Name: From Automation to Intelligence
Uploaded: 2026-03-14T08:16:09.572Z
Duration: 1 h 8 s
Description: From Automation to Intelligence

Transcript for "From Automation to Intelligence": Good morning, good afternoon, good evening from wherever you're joining us today. Welcome to our webinar, From Automation to Intelligence. I'll give a little bit more of an intro in a second, but before we do that, just some housekeeping items. Best webinar experience, please use Google Chrome or Fox. Audio will be streamed through your computer. There is no dial in for this, so make sure your volume is up and you can hear as well. You'll be able to download the presentation slides and some other resources via the resources widget that you'll see within the webinar interface. All of those widgets can be resized to sort of give you the best viewing experience for what you want to sort of give more dominance to in terms of your screen. An on demand recording will be emailed to the attendees here today about twenty four hours after the webinar ends. We are going to take some Q and A at the end of the session today. So please use the Q and A widget to get your questions in there and we'll do our best to get through as many questions as we can live. Otherwise, we will follow-up with answers to those questions after the webinar. So with that, let's jump in. Like I said, welcome to the webinar From Automation to Intelligence. I'm super excited to be here today with Walter Pasquale, who has done a number of webinars with me in the past. Always great to share the stage with Walter. Quick intro from my side. I'm the VP of Product Management here at Deltek, lead up a number of our product management teams focused on the architecture, engineering and professional services industry. And Walter, why don't you do a quick intro from your side as well? Thank you so much, Brett. Really great being back with you and sharing the stage with you once again. Well, for everyone who I'm just getting to know here, my name is Walter Pascarelli. I spent the last ten years working in AI strategy and policy. I've been working closely with Deltek on really trying to understand what matters for C suites leaders and for people who want to bring AI into their organization to really develop the raw capabilities to make the most out of this incredibly exciting technology. So we have a lot of ground to cover today and I'm very excited to be having this conversation with you today. Yes, definitely. So let's talk about the ground we're going to cover. So real quick, some agenda items here. We're going to start to sort of level set and set the stage for what we want to talk about today. The big focus of our discussion today is really sort of looking at companion AI versus agentic AI. We all know agents and whatnot are all the rage these days as well they should be. So we want to talk about the significance of that and why that may matter to you. And, you know, the really what we're seeing today is the shift from automation to intelligence. And I think that's absolutely critical for businesses and business leaders to understand how they can leverage that within their business. And then we'll talk a little bit about, you know, what we're seeing in the industry in terms of how leaders and how businesses are leveraging and navigating AgenTic AI. Now, from a sort of level setting perspective, you know, one of the things we want to talk about today as well is the C suites playbook on Gen AI adoption. Now, Deltek partnered with Walter over a year ago now to put the first edition of this playbook out. And we have gotten a ton of mileage, a ton of great feedback on this playbook. You can see the number of downloads as well as some of the engagement that we've had around various events and webinars that we've done in relation to this playbook. I've done a lot of sort of speaking opportunities where I've shared this playbook as well, and we've gotten some great feedback. The great news is, is that Walter has taken some time given the evolution of the technology and the AI sort of improvements and advancements that we've seen in the past twelve months. And we are working on an update, a second edition, the 2026 edition, if you will, of the playbook. And we'll talk a little bit more about the changes that we're going to see in that playbook, and give you all an opportunity to get access or early access to that playbook so that you can use it, within your businesses as well. But before we, sort of talk about the key topic today, I just want to do a quick intro to level set sort of where we are and why we're talking about sort of the shift from automation to intelligence. And if you've been and your business has been experimenting with AI, you're in the mainstream like most organizations and many organizations started with AI as sort of a productivity tool, whether it be drafting content or summarizing content or more intelligent search, getting through first passes on content a little bit faster with a little bit more efficiency. That's really sort of what the focus has been initially around AI and generative AI specifically. But the reality is, is businesses like yours, project based businesses, you know, you don't win or lose based on how fast you can write an email or how fast you can sort of summarize a document. Your businesses win and lose on decisions, on pricing, on staffing, on scope control, on forecasting how things are going with your business and, you know, catching risk early so that you can do something about it before it becomes a significant problem. So with that in mind, if you look at, you know, what's been happening around the industry and business in general when it comes to the ROI businesses have seen in relation to AI, you know, there's no doubt that the sort of general consensus out there is that, hey, we are definitely seeing positive ROI in terms of what we've been doing with AI, while it has been sort of largely focused on that more of that automation and that type of thing so far. Nonetheless, the fact that we're already sort of getting this traction and this