are not always right. They're called people. And so we're very good at managing risk in human processes. This is this is the entire structure of business as we know it today. And this is the sophistication that needs to be built into understanding and managing AI systems. And that is fundamentally why AI governance is an enabler and not a disabler to technology innovation.
SPEAKER_01Yeah, yeah, fascinating. So, you know, just summarising that is that we're potentially better placed than we think we are because we've got we've got frameworks that we could call upon around the cybersecurity piece and potentially expertise there. But at the same time, it's it's a very different beast and it can be unexpected. You mentioned there about the CEO and being quite concerned, um, and it obviously depends on the type of industry as well, in terms of the risk appetite. But another theme is around visibility that that companies can't govern what they can't see, especially when it the the usage of AI is scattered around the teams across many tools. So if we look at mid-market P backed companies today, how visible is AI in reality and where do you see the the biggest blind spots?
SPEAKER_02That's a that's a fantastic question, Steve. So I I don't see AI governance on the um on the management information that most um PEs are collecting as portfolio management information today. It does not exist on the scorecard, it does not exist in the mindset and the thinking of the PEs who are overseeing, governing, and managing their portfolio. And that is one of the key things that we would like to effect change in going forwards. Again, you know, seeing um uh governance as something that is effectively an it's it is an enabler to business because it provides the confidence to make informed decisions about what use cases to move forward with and what not to. But it also does manage inherent, uncapped risk for private equity companies with investments in these companies that they understand that the that the basics are being covered. It would be pretty unthinkable for any even halfway sophisticated private equity company today to not look for cybersecurity information in its management information, to not look for AI governance information at the same level is exactly the same. But because that understanding of AI governance is still nascent and developing, it has historically not been there, and we're still not seeing that in most organizations. So the practical from a private equity perspective, from a portco management perspective, asking for a governance framework and evidence of that working in operation in its portcos is clearly a starting point for private equity to under to get a grasp of whether that risk is being managed in organizations. When you actually come down into the portco and talk about, well, what does that mean? And you touched on a couple of points there, Steve. So there is both the small scale and the large scale piece. So the small scale piece is around shadow AI and the proliferation of unregulated and unmanagement tools, unmanaged tools being used by the employees of any portfolio organization. So an AI acceptable use case policy goes in, it's uh it's boilerplate text created on ChatGPT, it gets published somewhere, nobody ever reads it, but at least they can say that there's a there's a policy in place that it I'm afraid is not an unusual situation. It's been mandated that either the, I don't know, ChatGPT Enterprise license or possibly the Microsoft Copilot license, that's the primary AI tool. But there's no real carrot and stick there. People don't get given good enough literacy training to understand how to get the best out of the tool that they've been told to use. And the the stick isn't there to say, well, what is the implication of using that tool incorrectly? And worst of all, the the literacy training, again, coming back to my point about use cases, is understanding when it is appropriate to put certain kinds of data into certain tools. Like there is probably no risk in asking your personal ChatGPT account to draft an email to a new prospect as part of the origination process to try and generate a conversation. You know, that's that's basic marketing. To put your confidential information memorandum into your personal ChatGPT account, it's maybe a little bit more of a risk, and I could go up from there. So at that small level, carrot stick and education matters in managing and controlling shadow AI. At the larger scale, observability of AI features and AI products that are in production that are actually delivering some kind of core customer service, revenue generation, et cetera, absolutely matters. And by observability, I don't just mean the traditional stack of technology, uh performance, cost, latency, et cetera. I mean the answers that your natural language systems are giving to your customers, how appropriate they are, whether there is bias in them, whether they are accurate, whether conversations have been completed correctly, and whether the customers are having a positive experience in interacting with that system. That is a governance layer that is required for AI because of it's a natural language interface, which is generally not seen in the cybersecurity or or system monitoring world, and is critical for uh your governance platform.
SPEAKER_00I've worked trying to get these visibility pieces right for huge companies with huge stores of systems and vendors and that kind of thing. My focus has always been what's the best information you can gather quickly that doesn't just rely on experts together. So we have lots of expert, big, large platforms out there in the market, things like Credo and things like Fairly AI and things like holistic. But mid-market companies, they're not in the place generally to sort of take advantage of something that keeps track of all your AI and how you're using it. So it comes back to good old-fashioned discovery and inventories. And it's very unsexy, but it's so essential. But where it gets sexy is if you collect the right proxy indicators and direct indicators of how you're using the AI, what data's going into it, how big is it going to scale? Is it shop window? Is it client-facing? Is it just an embedded element of another tool? If you build up that picture with an initial heavy lift and an easy way to update it, the power of that aggregate intelligence, if you keep it, is enormous. So let's think in terms of portcos, a whole range of portcos. You can see how many of the AI initiatives are leaning on open AI. How many of those initiatives might come within the scope of GDPR using personal data? How many might be exposed if there's another Cloudflare outage that just took out OpenAI and anthropic models as well as a whole bunch of the rest of the internet? But there's also an upside to that. Even if you might have a use case that's constrained by current capability or investment appetite, even if it's not a now thing, because the models are all leapfrogging each other hand over fist, it's a store of strategic opportunity intelligence as well as if it's captured and retained. And it's really just taking that step change from doing a point-in-time discovery and gap analysis to actually keeping some of that information and plugging it into the rest of your governance. You've always got hard lines like do we need to do a data protection impact assessment, or do we need to work out if it's high risk or even potentially prohibited under the EU Act? I mean, again, that's not going to apply a lot of the time to these companies, but having an understanding of that exposure is suddenly power because you can have a confident conversation about your spread of regulatory and other risk exposure and get a lot of intelligence for your opportunity management as well. So groundwork is always where it tends to come back to with governance, and there are smarter ways to do it in a less painful way, and that's tends to be where I focus.
