How to Fund AI Without Letting Your Cloud Spend Spiral

Hello, everybody. Thank you for joining us for our live conversation today. We at Bluewave are partnering up with experts from CBTS, Expedient, and Bucher and Suter to discuss how you can fund AI at your organization without letting cloud spend spiral. And we have a great discussion ahead, but before we start, I have some housekeeping items to cover. First, we are going to aim to run about forty minutes today. This webinar is being recorded, and you will get a copy of that after the event. And lastly, there is a Q&A section in this webinar interface. So if you would like to submit questions as we go, feel free to put them in there. This is an unscripted discussion today. So if we run a little long and we don't have time to address those at the end, we will have our panelists get you an answer via email after the session. So with that, let's introduce our panel for today. First joining us from Bluewave, we have Solutions Advisor, Martin Gale. From CBTS, we have VP of Solution Sales, Jon Lloyd. From Expedient, we have Field CTO, AJ Kuftic. And missing today, he had something come up. Alright. Actually, we have him here. Thanks, Kenny, for joining us. From Bucher and Suter, we have Head of Sales, Kenny Barton. So thank you all for being here today. I will be your moderator along the way. My name is Stephanie Hamrick, and I am the Director of Demand Gen at Bluewave. So next, we'll do a brief round of introductions for those who don't know us. And for our audience, we're gonna try and stick to thirty seconds or fewer each so we can get right into our discussion. But first about Bluewave, we are a technology advisory with the mission to bring confidence and clarity to your technology decisions by partnering with IT leaders. We have experience across a wide range of technology areas from security to CX to cloud and AI and a whole host of others with both clients and advisers position across the United States. And I'm gonna hand it now briefly over to Jon to give us a short intro for CBTS. Cool. Thanks, Stephanie. Jon Lloyd, I've been here at CBTS, for going on fourteen years. We are a little bit of a unicorn because we do a lot. And so anything from complete greenfield custom app dev through, public hyperscalers, neo cloud, our own, private infrastructure as a service through networking and digital workplace. So your voice network SD WAN SASE, complete data center infrastructure VAR. And then lastly, security solutions, top down from MSSP through, you know, pro services and virtual CSO. And so what I always tell people is, you know, when they ask who's CBTS's biggest competitor? It depends on the conversation we're having. I think we're very unique that we scale, the entire IT stack to help customers reach their business outcomes. And then we embed AI into everything that we deliver. So our managed services, our SOC, all of that, we're using AI to make our people better, faster, stronger. Awesome. Thanks, Jon. And next, I'm gonna hand it over to AJ to give us a quick primer on Expedient. So hi. I'm AJ Kuftic. I'm Field CTO at Expedient. I've been here for six and a half years. We are a cloud service provider. Started all the way back in colocation as a dial up ISP and now provide everything from colocation to virtual hosting, backup, DR, edge, containers, cloud native, managed hyperscale all the way out to AI. And we're bringing cloud data and AI together because we see that as the way forward for everybody. As people start to do things with AI tooling, they need places to run those things, those AI workloads, not even just the things that are the models that people are using, but the things people make with those. So we are the VMware and Nutanix service provider of the year for 2024, and we were the back-to-back Nutanix service provider of the year, for 24 and 25. So this is something that we pride ourselves on, and we have a number of different solutions all backed as a service that help customers get to the outcome that they're looking for. Awesome. Thanks, AJ. And next, I think we're having some technical problems. This is live folks, but we're gonna see Kenny. Are you with us? Can you give us a quick intro on Bucher and Suter? I'm, Kenny Barton with, Bucher and Suter. We are based in, Switzerland. We're a Cisco contact center partner. Been a Cisco contact center partner for now twenty six years. We're the first of Cisco's contact center partners, and our focus is really helping customers to leverage Cisco's solutions, whether it's premise-based or, cloud-based solutions, leveraging their AI solutions and, their collaboration portfolio. So good to good to join. Thank you. Awesome. Glad to have you. Alright. So let's get to our agenda. Today, we're gonna have four sections that cover costs across the AI journey. First, visibility into where AI related costs start to show up. Then we'll cover how to approach funding conversations for AI initiatives. Third, what kind of governance is best practice or needed? And lastly, guardrails and accountability. How do you measure the success of your spend? So before we hop into our first section here, we are gonna set the scene a bit. I'm gonna take the slides down so we can get into our conversation here, and we're gonna run this first question lightning round style. So let's keep it all to one minute or less so we can jump into our main conversation. But I wanna help set the scene here for our