ROI when it comes to AI is definitely important from a foundational perspective and definitely important when people ask the question, which still gets asked these days, you know, does AI work? Does it make a difference? And clearly you can see from these statistics that, yes, it definitely does. And as we sort of look at the way different or the way businesses are using AI differently across their organizations, you know, what you're seeing on the screen here is pretty common in terms of what we're seeing. A lot of organizations are still sort of in that assisted AI world where, you know, humans are using the tool sort of manually, like I said, to do things like write email fasters and summarize content and draft first passes on content and things like that. But we are starting to see a shift in terms of automation, still humans sort of involved in the automation side of things. But you are giving the work to a certain extent to those digital workers, to those agents. But where the shift is really starting to happen and what we really want to focus on today is that AgenTek AI, where we're turning over, you know, more of the decisions, more of the work, giving a little bit more autonomy to these agents to let them do more of the work and leverage that intelligence that is really gaining traction within AI today. And there's two sort of phrases or terms that we want to talk about today. It's companion AI, And this is what you use for sort of high stakes judgment, a copilot, if you will, a second opinion, really human in the loop. And then of course, there's AgenTik AI, which is, let's call it bounded autonomy, where digital workers with some guardrails that we as humans can monitor, but they propose actions. They, in some cases, will take some of the actions if they fall within those guardrails, within the limits that we've put out there. So it's that companion AI and agentic AI that we really want to talk about today and start to understand how we can leverage, how businesses can leverage that agentic side of the AI more than perhaps they have in the past. And I think, you know, sort of a key insight to end on before we jump into some of Walter's content is, you know, the universal entry point for AI is sort of that time savings component. You know, 76.7% cite this as primary benefit. But I think there's a risk there. There's sort of a risk of getting caught in what we're calling sort of the efficiency trap and really only using AI for that efficiency piece instead of using AI in an area that I think can really prove very, very valuable around decision making, around reducing risk, not just doing work faster and applying AI to sort of the same way we've always done things, but really leveraging AI in new ways and focusing on decision making, focused on some of that more autonomous work, that more digital worker type of approach as well. So with that, we want to jump into a quick poll and then we're going to, you know, turn it over to Walter and we're going to get into some of these contents a little more detail. But before we do that, a quick poll, you know, how are you using AI today? You know, this will be a good sort of, again, level setting for Walter and I to to see how the audience is using AI today. So if you could take a moment to respond on the screen here to how you're using AI today. We've got sort of four choices, mostly manual copilot loose, some workflow automation, some are piloting agents, and, you know, some of you maybe have not started yet. So let's take a look at that the results so far to the poll. So a lot of the Copilot use and actually that 83.3% is, you know, pretty consistent with the number that we showed just a few minutes ago. So not a big surprise there. Great to see some of that workflow automation, you know, even that is pretty close to one of the numbers that we showed early as well. All right. So with that, I'm going to jump into, you know, the shifts that we're talking about from automation to intelligence, from sort of the doing work faster to the decision making and some of that more autonomous work. So, Walter, why don't you talk a little bit how we can cut through the hype and get into this more of this intelligence side of AI? Brilliant, thank you so much, Brett. Well, first of all, one of the things that we should really do to ground our discussion in what's going on is a very quick overview of how the field has developed in the past years. Now, most people are at the very least aware of some of the tools such as say Alexa or Google Home, Google Echo that were essentially these kinds of devices that we would have either in our offices or in our homes that we could ask questions to and we would essentially say, hey, what's the weather today or what's a good event near me and so on and so forth. Now, the one thing to explain here is that these kinds of tools were not really AI in the modern sense of the word. So essentially think of it like a massive decision tree where if you would ask a question, it would give you an answer based on some of the inputs that you would give and say, not this one, but this one. So essentially it's not really a reasoning ability that these kinds of tools had, but it was a lot more like if this comes up, then we go for this one. But really the past years, especially since late twenty twenty two and let's say 2023 has been the emergence of co pilots and essentially assistance. Now these are tools that are fundamentally reactive. So that means that I as a user have to ask it a question and it gives me an answer based on that. It can do some reasoning, it does a statistical analysis and it predicts essentially what is the most likely answer that a user wants to have from me. So that's where we've been in today. With AgenTik AI, some things actually change and the main thing here is that with AgenTik AI, we're not just talking about a tool that sits within your browser, but it's a tool that can essentially go out into the open web and autonomously complete tasks. So what does that mean? Instead of having all the data that is essentially within the model, it