Name: State of AI & Automation Report: Implementation Strategies from EMEA
Uploaded: 2025-10-02T10:31:04.992Z
Duration: 56 min 48 s
Description: State of AI & Automation Report: Implementation Strategies from EMEA

Transcript for "State of AI & Automation Report: Implementation Strategies from EMEA": Alright. Hello, everyone, and welcome to today's webinar. I'll give everyone a few moments to find their seats, get settled in, and then we'll kick this off shortly. Awesome. Why don't we get started? So welcome to our discussion on AI and automation in threat intelligence. We've just released a comprehensive study, on this topic, surveying over 520 cybersecurity executives and practitioners globally, including a 130 plus from Europe and The Middle East. So what we found reveals both significant opportunities and persistent challenges that organizations face regardless of their current maturity level. So we've got some great panelists, which we'll get to momentarily. But a couple of, housekeeping notes before we get started. So one, the recording will be sent out to the email you registered with. That'll probably be in the next twenty four hours or so. Two, we've got some great experts from our product team to sales engineering to customer support. So if you have any questions around how people are implementing AI and automation into their threat intelligence programs, you have some great resources to ask today. So please don't hesitate to use that q and a functionality, and get your question asked. And if we don't get to your questions in time, we'll try and do our best to to follow-up with you, and make sure we get your questions answered. With that, I'll move to introing our panel. And, Robin, I'll start with you. Okay. Thanks, Sam. So hi, folks. My name is Robin Brinkworth. I am senior product manager here at Recorded Future for AI and Search. And, yeah, looking forward to discussing this today. Perfect. And Chesca? Thanks, Sam. Hi, everyone. My name is Chesca. I'm a customer success manager at Recorded Future in EMEA, and I focus on working predominantly with public sector and critical national infrastructure customers. Awesome. Over to you, Lance. Thanks, Sam. Hi, everyone. I am Lance Moraitis-Jones. I'm part of the Recorded Future presales team. So I'm a senior presales engineer of Recorded Future, and I look after the implementation of our services with, our clients. So pleased to meet you all. Awesome. So let's start with talking about real world implementation stories. And they bring this to life with specific examples. You know, we know some organizations that are achieving dramatic results, say, 70% workflow automation, they're seeing 50% faster investigations. But what does this actually look like? So I'll start it off with Chesca. Can you walk us through maybe someone the customer maybe you worked with or worked in the past that is seeing success with AI? Definitely. I think something that I've noticed from my customer base is that one of the biggest success story or strengths with AI isn't anything flashy or headline worthy. It's more how well AI can provide them with situational awareness. So customers, for example, use our AI chat for various different reasons. That could be, you know, from summarizing the latest events or developments in the Ukraine conflict to, I'd like an AI report on the cyber threat landscape posing significant risk to the transport industry. It could be a whole medley of different use cases. But one of my analysts in particular or one of sorry. My customers in particular or the analysts on the team have set up to receive a weekly AI report on threats to his industry. And that comes in to his desk or to his email first thing on a Monday morning, and this is a report where he has vetted or tailored the sources exactly so we know he's receiving content from high fidelity sourcing. And he said that having that report as the first thing he does at the beginning of the week gives him that sort of holistic insight that he needs to direct the focus for the week. And I think, again, that's the best example where you see the symbiosis of sort of AI and human decision making, which is where you get that augmented intelligence value out of AI. Awesome. And I'll I'll toss it over to Lance. You know, when someone achieves increased workflow automation, which workflows do you tend to see get automated and which stay manual, and what do you see as kind of staying manual? Yeah. No. It's a it's a good question. I think, some of the things that we are seeing is that certain, let's call them mundane processes that security teams will rely on, are becoming more automated through the use of AI. So AI has become a big, force multiplier, let's say, in that space, because it's giving us the power to, or giving us our clients and, the industry in general, the power, to be able to streamline some of those processes. So streamlining is very important. So when you think of things like alert management, you think things, like, prioritization of events, for instance, something about recorded feature also to build into their services. That's a big part of the streamlining process and and, it has, made a big difference in that. In terms of, your second part of the question, which is what sort of state manual, I think still the decision making process is very much manual. So, I think whilst we still, as human beings, let's say, rely on AI to do a lot of the, like I said, the mundane tasks that we do, I think we still, we're not necessarily using AI to, say, what do we do next? Okay. We've seen this. What's the the consequential action of that? That's still something that is relied in, we relied on analysts and, and analyst teams to, at least feed that information back to decision makers within the company and decision makers make those decisions. So so that that is still a a manual process and I think for the foreseeable future at the moment, it looks like it can be manual process. Great. I'll ask Robin, you know, what separates organizations that achieve actual transformational results from those that you see kinda just get stuck with those basic efficiency gains? Yeah. It's it's it's a fun one, that one. The, the biggest the biggest difference here there is talent, I think, organizational alignment. So we see AI transformations happening typically in two different ways, and we've seen this internally at Recorded Future, and we see it across a whole bunch of different organizations we work with. One is there is the top down approach where someone's saying, hey. There is this amazing new technology. We should be using it. Like, folks go out and use it. And the the inverse of that, right, is the is the ground up approach. And you can have talent in different places of the