audience. And the question is this, how has the FinOps conversation changed over the last twelve to eighteen months around cloud and AI? And this one, I'll hand over. Jon, I'll hand this one to you first. Sure. Thanks. So I think one of the big things with cloud and FinOps, you had clear guardrails. Right? You had a specific amount of reserved instances. You had, tagging. You had policies in place to, spend some you know, an idle resource, turn it off after a certain amount of days. You had clear understanding and visibility in guardrails. In AI, things have shifted significantly. And so I know we'll get more into that today, but when I start talking with customers, I always frame this up with the concept of tokens. Tokens are not free. People always start with, I'm gonna replace ten people, and they realize they've increased that cost with tokens because they've got agents running 24/7. There's also the impact, you know, to make it a little more relatable whether you use ChatGPT, Claude, whatever your tool of choice is. Think about in your day to day life when you upload a PowerPoint and say or you go make this PowerPoint and it comes out and you go back and say, wait a minute. No. It was misaligned. Change this. That's more tokens. And so you think about a project isn't as hit or miss where cloud, the server's there, it's not there. It's spun up, it's spun down. How you roll out AI to your organization, you can have a huge surprise because even if you're on an enterprise version with five hundred dollars a month of tokens, two people doing the same job can consume those in very different ways. AJ, I I see you shaking your shaking your head there. It looks like an agreement. What do you think? Yeah. I think it's it's very much grown. I think it's not necessarily that cloud spend controls and FinOps have gone away. They're still there. And now you as an IT leader, you get to also deal with all the AI stuff. And nothing about the cloud spend FinOps side has changed. You don't change the way, as Jon mentioned, we're spinning things up. We're spinning things down. Where's the right place to run that workload? But now you have, everybody's running out and spending twenty dollars a month on cloud. We didn't even know that was happening. Or how am I this is this is almost coming back to the early days of cloud where every single team ran out and started deploying to whatever cloud they felt was the best one for them. And the only person who knew what the company was truly spending on cloud was Mary in accounting because she was the one who had to pay all the invoices. This is where we're starting to see a shift from my view in terms of AI tooling. Everybody ran out and got their own thing, and now people are saying, okay. We need to put some arms around this from a cost standpoint. Like, are we using the right tools? Is Claude the right one for everybody? Is ChatGPT the right one for everybody? Is Gemini? Is it rolling our own things? Do we need to buy hardware? All of the FinOps around it, to Jon's point about tokens, it's absolutely right. But most people don't recognize tokens because they're buying the, per-user accounts and letting the limits handle the token usage. But when you get into enterprise usage, that's where it becomes a problem. And so we kinda have to do effectively a rerun of what we did for cloud a decade ago to make this work. Kenny, what about you? What we've seen, honestly, from the, when customers are trying to adopt AI, and they're they're optimizing the cloud going from optimizing cloud to governing AI is they don't know what they don't know. They don't they don't know how that spend is gonna spin out of control if they're not careful. But, I guess beginning and with the end in mind with with governing AI is understanding what you're trying to accomplish, understanding your outcomes, what problem are you trying to solve with leveraging AI, and and then and then we'll we'll get into that more as we go forward, but, really understanding the the outcomes. If you don't understand the outcomes you're trying to solve for, there's there's gonna be a a risk of cost spinning out of control because it is a new a new thing and I'll say the AI gets you where you gets you way faster towards the end state, but that end state might not be the right direction. And so there's some there's some things we really need to help customers understand better. And lastly, what about you, Martin? Well, what I find myself involved with customer engagement, is if they're not embracing FinOps as a practice, as an area of attack to find revenue to fund the projects, to do all the things that the rest of the team was talking about. And it's something we've been doing in IT for a long time. And many companies that are cloud native or whether they're not cloud native or not, right, there's always opportunities to do cost optimization. And so now that becomes a mechanism, so they're a lot more interested. And we bring every one of these providers to those engagements that can help them because you can free up some cash that you can move over because you probably haven't had time to budget for those AI projects that you're being asked to deliver right now. So that's how I see a change in how the marketplace is talking to me. Awesome. And a few of you just touched on something, which is, you know, AJ, you just said, Mary in accounting is the only person who knows what we're spending. And I think that really speaks