can go on government websites, it can look on PDFs that have been published online, it can look anywhere where essentially there is publicly available information and that means that we're the one hand not limited to what is in the model, but more importantly, it can actually have some goals and some tasks that fulfills for us. Perhaps the newest transformation and I think that's actually a result of people using some of these AI systems so much is what we call Companion AI. Now Companion AI is slightly different from the traditional co pilots and assistants in the sense that it has a continuous relationship with us. It is context aware and critically and that's one of the main big technological changes is that it has memory. So it can remember the things that we asked it, our style of communication, the things really that matter to us and the information that we want to have highlighted in the answers that request. And to be aware, I mean, there's been quite a bit of hype in the past week, particularly with some of you might have heard the case of Moldbot that I think is really a case in point for how Agenetic AI has really captured people's imagination across business, across society, across academia. In case some of you might be new to this topic, Mobot was last week or actually the past two weeks, the most cited topic within artificial intelligence and it's a wild story. So let me break it down very quickly to you. Moldbot was essentially a GitHub project. A GitHub project that was provided by developers that was essentially an agent that could go out and do particular kinds of tasks for you and it became viral on the web after a user asked Moldbot back then still Cloudbot to go out and make a restaurant booking and a reservation for a user and it did it by getting access to various tools. One of them was called Eleven Labs, which essentially generates hyper realistic voices and it successfully made a booking for a restaurant. So it looked into the calendar of the user, it had a dialogue with the person at the restaurant and it made that booking. So this is kind of the first case where we can see, oh, this is really actually autonomous and rather than me having to say do this, do that, do this, it divided the actions in various steps. First it completed this, then it called the restaurant, then it looked at my calendar and so that makes artificial intelligence significantly more useful. As a result of that, it gained about 160,000 users in just one week. Now, a few things happen after that. Number one is, as you might see the name Claude is actually very similar to the name Claude by Anthropic and the developers of Clawdbot had developed it as or had called it Clawd because it would have claws or hands, so it could go out and be more practical, which is essentially a good thing. But in the second week, Anthropic launched a cease and desist letter to Clotbot, which they agreed they would change because the name was so similar, but in the time that they changed the name in that ten minute time window, some crypto scammers hijacked the social media handle and started developing a new token called CLOT, which they essentially did a classic pump and dump scheme where they kind of boosted it up and because the project had gone so viral by that time, it accumulated market capital of about $16,000,000 before crashing to zero. And then because things essentially hadn't gone far enough, in week three a young Austrian entrepreneur created a social media platform only for these clawbot agents which were essentially there and they could interact and what these agents started doing is that they essentially started having debates about consciousness, they even developed a religion whether we're worshiping a space lobster, so very strange story which prompted people like Elon Musk to say this is the beginning of singularity, Andrew Carpathi who was one of the co founders of OpenAI says this is a sci fi takeoff. All of that happened until January 13 and there what happened is that an investigative platform called four zero four Media revealed that a lot of these API keys, so essentially the passwords to control the agents were out there in the open. So it was essentially a massive data breach if you want or not even a breach, was essentially a massive data leak, a data offering and that meant that any of these agents were basically accessible to anyone. So people were essentially able to post anything and that meant that it wasn't just these agents posting, it might have been a human pretending to be a bot. Now, why does the story here matter? Apart from all of the shine and all the hype that it created, there is a couple of essential lessons that this tells us about the future of artificial intelligence. The number one is that autonomy is increasingly possible. So it's applicable, it's deployable and we've seen it happening across several use cases that we're able to audit. But the second lesson here is that while this was a very exciting project, there is really still a tension between on the one hand the safety of these agents and on the other hand the autonomy and that's not set in stone, that's a problem that will be solved that is being solved, but essentially the key trade off here is that if you have one of these agents and you keep it overly safe and you limit it, then essentially becomes a chatbot again. If on the other hand you give it maximum autonomy, can create, it can actually have a lot of independence, lot of proactiveness as well, but then the issue here is as well is that it becomes less safe. So we need to sort of slowly shift that balance. And finally, the response and the reaction to that is potentially the most interesting thing here and that I think really highlights how these themes around autonomy and agency have really become top of mind for decision makers. Critically, one of the main interesting things about Mobot and generally around the field of artificial intelligence as a whole is that we're really seeing a shift from artificial intelligence that is not just saying, can I do for you? But it's AI tools