organization, but you need to have that talent sustained through the organization of the folks who are aligned saying, like, yes. We should do this. How do we enable this? And it means getting the ground level folks excited. You know, I will say intelligence practitioners. I used to be one of them. There is varying responses to it to technology there at any given time. We all know and love our specific tools that we know and love. We love working with them. And so that's that's one side of the story, but then there's also the the kind of, like, governance enablement that someone from a more senior position might be able to do as well. And it's really making sure that, okay. Hey. If you are a CISO, that your director is on the same page as you, that actually you have a a couple of folks on the ground level who know how to work with these tools and can really push those systems independently. The most advanced stops as we have of AI and automation technology, whether it's AI or automation, but the most advanced ones are typically very small teams. I'm talking like, you know, one or two people who are extraordinarily capable and are empowered to go out and deliver value for their customers for sorry. For the for the customers, like, within the work broader organization. Yeah. It's interesting because if you look at what the report I've covered, it did find that kind of the smaller teams are actually seeing more value out of AI than some of the larger teams that we surveyed. I do. Yeah. I mean, the the some of, like, some of the clients I'm working with, like, the the the one per one, two person teams where they just have, like, hey. Like, I can use Claude. I can use Chargebee. They're building our own agentic systems. Like, they're they're pushing the limit of what is the the technology is capable of in order to basically automate their own workflows. But there's, like, the classic kind of, like, lazy geniuses where they're like, I don't wanna have to do this anymore. Like, get someone else into it. But I think what's critical is they're also empowered by their leadership to do that. And they communicate to their leadership well and effectively, hey. These are the risks implicit here. And I think it you know, it's it's different organizational models that allow for different levels of of technology risk. You know, they're they're in organizations where there is, like, a forgiveness of saying, like, hey. This is not gonna work every time. Like, this might break. Like, I'm not that's okay. Yeah. Yeah. That's a great point. So moving into the current state of AI adoption. We can start with talking about where organizations actually stand today. And the data shape shows, you know, widespread interest, but varied implementation approaches. But I think it is interesting to discuss that, you know, 93% of those surveyed consider AI important for their future strategy, and 86% trust AI generated outputs today and believe that they might see, you know, a 25% plus efficiency gains in the future. So I'll start again with Chesca, but what are your customer are your or are your customers showing any reluctance at the moment to adopt AI in the workplace? It's a great question, and I think I'd love to be able to have a concrete answer to give you, but I think it really depends on the customer in particular. So I've seen quite a range of attitudes. So some customers are much more cautious and skeptical, but others are much more eager to implement AI in its current stage or capabilities right now. But I think what is a common experience or maybe a more negative experience that users have had, which fuels that reluctance, is AI being too eager to please. So, you know, even these the cutting edge AI models, the ones that are around today, they're still prone to hallucinations. We can't that's not something that's going away right now. And so I think when a model is or an AI tool is giving you or or is telling the customer what they want to hear, that's when it's not augmenting intelligence. It's actually working as more of an obstacle. So I think when that is the case, that's put some of my customers off. But, actually, interestingly enough, speaking to customers about this, while you would think that puts all of them off a little bit, and it does for quite a few, one customer actually came back to me and said, I like that AI gets it wrong sometimes. And I thought that's an interesting approach. And then they said because it keeps analysts on their toes, it means analysts have to be have that situational awareness first in order to be able to see where AI is being useful and where it's maybe not so much. And so they actually said that if AI was perfect to getting things right all the time, the analyst would never learn anything. And I thought that was a really interesting approach because it is true. No one's perfect. Humans are, you know, suck susceptible to error as a machines, and it's what we learn from that that makes us better, which sounds quite cheesy, but I do think it's true here. That was a great point. And over to Lance, you know, what are you seeing in terms of the gap between having AI capabilities and actually using them? You know, what do you think might be causing that hesitation? Yeah. It's an interesting one. It's an it's a kind of constantly changing topic at the moment, around, both the industry and around and and and around, technology providers themselves as well. But there does seem to be a bit of a lack of trust, a, because it's new. Right? It's fairly new and, having something that you've not used in the past is going to have a lot more resistance, right, to, to adoption. We are seeing that within some security teams that there is still a lot of reliance on manual processes. Some organizations, I think to to Robin's point, some smaller organizations sometimes tend to, because of a lack of resource, have no choice but to adopt that a lot quicker, or others might have a bigger process that they need to refine, in terms of, the the whole decision making. So, I think there is a little bit of lack of trust to a certain extent, but, exist and and it's the fact that, you know, it's machines and and and machines making decisions or actually providing information. So, I mean, what Chesca was saying earlier about one of the clients saying, I'm I'm glad the AI gets it wrong. You know, some some clients, I guess, depending on, their their age as well. Right? They might look at it and say or eight in the industry, so they might say, okay. They might take more trust in terms of what the machines are telling them because they've grown up in an era where Google and and sort of search engines deliver the answers for them. Whereas some, users might, tend to always look at the response