to the visibility problem that's at play. So let's go ahead and get into our first visibility, our first section here, which is visibility. And the question I'm gonna hand over to you guys, and I'll hand this one to you, AJ, first, so you're ready. But the first question is, what are the hidden AI related costs organizations fail to account for early on? Yeah. I think right now, the big one is user costs against usage. So I rolled out this tool to everybody. It cost me an amount of money to run that tool. Am I getting actual value out of it? So that's the first one. The other one, funny enough, is compute. And not in the, I need big set of hosts to run my local models. It's now that people are five coding various apps and everybody thinks the SaaS market is dying, it's not. That's a different discussion. But we see this shift of, hey. Everybody's building these apps. They have to run them somewhere. And maybe that's the compute that they've got spun up in AWS. But for the most part, the people who are going through and building apps for themselves, they're not the developers. They're never in the world of how do I deploy this through our pipelines to get it out to AWS or to Azure using our corporate secrets underneath our billing structures. They're just going out and deploying stuff. There's a number of smaller providers right now. Not not Cloudflare is not small, but they're not the AWS and the Azure's of the world where you have Cloudflare, Fly. Io, Vercel, and a number of other providers who are like, yeah. We'll take your credit card. You can just run this container here publicly on the Internet. And now I'm losing the visibility of where all these workloads are running and what they're connecting to. So this cost doesn't really show up in the normal bills, and this is where you kinda have to have that the guidance on how do I go to I built this app. Claude made this app. ChatGPT. Codex made this app. But now what do I do with it? How do I hand that out to end users? How do I connect it to things internally without putting things at risk? And that level of visibility isn't really in place at a lot of organizations because people are just making stuff up as they go. So there does need to be a level of governance or process put into place to say, okay. You made this thing. Great. Good job. Now how do we release this? And you almost turn everybody into software developers in that sort of space. And that's scary for a lot of organizations because most people aren't great at their own software development, let alone how handing that process to literally everybody in the company. So this is where you need to have a level of, you know, understanding that this is that next step that people are going to and start to have those conversations internally to say, what do we how do we want that to work going forward? Kenny, what do you think? Well, costs the the, hidden costs oftentimes in what what we've seen is that customers don't understand and because a lot of the times we're working with contact centers and leveraging AI is either new or it's something that they are are there's new features and functionalities and ways to use it. And because of tokens and because of usage based costs, oftentimes customers are unaware of what that is going to look like as they start to adopt. And so those costs at times where maybe they have in in their mind, oh, well, we're not quite sure. Let's kind of make an assumption it's gonna be this, But maybe they are adopting it way more than they they had anticipated, and those costs start to really run up. And although it might be solving the the problems that they were trying to initially, utilize AI for, those costs can can get get a little bit onerous for for customers if they're not understanding exactly what they're trying to accomplish, have a goal in mind, and then and then build it and see that it's, you know, the usages of the solution is going to is going to be within reason. And so that's a big challenge that we've seen with customers that, have, I guess, hidden costs or the usage costs that that can start to spin out of control for customers. And, Jon, over on the CBTS side, what have you seen? Yeah. Think AJ mentioned it around governance, and I look at that really in two ways. The first is when we open talking about cloud spend, as much as maybe Mary was the only one seeing it, at least Mary was seeing it all. There was one person getting to see it all. And now we've got this challenge where one, you are seeing cloud spend go up because this department released an agent that's running twenty four seven, and it's pulling resources in your cloud, and your cloud was meant to run Monday to Friday, eight to five, and then be elastic and turn off over the weekend and save you money. So one, we've got twenty four hour agents. The cloud team's not aware of it. They got spun up. You're not spinning up cloud resources. You're simply pointing a user the same as I would. So you've got this problem where there's not a centralized cost center right now for AI. Mary's seeing the cloud cost go up, but you've got four different business units all rolling out AI projects independently. There's not even if there's a center of excellence around standards for AI, typically, there isn't a centralized cost center that's seeing it. And then the second component of that is how do you classify in governance? And this is we're starting to see more and more creep. AJ mentioned the