that now proactively say, consider this. And the main step change here that we saw from the main AI systems and co pilots here is that we see users increasingly having an ongoing relationship that anticipates needs, which means essentially we're interactive. We're having that tool that is increasingly getting to know us and it continues. It understands more and more our context because again, it has memory about us. It gathers data about us, it puts it in context. To give you an example, could be project assistance that remember some of our client preferences, financial advisors that might track our risk tolerance evolutions and really understanding the patterns that adapt to our knowledge gap. I might pursue projects in a particular kind of way, someone else might be doing that in different kind of manner, but critically the B2B distinction is different from the AI companions that we've seen like people asking advice on their personal needs is that can really help us understand better like our context within productivity. So what makes me work better? What is it that I need at what time and can you maybe remind me of this thing that I do. So the proactiveness is I think the critical distinction here. And if we compare it with for example traditional Copilot, AgenTeq AI, I think this is kind of like a very good overview that you see here. Traditional co pilots, I have to initiate everything, please do this. AgenTeq AI means there's some autonomous execution within a certain scope and with companions we're seeing increasingly that these tools are becoming more proactive and they really understand our needs and our context in a way that is different from what any co pilot so far has been able to do. So this is just a little bit that I wanted to share about the evolution of artificial intelligence and where we are heading right now, really watch the space, the interactivity and the autonomy is really the next thing that will be happening. Yeah, Walter, I mean, you and I have talked about the whole Moltbuck story offline and I think we're both fascinated by it. You know, whole crust of Farionism religion that you mentioned interesting. One of the other things I found super interesting was there's a thread on Multbook about where the best places for agents to live. And it was recommending that every agent, you know, pack up and move to Dubai. So it was interesting to see sort of how how that was coming to life in regards to these agents as well. So, yeah, it's a very interesting story. But as as fascinating and as sci fi some of it might be, as as you pointed out, it really represents sort of that next level of capability that that we're starting to see come to life. No pun intended. And I think that's that's very, very, very interesting. But aside from sort of the, you know, the fascination around it, obviously, we want to help our audience understand, okay, with these new capabilities coming to life, how can leaders navigate this and how can they leverage, you know, this more autonomous sort of companion AI? And before we do that, again, just to level set another poll, you know, we're curious, you know, how the audience is expecting outcomes around AI and what's being prioritized right now in terms of their use of AI. So if everyone could take a minute and give an answer to this poll before we jump into a little bit on, okay, how can businesses and leaders navigate this new world that we're seeing come to life? Then there's a couple options here, you know, time savings, better decisions, risk reduction, new capabilities, revenue growth. You know, all of these are things that that Deltic is seeing amongst our customer base. So we're curious to sort of understand how people are using this. So as the results come in, time savings, again, not a huge surprise there. That's sort of become table stakes when it comes to AI today. Great to see more around better decisions, new capabilities and even the focus around revenue growth. All right. So with that, you know, Walter, let's talk about this C suites playbook, because when it comes to enabling leaders to navigate generative AI and some of these new capabilities that are coming online. This is a great tool, to not only educate, in my opinion, but also to give, a bit of a platform, if you will, for businesses and leaders to sort of, you know, understand how they're doing, where they could do better and what's important. So can you just take a minute and give us a little bit of an overview of how this playbook works, what it covers and how businesses can benefit from it? Absolutely. And Brett, one of the things I immediately agreed with you and also seeing at the results of the poll, I think the better decision is I think one of the most valuable things that I think we're seeing out there and also the revenue growth is very much in line with some of the patterns that we've observed. Now, as far as the Playbook is concerned, we launched about a year ago in fact, the C suites Playbook on generative AI adoption. Exceptional feedback, everyone that we've been sort of in touch with has found it useful really as a tool for introspection and asking themselves where do I currently sit and what is it that I do need to do next. However, as we all know and as I think as we've kind of been discussing as we've been seeing throughout this webcast so far, is that the field of artificial intelligence is changing and that introduces essentially the need for looking at the field of AI through additional lenses. Now, we still believe that the original framework, which is essentially readiness relies on fundamentally five pillars. So, we want to have the right kinds of strategy, we want to have the right kinds of infrastructure and integration, critically we want to have the right kinds of compliance and governance, the monitoring and evaluation because sometimes when we speak to people they're like, okay, we brought AI in and kind of like, yeah, we're done, this is it, we're now AI ready, but it's an ongoing thing, and then finally the culture. Now, the new framework provides like the previous version