of the skepticism and validate and check. So, you know, having access, I guess, to data sourcing is important to understanding where the the AI the AI tooling that you're using is deriving the information from, is something that, a lot of, a lot of a lot of organizations are quite keen to have that that that highlighted to them. And and I think one of the things that, in our capabilities is becoming more streamlined is around how some of the technology providers still, but, organizations will be using to protect their infrastructure. Right? That's gonna be using a lot more AI. So it is it is a constantly changing, and evolving situation, but I think trust will eventually go up and it will go up, especially as more adoption takes place and as, like I said, more security providers are also using AI themselves. But, it it it just, it it it like, very thing that this thing is is going to it it it's very hard to how quickly will it will it will it will take to implement, but it's something that, is, is going to take off, as well. So that's some of the thing things that we see, on our side. It's it's, it's a bit slow, but, but not always, not always a bad thing in my opinion to do that. Yeah. And it's interesting because we, you know, the report found people do seem to to trust the AI outputs, but then it also found that they're still doing a lot of those those processes fairly manually, instead of, you know, using the full capabilities of AI or automation to to offload some of that manual work. So we'll get more into the, you know, the trust versus control challenge. And, Chesca, I'll pose another question back to you. But gover like, you work with a lot of government clients. A lot of them probably have unique requirements around oversight and accountability. So how do they balance some of those automation benefits with, again, those compliance needs? It's a good question. And I think, obviously, with government inherently, there are, at times, more guardrails and more, you know, hoops to jump through before getting new technologies or new processes implemented. From sort of a regulatory point, some customers are saying it's quite a challenge. So one of my customers who works in a very secure environment basically said AI can be used on a smaller scale, but widespread AI usage across his organization is, he said, very difficult, if if not impossible as of today, because of security protocols. Now an example he gave was if the team wanted to use an LLM to help with, let's say, writing, summarization, for them, that would have to be hosted on a local service with no Internet access. And, obviously, with the world we live in, that poses a lot of challenges. So I think from a sort of the public sector perspective, that would is the main the main issue I'd say would be around approved implementation. But some other organizations, again, it depends on the the country. It depends on, you know, what's happening on the national as well as the local scale. But some another one of my colleagues has said that their organization is applying the same regulatory guardrails that they would for the adoption of any new major technology. It's not gonna it's not AI is not its own special use case in that scenario for their organization. And so with that, obviously, yes, there's still guardrails to follow, but it's not made maybe as challenging as one might think. So I realized again with my aunt, so I'm saying, well, some do this and some do that. But what I've learned as well working with government is there is no uniform approach to any forms of technology or intelligence. It's all very much case and nation specific. I hope that answers your question a bit. No. Certainly. Definitely does. And, Robin, you know, you're actually building this. How do you design AI capabilities that provide transparency and control, and what does your explainable AI actually mean in threat intelligence? Yeah. It's that's a that's a fun question. I think the the first thing to recognize here is that the task we are asking any kind of generative AI to do vary considerably. And the the level of transparency on that is, also variable as to how much, like, we need to to to provide and how much we actually physically can preside provide. Lots of people will know here that that LMs are functionally black boxes in lots of senses. They're they're not entirely black boxes. There is some kind of discoverability going on there, and we have the offering of of reasoning models now as well. But the, what we tend to take as an approach is is the the level of consequence of the decision, like, the the level of analysis that's being offered as as that increases in in terms of sophistication or or kind of, like, consequence of output, offer more transparency as much as possible. For example, in the platform at the moment, if you ask for a report, you will be able to get a, you know, full scale intelligence portal, basically, any given subject under the sun. But in order to be able to vet that report, it needs to be fully attributed. Well, we fully attribute things. And then the fun thing here is that, well, AI has has no issue with attributing more sources than a human can ever read. Right? And so what we try and do is is make sure that those queries are also fully accessible, and and kind of like, otherwise traversable by a human user. The other risk here is is just point out, like, we're talking specifically about AI here and and I'll scope this down to generative AI. But what we're more broadly talking about is, like, the adoption of, like, probabilistic decision making. And, like, that decision making is happening everywhere in our platform and other platforms users are using all day every day, whether it's machine learning models, whether it's it's, you know, recommendation models. The level of transparency you need to offer there is variable depending on, like, the consequence decision and how much it matters. Netflix, for example, needs to offer very little transparency on its recommendation algorithm as long as it sees as a happy. Right? There's very low consequence to being recommended the wrong movie. Whereas if we are recommending or suggesting or transforming some data, there is there there is consequence for that in our space. This is, to my mind, the the most joyous thing about getting to design AI systems in in security is that you live in a very high consequence world. And that means your your design needs to be accept there is there is an exceptionally high bar for successful design. So there's a there's a lot of joy in that for for from my perspective. I don't know. And but Netflix, you know, recommends the wrong movie. I don't know if I'm that happy about it, but I see your point. Let's let's kinda shift to some