SaaS space. We're seeing this to creep in the enterprise of your SaaS providers are coming to you and saying, we've got these new features. We're embedding AI. So whose budget is that? Is that the AI budget? You know, if we add on this AI module in ServiceNow or Logic Monitor, pick whatever your SaaS tool that helps run your environment, Should that be hitting an AI bucket? Should it be hitting your network observability bucket? Should it be hitting and so again, I think comes down to governance with visibility. How are you defining AI? Is it LLMs? Is it agentic? Is it app developers? Define AI, decide where it lives, put guardrails and constraints and governance around it. And right now, customers don't have that, and the people don't have there's not a single source of visibility. Yeah. I see you nodding really emphatically, AJ. What do you think? Yeah. I think from, like, the feature, the one I always pick on is Zoom because Zoom has their AI companion. And we get asked a lot because we have an AI tool that can do summaries and all that sort of fun stuff. And when we have that conversation, like, why would I use that over your tool? And I'm like, because it's native. It's just in there. Right? It's it's a good to be the king situation. You click a checkbox and it automatically transcribes it, figures out all the action items, and sends you an email when the meeting's done versus, hey. I'm going to manually tell it to transcribe, then I download the transcript, then I run it through insert LLM here to then generate an email and then send that out. And there's some automation you can build or you tick the box and it's done. And you don't have to have god level rights in order to do that inside of Zoom when you wanna do that across your company. So there's a number of those sorts of things where in my mind, John, that that's the sort of thing where it falls underneath the like, for Zoom, it would be something like a workplace group or who's running end user computing because that is an end user tool. It's a function of Zoom versus saying, hey, AI team. Can we get budget to turn on this feature of this thing that you don't necessarily know or care about? So I think there's some I think, terms of the AI features versus AI tools, that's where I think some of that cost split goes in. That's a great point. Yeah. Definitely. How about you, Martin? Well, I think what I'm seeing now is for organizations that are just beginning that journey. And it reminds me of a conversation I had with a CIO and the CISO for an organization that really hadn't deployed it. They probably dabbled a little bit in some chat GPT or Claude type applications. And they said, "Don't you just have a tool that you lay down and it does it all?" And I was like, "Well, no, not really." And I think so people's understanding of all those components. Let's say, for example, they have a bunch of subordinate companies or done a bunch of M & A, and they have ERP systems, and they want to get that data from point A to point B. There's an ETL or a way to get that data over to what there's I hear the term data lake. I'm going to put it in my data lake. What is a data lake? Does it just create itself? Is it a file repository? Is it a SQL database? And then the infrastructure, again, that you need to do all that. So there could be some transport costs, some software components. There may be organizations that do some or multiple of those pieces. And usually when I'm going down that journey and that conversation, people start saying, Well, how do I do my AI agents? I'm like, Well, that's a whole different discussion, right? And it's like, we have to do a discovery, find out where your friction points are. So there's a whole bunch of parts, and that's not even talking about the governance and the security. The security team is going to be concerned about, do I have the right MSSP in place that has the ability to deploy software that can identify shadow AI and things of that nature, right, and deploy the right kind of policies to prevent data loss and whatnot. So again, I was backing up a little bit from the customers I talked to. Generally, a lot of them, when I used to talk about where are my switches, where are my firewalls go, where do my servers and my compute go, right? Now we're talking about how do I deploy it in an overall architecture, And there's a number of costs that they have to think about. And again, that usually leads me back to the conversation I talked about earlier. Are you doing thin ops? Can you save some money somewhere else to help pay for the things that you need to do right now? And I think I'm gonna hop in here because I think that where this conversation is going leads really well into our next section here, which is funding. And I think what you were just saying, Martin, where people are asking the question, isn't there an AI that just does everything? Well, if there was an AI that just does everything, maybe your budget would be a little larger. But instead, IT leaders are needing to go to leadership and say, I need funding for these AI initiatives. And it can't just be an open-ended, "I wanna fund everything". Right? They need to think about their priorities and what they're doing. And I'm gonna hand this over probably to Kenny first for our first question in this section, which is how can IT leaders go to the board, go to the c suite, and avoid treating these asks to fund AI initiatives like an experimentation