a self assessment tool which we have now expanded into a couple of areas that we think are critical. Now, number one thing is the shift towards AgenTik AI and that means that essentially we're not just looking at AI as the static tool, but we're looking what kinds of questions around autonomy are we asking? What is the scope of action that we give these agents to be able to act and how much autonomy and how much independence are we giving it and are we thinking about this intentionally and with the right mindset. The second point is one that has almost been somehow like everybody's dirty little secret and that's the fact that people increasingly rely on it for decision making and I think we saw it in the poll that we just put up just a minute ago that people really use it for better decisions. There's a right way of doing that and that means that essentially we look at for instance how do we discern the outputs that AI provides, what is really the right workflow for us to ensuring we know how to interpret the outputs and there is also perhaps a less ideal output which is essentially that we use these kinds of advice that AI provides us without actually thinking ourselves, which is also something that is known as cognitive offloading. Now, this is something just to put it in context that is not just an analyst that is essentially, oh yeah, I want to outsource my work to an AI There was just a case coming out a couple of months ago where actually the Swedish Prime Minister admitted that he was using some of these language models for bouncing off policy options. There's been a lot of pushback against it publicly, even though I'm convinced that there's going to be a point in time where eventually people will be expecting to essentially bounce off thoughts and ideas to some of these language models once they become trusted and advanced enough. But moving on beyond that, what you can see here is that we really want people to use these tools to really understand where they sit on their journey and the idea here is that this is a tool and a self assessment that you can use individually, this is one that we encourage you specifically to share it among other leaders in your firm so that you can really debate this and create almost like have this diagnostic effect of where you are. This is I think really the benefit of that and you can see here we have a number of questions that we ask, for example, do you understand how generative AI can create new competitive advantages and we look here at use cases, we see here whether there's some goal settings and KPI setting and as you can see here, there's also an additional questions that we included for agents, which is whether you know and whether you have specifically designated in your strategy where these AI systems can actually act. Now one of the things that I think is really great about these tools is that we then provide also like some actual roadmaps. So we don't just say this is your score, good luck, but we actually also provide like really some actionable key takeaways of what you can do on the back of that. So diagnose the issue, provide some opportunities and some options for moving forward. So this is just the playbook in a nutshell. We're very much excited to share with you this new version. As I said, the previous version had exceptional feedback and I'm confident that this version here will really help you stay afloat and provide you really with an understanding of what's to come in the next couple of months and what you should be looking out for. Yeah, so while our audience responds to this poll and the other thing I would just throw out there, you know, if people have any questions they'd like to throw at us, go ahead and fill those in the Q and A widget, and Walter and I will get to those in a few minutes. But Walter, I've got a few questions, sort of follow-up questions based on your initial content and on the playbook. You know, one of the things that I've been sort of communicating and as I talk about the playbook is the assessment and the score you get from it is certainly interesting. But would you agree that the process of sort of doing that assessment and socializing that within your organization is almost as powerful as the score itself? And how do you view that part of the playbook? Absolutely, I think if anything, and I think you're spot on Brad, I think the playbook itself can be used as the North Star and almost like maybe if we even wanted to go further with that, it's your opportunity to think about it. It sets the context and it's almost like, okay, we have this tool now, now let's debate this. The friction between me saying, I think we're sitting here and someone else saying, actually, I think we're here, allows us to really uncover what is going well and what we need to improve. And I think perhaps the number one thing that I think people liked and found valuable about the framework is that it kind of gives you no excuses and because the questions are targeted in a way that it's not just, yes, we have an AI strategy, but it's really what does it look like and what is a good approach and what are the key things that you need to take into account. Now, we've designed it to provide some flexibility and I'm saying this because we cannot provide a cookie cutter approach to AI adoption or to AI strategy because if everyone develops the same AI strategy, then kind of nobody has an strategy, right? So the one thing that we're doing here is this is your opportunity for conversation. Yes, you can use it as an intellectual exercise out of your personal interest, but really the benefit of that is in the conversation that you will have with your co workers in that friction. That's where the space of growth lies and that's how we encourage you also to use it. Yeah, So a couple other questions Walter I have for you based on some of the content you covered and just some of the conversations you and I had leading up to this webinar. The talk of agents obviously is all the rage these days, right? And it's almost gotten to the point, at least in my experience, that