of the implementation barriers, that we've seen come up in the report. And, you know, when we look through the the data in the report, it definitely identified some clear patterns in what's blocking expansion. So some of the the top ones are are loss of analyst control, limited internal expertise, inflexible vendor features, and compliance or data privacy concerns. So, Lance, I know you work with a ton tons of customers around Europe. But when prospects cite limited internal expertise, what do you think they're actually missing? Is it the AI knowledge? Is it integration skills? Or or something else? I think it's, fear of the unknown in many ways. It's just there is a lot that has first of all, the evolution of this has been very fast. Right? And the movement and the adoption of it naturally kinda follow suit as as evolution moves. So I think, one of the things that we see or we get for him sometimes is that this concept that AI requires a certain skill level. It's moved so far, but, you know, we need a certain skill level. When in fact, what it's really doing is breaking down the skill barrier. Right? It's it's, removing, if you like, the need for skills. We see this both from, an analyst response perspective. Sadly, we also see it in the the, in the threat actor space as well. We see that that also, has a negative consequence of having availability to AI. So, so there is there is that that skill barrier is dropping. So, it's less about the skills, and it's goes back to our previous conversations about the trust. And it's about, you know, is the arguing in the concrete knowledge and the information that they need. And I've had, you know, people come up to me sometimes and say, this attack was done by this threat actor, ChatGPT tells me, and I trust ChatGPT. And you go back and you say, well, are you sure that's correct? Are you have you looked at what information ChatGPT is is is is drawing on, to be able to provide that information? So, some are quite happy to take that that that kind of, like, that that verdict of that of that information, others less so. So when we see that, the other thing that analysts are always gonna grapple with is, also to say, more teams is the loss of control. Right? You you mentioned it there in in or you mentioned it in in in the reports as well. Loss of analyst control probably was the biggest thing. I think it's again, it goes back to the whole thing about the unknown. You need to trust the process. You need to trust that the machines are making, quantifiable information sorry, informational responses that they can quantify and they can validate. Validation is is is a big thing for them. So that is, again, another another one of the reasons, like, where the gap is. The gap is about trusting, those, those those machines. So or or or those decision making processes have come out from machines. So that so so so that that that that really is it. This this this should, in future or as we start adopting this, we should see a a a bigger drop, if you like, of a skill barrier. Yes, there will be bigger kind of like provision of control over to, AI as it takes on more responsibility and more processing away from, analysts. But, the only way we will establish that trust or build that trust is in whether a sourcing, as as Robin was saying earlier, is is, the sourcing that we are getting information from is is is is good sourcing. And secondly, but it's giving us correct answers and we're validating that those answers are correct. And that's how trust in general gets built over time. Not it's not a AI thing. I think it's trust in general. Yeah. Great great points. And, you know, one of the last points around, concerns was was regulatory concerns. So, Chesca, to go back to you, you know, are you seeing some of your government organizations encounter those significant regulatory barriers to adopting AI? I think I think, yes and no. And I think something is, again, without sort of, you know, copying Lance's point, it does also come back to trust. So a particular different differing regulatory approach approaches from different organizations or nation states will depend on how much they trust AI. And I think something that is an interesting element when it comes to regulations, and this will be the case of many elements of organizations, not specific just to government, is around the need for ethical use of AI. And I didn't really maybe quite naively think about that until really seeing customers exposed to it and then telling me about how they can or can't implement it. So I know that, actually, there was, a GCHQ paper quite recently on the ethics of artificial intelligence, which is talking about the need for a governance system to manage AI, and with that would include specialists around minimizing bias and discrimination. And so what I've had with, various organizations across government, across c and I, which called national infrastructure, is the concern here about relying on the right kind of AI. So if they're going to implement AI, it needs to be AI that has been is both capable at where it can be of sort of applying, sort of basically, it's capable of removing bias as best it can. I also think another element that I've seen in terms of regulations is also around the security of AI practices. So government customers, CNI customers have all said that there is a whether they want to use AI or not at this stage, they need to be aware of it. They need to understand and keep up to date with AI because it's being used by the adversary. And I think that's one element of the security concern. But when it comes to implementing it in house, there's also a security element there. You're opening yourself up to a new attack surface. There are a lot of those practices. And as Lance, Robin was saying, this is going so fast that I think a lot of my customers, especially at leadership level, are conscious or should I say maybe more cautious to implement something that's growing so quickly just to say they don't mention or just to be able to get the benefits without doing the proper, sort of due diligence, shall we say, especially when it comes to security. I think that's a huge concern because, understandably so, a lot of government customers are, you know, susceptible to targets, and they wanna make sure that there is nothing they're bringing in house that's going to make them all vulnerable. If I can also interject on that, Chesca. I think one of the things that, as you were mentioning there, that I was thinking about in terms of how we view AI in general, and this is not just purely from, how the cybersecurity space sort of uses AI. I think there is a bit of a fear to it as well because of the way that, let's say, it has been, talked about in media and films and