budget, like a never ending bucket of money? How can how can they approach those conversations? That's a that's a really good question and I I kind of equate it to all all companies go through periods of time where there's a swell in work and things that need to be done and maybe not enough people. And often used term is, well throw more warm bodies at it. I was in the Marine Corps so that's a common theme or common terms. Just throw warm bodies at it and it will solve it. In hiring practices, we don't do that. We set up a role and responsibility, a vision of what that role and responsibility is going to be doing to then solve those problems. But if we just throw warm bodies at it without qualified people, the qualified people that are already in place will be stretched more and more thin and it will pull them away from being able to do the good work. So we wanna be, you know, the way I see this and the way that our company has been seeing this within organizations is understand exactly what the role and responsibility of the AI tool is going to be. Identify how you're going to execute on this. What are the outcomes? Build out the solution and the and the and the path to that outcome with AI, and you'll have and then and then report on it properly. So understanding your business, understanding the outcomes that are needing to be met, and then align the AI within the parameters of solving the problem. And then, of course, governance should be a top priority in this scenario. Otherwise, AI is gonna be the new warm body of our generation. Throw more AI at it and it will solve it. It won't. It will cost more money and it will stretch your people thin and the execution will not be there and and and it will end up being like just hiring a bunch of people without qualifications. You'll you'll have a lot of of of over overrun of of of costs. Jon, what about you? Yeah. So we've talked about, know, a center of excellence, visibility, understanding, guardrails, right? All of the foundational stuff. Let's say you have that. We do a lot and we talked about this earlier, lot of these we call self funded projects. It might cost you a million dollars, but there here's the real ROI. Now here's the difference between our million dollar projects and, you know, some acronym consulting firms. Right? It's a couple of letters. They're gonna come in and go to a million dollars and it's six months. And at the end of it, you're gonna get a PowerPoint that hopefully they've changed the company logo from the last time they gave that same PowerPoint out. Our million-dollar, six-month project, we're gonna do in a modular fashion. So the first phase might be a two-day workshop for fifteen thousand dollars to even understand, is it worth going down? Is it worth proceeding? Then we're gonna move into a proof of concept, and we're gonna define what is a proof of concept, time bound, what's the value you're gonna get? Then we move it into a pie. And we give customers along this way. So ours might be seven, you know, epics, phases, whatever you wanna call it, stories that are you can walk away at any point. And sometimes the best value in AI is spending thirty grand to realize there's no value in that AI project and not going further. We see customers get stuck in this. It's almost working indefinitely. So let's define what is almost working. What is the time? If it's not producing this yet, do we walk away after two weeks, three weeks? And so I think that's a really key component of defining what is success, defining what's almost working and then also going with and building part of our modular approach is building a real business case, real costs, real expenses, real projections that will allow you to make that decision. So I think for us anyway, it's understanding along the way, the best project is the one that you didn't spend a million dollars on to realize it was gonna fail. And so much of our competitors in that space wanna position that million dollar PowerPoint. Yeah. I think it's so powerful to be able to say sometimes you learn the lesson and you walk away and that's the lesson right there. AJ, what about you? Yeah. I think there's a if you don't want it to be an open ended budget line, which in our experience, it's not anymore. It was a little bit two years ago, but nobody's just throwing money at the wall now. We're we're doing an accelerated we're doing a speedrun effectively of it. I keep coming back to it, but it feels like a speedrun of what cloud the cloud journey over the last fifteen years has been. It's, hey, we're gonna just go out to the cloud where everything's gonna go out to the cloud. Oh, hold on. That expense is that's expensive. Hold up. Let's pause there. Let's maybe take a step back, and let's think about why. And it's the same thing. If you don't have a use case, it's a bad idea. So if you're going to go out this is why when we come in, a lot of it is starting with stabilize. Hey. You've got a million different AI tools. How can we put help you put the arms around it from a security observability standpoint and still not remove the feature set that your customers are using today, your customers being your end users inside the organization. And then, hey. Let's figure out what are the big use cases. We need to connect that data together. Right? You probably have a whole bunch of stuff in SharePoint that you want to use. So how do we connect the data in SharePoint? How dirty is the data in