everyone's calling everything an agent, almost to the point that nothing is an agent anymore, so to speak. And when I talk to business leaders and other customers at Deltek and whatnot, there's there's sort of this common question or misunderstanding, whatever you want to call it, that what what really is an agent and and how do companies build them? And I'm curious, how how do you react when you hear that? What is an agent and how does a company go about building? Because it is sort of this nebulous thing out there for someone who's not seeped in the technology. So how do you respond to that? Yes. So there is essentially perhaps an analogy that is sort of widespread and that's before we go into the technical elements of that. The usual analogy that we hear people make is basically having a remote intern that works for you and that is on some level you could say humanizing it and I will be very careful with that because we shouldn't humanize it, it's still a tool, but the idea here is that we have an assistant that allows us to be more Now to give you an example, we are very heavily involved in developing explainer videos for understanding AI on a technical level and I had one of my analysts that essentially developed an agent that interacted with script writer, with our designer and with our video editor and so she essentially will be saying to this agent, go and provide these various instructions once one person gives you some info you go to this other co worker and you're essentially able to interact with. Now this is not fully autonomous and I also wouldn't want this to be fully autonomous because I think there's still some space for tweaking required here and so that we can really work together but speed things up so that when I'm traveling for example, I don't have to check my inbox and provide some very basic answers that maybe I could do in a minute while maybe I have an agent that does it for me. Now, these are low stakes decisions. If something goes wrong here, there's nothing sort of dramatic that happens, but sort of the idea of a digital assistant, virtual assistant that works for you and can answer the basic questions is one thing that I think is fantastic about it. So when we look at adoption and it is obviously the question of how do I build an agent? There's a couple of approaches here, right? There are some people that develop kind of these custom GPTs that was all the rage a year ago where people can essentially contextualize them and give them a bit of autonomy, but really from some of the newer studies that we looked at two weeks ago in early January, the majority of organizations that are actually more successful with that and I think it's about 76%, it might be 72%, something around that range, the other ones that actually partner up with other providers and the reason for that is that you want to be able to specialize, you want to be able to really understand how do I do that and me as an organization, as an advisor, as a consulting leader myself, I do not always have the full understanding and the capacity to develop that. So the adoption is one thing that ideally you want to do this with trusted partners that are out there in the field. That's not my opinion, I should say. Those are not numbers that are just fallen from the sky. Those are numbers that we have from several reviews and literature that we've been scanning for the past months. Yeah, and I think that makes or is aligned with a lot of what I'm hearing and quite frankly, experiencing ourselves, you know, Dell Tech for our own internal processes and things like that. You know, we're leveraging key partners to build some of the agency and agents that we're using to, you know, sort of create that remote intern, like you said, exactly. I just as an example, to put a finer point on, you know, what an agent might look like, I've started to do sort of an internal podcast for my teams on AI just to keep people updated and, you know, what's going on, what are sort of the internal use cases that we're solving in Dell Tech, what's the industry seeing. And I actually have an agent that goes out there and does research for me and sends me a package back that I can use to sort of build this podcast. So it's yeah. Is it is it doing it completely on itself and just generating the podcast? No. It's doing some research based on, you know, prompts that I've given it, and then it's, you know, sending me a package back that I can use as a as an input to the to the eventual podcast that that gets produced. So and we're leveraging, you know, tools from partners, our trusted partners to to build those types of things as well. So, yeah, I agree with a 100%, you know, partner with the vendors that you trust because chances are they're going to give you that agentic capability that you can start building things for your business that are specific to your business and probably leverage tools that you maybe have already invested in. So a couple other things before we maybe get to live Q and A. There's all kinds of things I want to ask you. But the decision making piece, I think, is key. We've talked about that a few times. And I think that's really a huge opportunity for a lot of businesses and business leaders. Where do you see the companion AI that you talked about driving sort of the biggest decision quality gains, if you will, first? What are the key sort of use cases there now? Yeah, so the number one thing to say here is that there's kind of two layers. One is me as an individual and then is one me as an employee essentially, right? We ran a study in November that we asked people what to use AI for, what questions do you ask it? I think the answer to summarize it or the way to summarize it is more what do you not use it for? I think that might've been a better approach for us because we have seen people across, think it was 1,500 people across The UK and The US using it for everything and I mean really everything, health advice, financial advice, legal advice, their personal health, their relationship with their friends, everything. There was at about, I think