and everything else in the build up to, to the adoption of AI. But But one thing I think that I like to look at AI is it's reducing your research time. Right? So years ago, you'd say, hey, how do I, how do I cook this dish for instance? You'd go and buy a cookbook, you'd read through it, and you'd get the information you need. Then we have search engines. So you'd go online and you just Google it, and it'd give you a list of, of information. And now we just ask Alexa. Right? Or we just start, you you know, the the process of that research is cutting down, cutting down. And now we're at the point where effectively AI is like, you know, is pointing to, hey, you need to look at it. This is this is where kind of like the issue is. This is kind of what forget everything else. Everything else is noise. So I think it is definitely going to assist us tremendously in many ways. And I understand that there's I'm actually gonna be caution, on on that front, especially in in our space and and in the security space. But, it it's inevitable that it's something that will make, our lives a lot faster and a lot, easier in terms of our decision making. It's designed to make our decision making easy. That's the way I view it rather than, necessarily take decisions away from us. It's it's it's more about knowing the exact information so we know, when to make decisions. Sorry. I just thought I'd lay that instead of reference that because he was saying very about the the rep regulation side because regulation always kind of tries to catch up with, with reality in many ways. No. They're always nice. Thanks for thanks for adding there. And and then before we get to our, you know, final section and talk about where we're seeing AI deliver value versus where maybe it falls short, definitely, please feel free to use that question and answer box, and drop in any questions that you might have for our experts. But, you know, we we look at the report. The the data clearly shows patterns in where people are seeing AI succeed for them and and where maybe they struggle to see value. So there could be quick wins or or longer term opportunities, you know, when we look at some success areas, being able to score threats and and automate the way, we're able to, like, mark something as critical versus more informational, be able to summarize reports, which we've kinda talked about in this, webinar or data normalization. But there's also a lot of areas with less traction, whether it's cross team collaboration, which is interesting one, or maybe some more strategic intelligence initiatives. So I'll start with Robin here. You know, why do you think AI succeeds in some areas but might struggle in others? Do you think it might be due to technical eliminate limitations or how organizations kind of approach the initial implementation? Yeah. This one this one is is is you'd in charge entrusted with building some of these tools and trying to get them to work. It's, it's always a fun, a fun story. I think there's there's two there's two angles here. The the the task to which AI succeeds very effectively at the moment are often those tasks which require at least context under which to operate. So data normalization, enrichment, kind of fundamental, like, NLP tasks that we've been able to achieve, for a while, like, generative AI is also very good at those as well. The the the other thing is is those things don't require context or at least that much context. The tasks that require significantly more context, so whether it's strategic intelligence initiatives, whether it is cross team collaboration. Because What you're really saying about about when you say cross team collaboration, but strategic, you're like, hey. This is my organization. This is the intent that we want to go in as an organization. And for any organization, getting on the same page there is a task in and of itself, having an AI deliver successfully on that initiative, deriving that intent or value from, an organization is is pretty hard. Like, often that's simply not encoded in machine terms. And and that is the other if. It's like, what is encoded in machine terms? Like, what is actually available to, an agent or an LN to actually employ? And so we see this approach. You know, you will hear more from Recorded Future about MCP and on our strategy there over the next little bit of predict. But the the availability of tools and and their encoding is often, like, the the functional limitation. And for some some use cases, like, you really don't need a tool. You you you wanna ask a question and you want to be able to retrieve some data and get a response, and it's it is fundamentally that simple. But when we move into AI taking decisions on behalf of a user that that that are multistep decisions or require multiple components of context. Like like, how well is that context been modeled? Like, context modeling is a is a buzzword for a reason. How well have we managed to, like, attribute the right tools to the right task? Like, that tool task matching, like, that is something that any human learns over, like, the scope of their lifetime. Right? They're like, alright. I have a spoon. Do I use a fork or a spoon to eat my soup? Like, you see that with children. Right? And and the the kind of progression through our lifetimes as we enter the workforce of, hey, I have this tool available to use. Is this the right task to pass to to to apply it to? Like, LMS, the management of that tool use is is is part of this story. And we see those more sophisticated use cases necessarily requiring additional work. And I think that's the that's the progression there. It is not in any state to my mind that you're, like, genuine capability gap that they will never overcome. It is it's purely an adoption card story of saying, hey. Like, the technology is maturing. The, like, audience to which the technology is available is also maturing. Like, those two go hand in hand. It's it's about as simple as that. Good points. Hey, Lance. When you think about demonstrating some of the capabilities that Recorded Future bought to bear, well, what's their typical customer what is typical reaction that you're getting? Yeah. It's interesting one because, I think, within Recorded Future, we obviously, we've been using machine learning, kind of the name before AI became a call term. We've used that for, for nearly a decade. So the platform really delivers a lot of these kind of AI verdicts, if you wanna call them, around what an analyst should do and shouldn't do next or how they would fret profile, for instance. So, we've had it built in our platform for quite some time. So when I'm doing a demo, for instance, AI is a big part of, like, the output of what they're seeing in many ways. It's it's, AI has