SPEAKER_01Is this something you you feel a P firm should mandate across their their portfolio? And another question on top of that is that surely there's then a um an opportunity to use the fact that they have um that they are part of, say, 30 firms that are um using AI and and they have a uh spreading their risk across the small and large scale, is that they should be helping each other. So they should be learning fr from that.
SPEAKER_00The the the economies of scale you can gather with doing this kind of thing really depends on getting normalized data without killing people across across the piece. We all know from interacting with data our our entire careers that um a small set of quality data that's sustained and maintained and useful is better than trying to ask everyone all questions. And this is where governance tends to fall down. If you get somebody who's not experienced come in, they might work from the top to the bottom of an AI governance policy and not actually get that first foundational piece of intelligence that you can maintain and then build out from. So short answer is yes, I think everybody should be collecting a baseline of data, but the big question is what baseline? And that's where the research done across standards comes in.
SPEAKER_02Yeah, I mean just at a very simplistic level, you're just keeping this purely from the private equity perspective. So your question is was should all PE firms be um asking for data on how AI risk is governed in their portfolio organizations? Um well, let's just take out the AI bit. Should PE companies be managing risk in their portfolio companies? There we go.
SPEAKER_00Should they should they know what's out there, what's being used, which vendors are attached to it?
SPEAKER_01Yeah, yeah, absolutely, absolutely. It's almost like a bit of a therapy session this. Let's let's um let's focus on the upside because this is a this is an opportunity as well, isn't it? And um, you know, good governance can be and should be an enabler, not a blocker. How do you link individual AI use cases to both ROI and risks so that governance becomes a way to prioritize the right things rather than just saying no?
SPEAKER_02What we seek to do in helping organizations to set AI strategy is put risk, um, uh risk and reward together to help a C-suite make the right decision. We're looking at developing return on investment cases for any AI use case, whether that is delivered in revenue generation, cost savings, efficiency improvements, whatever is meaningful to the organization itself, and matching that against the risk and the risk mitigation strategies that can be applied to uh to that use case in order that you can make a clear business decision about whether something is right to go for. Some of as I said at the start uh the start of this session, some of the most valuable opportunities carry significant risk. So it's you know, it's absolutely not the case that we say, oh, we just don't do it. It's about how would you manage that risk and carry it forwards. And with most, you know, it's it's like any other risk in business, recognizing it, being able to s to um um to minimize the uh the chance of it occurring for a reasonable degree of spend, monitoring the situation so that you can get ahead of it should something start to uh start to happen. You know, all of the or uh and including just insuring against it as the as the absolute base case from a financial perspective, these are all really solid strategies for uh for handling risk in in business. I think, you know, we when we're talking about governance as a positive, as a as a way to allow you to move forwards, um, I think I feel really strongly about this from the perspective of the uncertainty that AI produces inside organizations. We live in uncertain times. AI is still proving itself, and there are many use cases which people say, oh now oh, AI can sort that for us, where it patently can't. So, you know, we some in some areas we have over overinflated expectations of what AI is going to do for us. In other areas, we have this fear and uncertainty and doubt that around the stochastic nature of AI and oh my god, are we gonna put it out there and then it's gonna hallucinate something and our reputation is gonna be in tatters, and therefore we should just shouldn't do it at all. And walking that middle ground means being confident about the ground that you're walking on. And it's only by having a solid, pragmatic, practical framework to recognize risk and manage that risk that you can have that confidence. So for us, that governance piece is a necessary plank in making good, solid business decisions that you could be confident upon that is going to deliver value for you in the future. All business involves risk. It is simply a matter of understanding that risk, being able to talk about it clearly and cleanly, and doing something about it. And that is why fundamentally governance is not just an oh, if we have to, it's not just an oh, we do business in Europe and therefore there's this legal requirement we have to deal with. It's it's not lip service. It is the way to make the right decisions and to move forward as fast as possible.
SPEAKER_01Are you able to give us an example of what it looks like when governance becomes a genuine commercial differentiator?