on average between 65 to 70% are doing that at least once in the past three months. These are high stakes decisions. The second point that I think is interesting is we then ask them based on some traditional sources of experts, when you have an output from an AI companion and when you have a recommendation from a doctor or a financial advisor or any kinds of traditional or trusted friend, who do you prioritize? And about a third had at least once prioritized the output from an AI. Now, this shows that if you're a business leader, people might not be open about it because there's still a bit of a feeling of oh, I'm using AI and so on and so forth, but people are using it in every area and most likely also in your business. So you want to get a hold of that and develop the right policies and again based on the playbook ask the right questions for that. As far as individual adoption within a word context is concerned, there's a couple of interesting points here that I think are really interesting. All of the evidence that we gathered throughout last year and I'm talking industries such as financial services, consulting, research science, even creative industries shows that the top performers who are able to ask advice to an AI system and do it in the right way, increase their productivity by about 30 to 50%, massive. People who use it incorrectly or just outsource basically decisions to them had either no improvement or a maximum I think in investment advisory and in consulting, was about a maximum of 6%. So there was a couple of things here that this tells us. Number one is, I think this kind of like brings us back to the question about the future of work and jobs and so on. Your excellence cannot be automated. The excellence that you have as a human, the knowledge, the ability to curate and select and discern, that's the number one skill you want to focus on and that's independent of AI. Because the people who are usually able to use these companions to generate tangible outputs to bring it back to what our respondents just said to create better decisions are the ones that on the one hand have good reliable tools, but at the same time the one who know how to discern, to pick among infinite outputs say this one, not this one. In the same way that when I hire an electrician who comes to my flat and turns a screw and charges me $500 for that, it's not the fact that he just turns a screw, it's the fact that he knows which screw to turn and that's kind of the same philosophy that we have here. So the number one thing here is that the skill that we need to practice is the discernment, it's the judgment and it's our ability to have trusted agents and companions that we know how to communicate with safely. Yeah, the discernment thing I think is key there. You made another point that I think is important to say more often, quite frankly. I'll say that I'm one of the people that falls into the category you described in terms of I use AI for everything, you know, personal health, personal stuff, work stuff. I bounce, you know, I sort of see it as a strategic adviser. And as you've mentioned earlier, it has memory about me and how I work and whatnot, and I bounced thing off of it. But I think I have enough discernment to know, you know, which stuff to use and which stuff not to use. And the other key thing I think you said there, like I said, I think needs to be said more often. I don't hold any shame for the fact that that's how I use AI. I think, you know, it's okay that we're using it that way. Makes us better. It makes us more effective. Gives us it gives me sort of insights I probably wouldn't get otherwise. And quite frankly, I don't think that's anything to be ashamed of. So I think that needs to be said more often in the context of how people are using AI these days. So 100% agree. A 100% agree. Definitely. So we got a couple of questions. You know, one of the things is one of the questions was when will the 2026 playbook be available? We're probably three to four weeks away from that. You know, you'll if you answered yes to the poll that's on the screen now, you'll you'll definitely get an early preview of of the playbook. But like I said, it's probably within the next month or so. There is another question there. And Walter, you know, well as I do, we can't talk about this topic without talking about security and governance and data privacy and all those types of things. So I'm going to use one of the questions that one of our attendees posted to sort of let you shed some light on this topic itself. But, you know, this question is about metrics around security and what should be considered there. I guess, as you think about everything we've talked about today and the governance and security side of what's your top three things that you would recommend our audience when it comes to monitoring the security and the governance within your business and the use of these tools and these concepts? Thank you so much, Fred. Yeah, I think I see here it's a question from Michael Green. So the number one thing is it starts with a strategic question and the number one is here is I think you'll be able to see as well in the playbook is obviously the scope and the specific areas that you want to actually designate to an agent to be able to act on. So setting those boundaries really, those guardrails, that's where it really starts before you actually even deploy it. So identify first of all which are the spaces and the tasks that are essentially could benefit from more autonomy and from automation and try and limit them. So I think that the key thing here is as well and I think that's kind of another word for saying use cases, limit them to specific kinds of things that you can then test them out and that reduces the risk of an agent essentially going into areas that maybe you do not want go through. Number two, and that relates to that specifically, essentially the key things that makes an agent interesting, but at the same time creates the risk are three things. Number one is that it needs to have access to personal information, it needs