played a big part, and I don't think that's just a recorded feature thing. I think that's the case of, you know, many tools out there now as they're all starting to adopt this to make things a bit more streamlined, and easy, for for clients to to make decisions. And ultimately, that's really it. It's it's it's the purpose of AI is to make you, you know, provide you with information to make faster decisions. What I'm seeing kind of mostly, used in demos or where I get more feedback in demos or where I find more value is showing some of the let's call them quick wins that you can get with AI. And that is I mean, Chesca commenced this earlier, the AI report generation, for instance, having the ability to go to a tool and say, tell me, you know, what are the major threats that might impact my organization over the next, thirty days? You know, something, you know, something like it's at an executive level that you might, get asked, about your company. And we can deliver the the reports of that in a in a in a a very automated way and in a, continuous, recurring way so that they get that data without having to spend time, you know, researching that or pulling data from different sources to to provide those reports. And that's a huge, time saving, right, for for analysts to do. So that's some of the things that, you know, I think everyone tends to accept that that's a great benefit of of the use of AI, and can help. Now, you know, it comes up over the question of the sourcing. Our platform exposes the sourcing as as as best as as we can expose that sourcing, so that it gives that confidence and kind of brings down that that that trust barrier that, that exists, around AI. So those those are the things like faster reporting is is is a big thing. When it comes to some of the, let's say, the, the proactive, let's say, threat detection piece, a lot of it, a lot of the stuff that we'll do in them is we'll very much focus around these are kind of like the main, threat vectors or the threat actors that you should be aware of. And we provide the full workings out of that, which which of course, again, builds our trust, like I said, with the AI. Where we get, I guess, a bit more of skepticism is, like, I think I mentioned this in the beginning, it's the follow on actions. Right? What I meant to do next? And I think that is the nirvana of where, I guess, we wanna get to with with AI is to not bother with all of this visualization of information is to say, right, I get all of that. I just wanna know which server stuff I need to fix on my infrastructure to protect myself against a potential attack. And that's kind of where, let's say, the the skepticism is whether it's actions to consider here is x y zed, but, you know, x and y are pretty much self explanatory. I kind of know that those are actions that I have to do anyway. The zed could be interesting. And I guess the conversations that we have around that, especially when we're doing demonstrations of the product is, well, this is based on information that either, you know, the platform is gathering or information that you as an organization, are supplying us with. Right? So it has to be a two way thing, in terms of that collaboration, the cooperation with, with the organization to ensure that we are capturing the stuff that they really care about and the intelligence goals that they really want to, to to implement. And then we can provide those kind of, like, next action steps in a more accurate, fashion. So I think it's mostly stuff that we've covered pretty much in this in this discussion already. But things like trust is kind of like a common theme through trust on on those actions, for instance, are where we might see a bit more skepticism, especially in demos, about what we should do next. But another more mundane task, I said mundane. I mean, report creating, I wouldn't call it mundane. You know, it's it's, you know, I I found it mundane doing it for years, but but it's, that is definitely a time saving that, AI can, deliver, on day one. So yeah. Awesome. And, Chesca, you know, have your find your customers have changed their approach to AI in recent years or maybe even just recent months? Yes. I think there's two elements to it. I think that what I was saying earlier around how the adversary is now using AI, you know, we spoke about it reducing, the sort of the skill level that you need to be able to manipulate or, you know, effectively use this tool. If we're doing it, the advisory is doing it. The adversary, sorry, is doing it. And if we think we're moving that fast, the likelihood is they might be faster. And so I think that's been, even if customers are, you know, don't want to implement it in their own in house tooling, they know they need to be up to scratch on it because they can't they don't want the, you know, the threat or the threat actor to beat them to it. So I think that's definitely something to bear in mind in terms of how whether they want to or not, they have to join or they have they're having to ride the AI wave because we can't I think it's I think we can all say AI is not going anywhere. It's not disappearing. Whether you want it to or not, it's part now of I mean, obviously, the industry we're in of intelligence, it's a part of intelligence now, whether that's for good or for bad, however you see it. But I think also something that I wanted to mention, you know, going a bit back to Lance Moraitis-Jones sort of, you know, sort of quick or smaller wins, is that with customers who have originally been more hesitant to adopt AI, try it on a small scale. You know, always makes me think of when I remember when my sister was running a marathon, I had to have a four year, but I do it step at a time. It's not about how far it is. Just try a little bit of the time and you chip away. And I think it's the same here. So some customers who have said, well, we can't do it because we can't get rid of trust. We have to have human decision making as a priority. But then they try just little elements of AI. So an example would be around for the more maybe technical analysts in writing and understanding code and query languages with AI. So one of my customers actually said that their engineers and analysts who use KQL, they use, Copilot, Microsoft Copilot, and that can really help them. Even so it's like having a KQL expert looking over their shoulder. It's not to do the whole process for them. It's purely to help them understand this one element of code or help them, you know, interpret something. So I think my customers generally, when they've done little elements of AI or even seen how it's worked outside of work for work, maybe it's with your health care or with fitness tracking or whatever it may be using AI in that