SPEAKER_02Yeah, absolutely. I mean, so this is this is a project that Sarah's deeply involved in um at the moment, and that we spoke about as a case study at the um Pathfinder Strategy Day, Steve, where we're dealing with an organization that works into a heavily regulated industry and it is a producer of content, and it is seeing uh strong pressures in two heavily competing directions from its customers. On the one hand, they're saying to this organization, we pay you for your human expertise. So, you know, I want this output to be human driven. Don't just give me AI copy, because otherwise I could go anywhere for that. I need I need your best people to be doing this work. And then on the other hand, the contracts team are saying, right, well, you're using AI, so you're vastly more efficient than you were, so let's take 30% off your pricing, shall we? The organization could just get into a position where they go in a race to the bottom, trying to automate everything that they can in order to deal with this downward price pressure and just still win the work on whatever terms are necessary. They could take the complete opposite view and say, right, well, we just won't use AI. We will be that human-led organization and will live and die by that sword. And ultimately they will die by that sword because the efficiency that can be driven in particularly any kind of technical writing is clearly going to drive efficiencies in other businesses that they can't achieve. The way to be able to stand up to their very large clients and look them confidently in the eye and plot a middle ground there is to be able to talk about the genuine risks and the benefits of the technologies that they're using, to have a strong governance framework in place that shows that they recognize the risks, they manage and minimize those risks, so they can give the best possible service at a competitive price whilst retaining the value of their humans. So augmenting their human staff with AI and not replacing them requires that risk management and governance to show that they're they're walking that right line, convincing their customers they're still getting the best out of the people, but they're getting the best out of the technology as well.
SPEAKER_00One of the big things that we're finding is that all companies of no matter what size are working out what requirements they want to pass down to suppliers in terms of AI governance. Um we're we're seeing hand-over-fist new contract and master service agreement clauses coming coming. In a lot of cases, I can I can see because of the journey, have been written by legal teams that are trying to cover every front. But just as a live example again, you've got different usage scenarios for AI, and we all get very frustrated by people not sort of splitting this up into are we talking about generative AI or are we talking about machine learning or are we talking about a specific SaaS solution that might have a bit of AI in it? They have got a clarification going back to some big companies soon saying, what exactly are you expecting us to do? Are you expecting us to do governance on the big foundation model providers? We suspect you're not. Are you trying to rule out us just using productivity enhancement in our processes, which we would expect you to view in the same way as using the internet, using Microsoft Office, using a cloud service? Can we start that conversation? And I think that conversation is still very early. And the way to be on the front foot and potentially a strategic leader in that area is to A, keep your domain expertise central to whatever it is you're discussing, and be very clear on what the edges of the tool you're using and the edges of provided governance to that point are. I think there's uncertainty is also an opportunity, and the first people to see a clearer path through are going to have a distinct strategic advantage.
SPEAKER_01Yeah, and I think the the the uh example you gave there, just having that position uh around how they're going to uh utilise the technology becomes a differentiator as well, doesn't it? Um that they've they've made a definite decision and this is what it is, and that flows through to their their customers and across across the business. We are coming near to time. Is there anything you want to to add before we before we go? I know there's a lot here. Perhaps it's um you know it's first steps, next steps for for the listeners here.
SPEAKER_02Yes, well, um I think uh the the clear message that I've tried to get across today is that um firstly AI governance is not something just to pay lip service to. It is absolutely critical that private equity companies recognize that there is risk in these use cases in their portfolio companies, and that needs to be managed in some form. The second is that implementing a framework for AI governance can be a very light touch for a lot of portcos. It is just about doing the right things. I spoke about that back office piece and shadow AI, the carrot stick and the education to go with it. These are things that can be done entirely internally to an organization. It doesn't need consultants to come in and help, although obviously we would be delighted to help if anybody wanted us to. But they are things that need to happen within organizations and it and it's crucial in order to get the best out of co-pilot rollouts, for example, which frequently uh I hear about being rolled out across organizations, costing quite a lot of money and license fees and showing very uncertain results because it is just a tool. It's not well managed as a delivery. So get those basics in place, is what I'd say. Talk about AI risk and AI governance at a portfolio management level. Talk about it, about it with the C-suites in your specific port co-organisations and understand whether there is a meaningful framework in place and consider how to deliver consistency across the portfolio as a way to move forward.
SPEAKER_00Yeah, and I just back up pretty much everything Graham just said. But the first thing is to try and create a common level of knowledge and have conversations with your people, the amount of first requests that are, but I'm not sure how I could use it. Um, so it's capturing those success stories, putting some edges around it. The boundaries are actually accelerating uptake because people are more comfortable that they have permission to use it and they know where the edges of the allowed are currently. We want to avoid a situation that we had with internet usage where we sort of hurry up and use it and then blocking it and stopping it with nothing really in between. And all of that comes down to risk appetite and exposure and opportunity. So the quicker you can get to that visibility of what you're trying to achieve with what tools, the better place you're going to be at.
SPEAKER_01Graham, Sarah, thank you both. That was an excellent conversation.
SPEAKER_00I've really enjoyed it. Thank you, Steve.
SPEAKER_01Yeah, thank you, Steve. Most enjoyable. Yeah, you've really helped reframe AI governance and um definitely not a drag on innovation. So, well, thank you for listening. If you'd like to learn more about Atacop, visit Atacop.com or connect with today's guests on LinkedIn. All the links will be in the show notes. That's it for this episode of AI Pathfinder for Private Equity. Thank you for listening and see you next time.