to have access to the openly available web and it needs to be able to interact in ways that essentially are autonomous or independent. These three things put together are the ones that you want to control. So you want to limit from the point of view of data what kind of personal information or in the case of corporates, what kind of organizational information can it have, so the limits to the data. Number two is when we talk about the openly available web, I mean the internet is pretty vast, which areas do we want it to actually be able to tap into, so there's a spatial dimension here and finally there is then the question of who do I allow it or what do I allow it to interact with. Are we happy for an agent to go out and interact with anything or anyone or do we want to limit it to particular kinds of people and spaces that are out there? This is kind of like more if we use it booking things or making interactions, I would strongly advise not to interact with a client on that depending on the use case, but if we want to sort of create some limits, have to say the personal data, the spatial limitation, the personal limitation. And the final thing perhaps to add is that when we see some of these projects on like GitHub that are gathering massive starring, massive new users. The mindset that I usually like adopting for these things is to essentially see this more as an experiment and it's an experiment because it's developed by communities, people can work on them, but I advise people to say let's just use ClothBob blindly that we found on the GitHub project and bring it into organization because we saw the spiral video. The key thing still here and I want to be the person that kind of slows people down, but you do want to rely on trusted advisors and trusted partners because it helps us outsource some of the thinking ourselves and it also brings in decades of experience in doing that. Yeah, I think again, the trusted advisor, trusted partner is a key piece there when you're talking about these types of things and, you know, being able to sort of anchor it in that trusted relationship and trusted toolset versus, like you said, letting Claude Bot go loose on whatever it wants. All right. I know we've gone a little bit long. Walter, we always run out of time and not be able to talk about everything we want to talk about. But I want to give you a minute just to sort of give your crystal ball view. Obviously, this year coming into this year, the companion AI, the memory and all of that is, like you said, all of the rage. Crystal ball a year from now when we do the third edition of playbook, we get a chance to do What do you think gets added? What's next? That's an excellent question. So there's a couple of things that I know, for example, venture capitals are investing heavily in at the moment. One of those areas is something called unified memory layer and that means essentially that instead of me having a one to one conversation with the agent, it might be other people in my organization that can get access to this, so that the memory, the information it has about me as a worker is something that can be shared more broadly. Now the good thing about agents is that we don't have to argue over staffing like we have to argue maybe within my firm of like, oh I put this consultant here but I need them there and so on, but essentially that means that the spread within an organization could be significantly enlarged. If there's two trajectories that will be continuing even exponentially over the next year, it's on the one hand that I think the autonomy will really increase and that means that we can provide more precise instructions as to what these agents can do and dare I say trust them also more with fulfilling certain tasks. There is obviously always and that's something that every organization will have to figure out for themselves, there's always like the chart that basically says the more autonomous an agent goes, the more you want to understand the use case and the sensitivity of these things. So whether you're in construction or you're an architect or you're a consulting leader, you want to find that balance between the autonomy and safety and you want to do that yourself, want to have ideally sort of like a really like an honest look at what this looks like. And the second thing is thus interactivity and the interactivity is important because on the one hand this means that the companion AI, the closeness of the relationship that we develop with these tools and again, not humanizing them is something that's only going to advance. In part we see that the language models and how they're being updated, they are increasingly going towards that because it increases user engagement, but really understanding that discernment as we entrust these companions more and more with our information, with our preferences and knowing how to interact with them, that is going to be another major trend that we're going to see. So essentially to sum it up, it's going to be autonomy and interactivity. Those are the two things that I think our next playbook will include potentially even more of. All right, well I hope we're back here a year from now sort of talking about edition number three with those additional pieces. Walter, always a great conversation. Appreciate the time, the insights, the expertise. For our audience, as I mentioned earlier, you can download the slides through the resource widget. You'll be getting a notification of an on demand recording of this session that you can share with others if you'd like about twenty four hours from now. And again, the playbook, if you answered yes to the poll earlier, you can certainly get access to that playbook. If you didn't answer yes, you know, go to Deltek's website and you can learn more about getting access to that playbook as well. So again, Walter, thank you so much for the time. This was a great conversation. Thank you everybody for attending today and I wish everyone a wonderful day, evening, whatever time zone you're in today. So thanks everyone. Thank you, Brett. Thanks everyone.