element. When you see a small benefit, it actually goes quite a long way, if that makes sense. That was a great point and, helpful advice for my marathon next month as well. So So appreciate that. Well, thanks for that and to our panelists, for walking through some of the findings from the report. Before we hop into q and a, and again, please feel free to ask any questions that you might have. I wanted to direct you to the documents tab where we have a link to our predict twenty twenty five event. So actually, our New York event is coming up next week, but our London event, which I'm sure many of you on this webinar are slightly more interested in, will be on October. I believe that is our first two day event for our London Predict event. It's always a great time with a lot of expert speakers, practitioners. You get to meet a ton of people from Recorded Future as well as network with people across the threat intelligence industry. So, if you do have any interest in registering for that, please check out that note in the, documents tab. It should say predict 2025. Awesome. Yeah. You get a chance to meet us in person as well, which I'm sure you'll be looking forward to. Especially after this. Alright. Yeah. Thank you, Marie, for adding. There is also a complimentary code to register for Predict for webinar registrants and attendees. That code yeah. If you are watching this after, you can still use the code. I just wanna make that clear. But that code is pre twenty twenty five red, reddisher, and you can do that at the link. Great. So to so to hop into the, q and a, you know, what are what do you think are some overlooked quick wins where AI could deliver, I don't know, immediate value, immediate ROI, for threat intelligence workflows? Kinda pose that around to the group. I'll I'll I'll hop in on to You go ahead. Yeah. Yeah. If you want to go. Like, it's so so the the two obvious ones are, like, summarization and reporting. Like, summarization, there's a lot of data out there in the world. Like, you don't need to read absolutely every every every time of it. Like, it's that that that is is is really obvious. And the the next thing is is obviously, like, the reporting capability we talked about a lot is it's essentially summarization of scale and summarization from from relatively high quality data sources. The the third one I'll I'll mention, and and this is, something that I think Recorded Future have been doing for a very long time that our clients have been doing, but aren't isn't necessarily kind of encapsulated in an AI workflow is very simply entity enrichment. Like, tell me more about this thing. Mhmm. Super valuable, really low key, analysts have all the opportunity in the world, so that's it. Not a complicated workflow. You don't need to be building 15 different agents all talking to each other and implementing every tool under the sun. Every everyone's everyone's a winner with that one. Yeah. I know. I I I was gonna say very similar things, to be honest, Robin. I think that that's that's pretty much what, one of the the main quick wins is that that we have seen. Other things that we are starting to see, and this might not be a more widespread thing. These are maybe a couple of, call them, outlandish examples where, clients who perhaps are in a, high-tech space, let's say, or or in the financial technology space as an example, that might be adopting technologies at a far faster rate than others. They because that's their business model. That's what they have to do. They might be accessing a lot of data, a lot of this raw data. So AI becomes a very big, it becomes transformative, right, for them. And they're looking at things like return of investment. They are looking at ROI. Right? They're looking, well, how can this we have all these data points. Where can we streamline processes? Where can we, you know, ensure that we are spending less money as an organization or, you know, where we should be spending more money as an organization to to improve our, to to improve our growth right as a company. So there is some movement towards, and again, this is probably very early stages on some of these clients there and and some of these sort of these these spaces where where clients are using this, to effectively use AI to also, you know, to customize it for their own business needs, let's say. Right? And that's kind of where they can build things like ROI models. I know ROI models are, you know, exist today and a lot of them use sort of, similar things on that. So, that's quite interesting. Awesome. And see if you start to, more come through the chat. Interesting question here. So do you feel there is a danger of having kind of a two classes of enterprises? You know, those that can afford to utilize AI tooling and those that maybe given their budget constraints can't. And, you know, is there danger that those organizations who can't embrace AI tools and platforms could get left behind. I think I just dead one. Sorry, Lars. Go ahead. Just get that. That's fine. Honestly, away from the technical perspective, I think it's is not just specific to AI. I think this is specific to all technologies. And I think if we think about threat intelligence primarily, A lot of what I know we are biased. We work in the private sector, but a lot of a lot of the resources that you get for threat intelligence can come from private sector partnerships or come from, you know, new, technological productions and things like that, which all do cost money. And I think whilst I realize I'm not reanswering the question, I don't necessarily think this is an AI specific problem. I think this is much more universal to not just our industry actually, but to all industries is, at the end of the day, it comes to if you're aware of what you want to implement and if you can implement it. And I think a lot of that is finance and a lot of that is also if it's not broke, don't fix it approach. And so, yes, I do think that there is a danger that this might happen here, but I think that's a danger that that might happen in many spaces. Yeah. And I I to to kind of second Jessica's point, but get into a little bit more of the, like, the AI weeds of things. Like and it's a really good question. I think the, the cost of AI are often, in terms of the public perception of them, seem to be very, very high. You have all of the the major AI providers talking about all the provisioning so many billion dollars worth of additional compute. NVIDIA's stock price is the kind of, story itself here. The the flip side, I think, is actually, like, AI is extraordinarily accessible for for a relative massive, like, the the cost that is at initial ingestion. Like, one of the friendly things is it's rel like, the the the, development of, like, usage based pricing models. You can stick your feet in Adam and and not get scolded. You don't you don't have to jump in all the way straight away. And I think that means it's a little bit more amenable. I think the other thing here is there is there's two tracks here in terms of organizations. There's there's the kind of buy side of of a lot of organizations and there's the build side as well. I think going the build side, pursuing that as a as a story, I think you need to have the talent in house, and the the talent is the expensive part there. Yes. There is some kind of additional compute costs and and and data engineering, whatever else that may come associated with that, but I think your your greatest cost is talent. If you go the buy side, then obviously, like, it's it's the commercial story. Right? And and and how do you get the biggest bang for your buck? The risk, I think, is not actually necessarily organizations getting left behind because I think the market will pick up those organizations. I think it's the the risk is that those organizations suffer at, like, at the federal consequences because they don't have the right tooling. Right? Because, essentially, the threat act innovates faster than them. And, ideally, you either have the account in house to to build whatever you need or you have, you know, you go to provide a, like, recorded feature to to buy what you need. I think that's the the challenge there. Awesome. And I know we're coming up, on time here. So last is we'll get to this one last question from the chat. And, again, thank you for everyone who submitted a a question throughout the presentation. Lance, I'll post this one to you, but how do you go from a lot of weak signals to a result that is trustworthy? Yeah. That seems to be one of the challenges we we we face, on on our side as well on Recorded Future side, to be able to provide, if you like, high fidelity information, right, when you are looking at data coming through. So weak signals is, you know, doesn't necessarily trigger anything. Right? It doesn't trigger your instant response tooling. It's not gonna trigger anything like that. So context is important. It's it's whether we, whether those weak signals lead to something else. So weak signal in itself, on its own, might not be a danger. Let's say an ILC, right, or an IP address that has been used in the past for, you know, command and control server that hasn't been used in two years. Right? That's not a threat. Right? As far as, we are concerned. So, but it might be the start of something. Right? If you start to see a lot of that, you're getting those indicators in your environment. And and the way we would look at that is to see, well, do we have a server set up on the front end of that IP address? Do we have a mail server set up in front of IP address? Is there, you know, we look at the way that, sort of domain abuse occurs or type of squatting occurs. It tends to be first with domain gets registered and then nothing else happens and, you know, you are sat there with someone trying to emulate an organization's domain, but it there is no risk. Suddenly, this IP comes into play, which, like you said, it was a weak signal, but our weak signal might become a more powerful signal because it was used as a command and control server. We have got indicators now that it's being used or primed for weaponization or being used again, but it's not active. So weak at the moment. Right? But looking at that further context, that's kind of where it now becomes an alert that you should be concerned about. This is just one example. Right? There are other, combinations, if you like, of of indicators that together might make up, a, a threat, right, let's say, to to the organization. So, AI is gonna be a big driver for that. This is not just a Recorded Future thing. This is gonna be, I think, in general because we're gonna be seeing lots of those and we need to sort of filter and sift through the ones that are, more valuable to your organization. And that goes back, in in part to what I was saying earlier. You know, it has to be collaboration. Right? Because let's say if you are an organization that, you know, maybe let's say you're a service provider as an example. You have a lot of infrastructure data that's that's, gonna be constantly abused potentially, by by by threat actors or as a target for threat actors, let's say. So you might have a very different approach to this, and you might have a lot of weak signals that you need to be able to manage. Other organizations might be small in size or they might have very different, business models. But for them, it's a case of looking at stuff that's industry specific. Right? Specific to their organization, to their industry. So getting those signals build a bit more relevant to to to the organization and to the, and to the industry in general that they're part of is gonna be more valuable. So it's about kind of taking that big funnel approach and sort of sifting through it, and that's a big part in the fact of what, AI is gonna do, and it's doing it already. I mean, we've been doing this for quite some time. We've been looking at the high fidelity indicators of stuff that's actively running. We put a a a risk against those. But also, like I said earlier, we do look at context as well. We do look at whether there is a bit of weaponization happening in in in in in in the use of that IP address that that gives us a little bit more, risk analysis on that. So so so that's, I guess, a very good example of how AI can be used to start to build the results trustworthy. Now again, I I I'll give the the naysayers among you might say, but again, you know, you that might be wrong. It might not be necessarily the same the case with every indicator. Completely understand that. But it's something that it's going to be refined and the more that these machines that we are using today to, provide these results get more of this data and we start to understand this data, a lot better. It's gonna become second nature to us in the same way that, a Google search becomes second nature to us today. Perfect. And I think that's a a brilliant place to wrap, but, appreciate everyone for hopping on today's webinar. Again, if you haven't registered for Predict or you're interested in registering for Predict, you can use the code, in the chat there pre 2025 for our Predict London event on October 2022. Lost it there. And then, yeah, thank you very much to our panelists from across Recorded Future representing the product team, our sales engineering team, and our customer support team. It was great and amazing to hear all of your incredible insights. We really appreciate it. Thank you, everyone. Thanks very much, for giving us time. Thank you.