SharePoint? Spoiler, it's very dirty. It's always dirty. It's a dump it's been a dump forever, but we're gonna try and figure out what are the good pieces that are in there and put them into the right place. And then you start to modernize in terms of the agentic flows. Hey. What is this thing that you struggle with? We actually did this internally, and we get notifications from our connectivity vendors. We have fourteen data centers across the country, so we have a lot of connectivity coming into our data centers. We get emails from them all the time saying they're working on some connection that comes into our data center. And they basically ended up being about a thousand man hours a month of going through, figuring out what's in all those emails, sending out notifications to our customers saying, hey. There's a change that's happening on that. Just be aware of that. We don't have anything to do with it. They're doing that change, but just so you're aware. And then we had somebody one of the people on that team said, I wanna automate this. And so they brought in we have an agent platform. We brought in we put an agent in place that watches this mailbox, looks for a particular vendor. We started with one. We didn't try to figure out all of them. We said, okay. This is the one. We'll just pick it was let's just pick on Zayo because they're a big connectivity provider across the country, but they have a fairly standard form that they send out. So we go, okay. Let's look at Zayo. And now we ran through all of the Zayo, you know, request and notifications that came in, parse them out, automatically send out things to our clients. We save nine hundred and sixty hours a month of work. Wow. That is a huge savings. And now what do I do with that? I take those five people that were doing those, and I can put them on other things. You don't let them go. You put them on other things because the the laundry list that you have internally doesn't shorten. It's never complete. So if you can free up those hands to do that, that's a use case that then provides an immediate value. And I can say, hey. Remember what we did here? Now let's do it with this other notification, or let's do this with this other use case that we have because it's a pain in the butt for us. So there's ways to do like, if you go if you start at the use case level, you will have far more success at getting funding, especially when you can really, like, box it into this is specifically what we wanna spend the money on. Martin, I'm gonna hand over this over to you. AJ just said a lot of dirty data in SharePoint. I know you love talking about that, and that's not confusing. Well, I I wanna I wanna talk about two different use cases where, you know, I personally have gotten involved with with clients to help them get funding. One was internally, and it really plays exactly on what AJ was saying, is creating a use case within. And then it's, you can't prove something is working if you can't measure it, and then you can use that measurement to improve it, right? So I was talking with a customer and they had a board, and their board was like, how do we even know this AI is going to work, right? And so I think deploying a tool in a proof of concept with a way to measure the use cases to show quantifiable results based on users. And again, you can tie that to some outcomes and some things like that, and that's part of your discovery. Where we lean in is try to help customers understand where their friction points are, bring a tool like Expedia was talking about with their tool set, and show them how to measure and present that result back. And that way, if it's a board of directors and they're saying, well, how do I even know my people are using this for productivity? Well, here is your proof. Here are these people, they're tokens, this is how they're using it. And maybe there's two or three people or a subset where you can say, these people aren't, and their productivity is less. So that's one way internally, I think that I've had some success with a CIO getting internal funding. The other way is depending on their choice of what their AI toolset is, right? So let's say, for example, Copilot inside a Microsoft ecosystem. There are ways, and I'm pretty sure every single one of these providers here can help you, is to get funding from Microsoft. Why not have Microsoft pay for some if you're adopting and you're planning on spending a lot of money inside their ecosystem, is they can help you get funding from those providers to help fund that project, get it kicked off the ground, help with deployment, and those kinds. So those are two different really relevant ways that I think you should be considering, like how can I get some help if I'm trying to fund something and get it kicked off the ground because I have some people who are above me who are signing the checks that maybe are not quite sold on everything we're doing? All right. I am going to take us here into our third section for the day, which is governance and how you should be thinking about it on your AI journey. And as we've kinda seen thus far, everything really ties back to governance. It's impossible to have any of these conversations about any of these other topics, visibility, funding, without talking about the governance piece. It's so fundamental. It's so important. This one, I'm gonna hand over to you, John, first, and the question is, at what point does AI experimentation require formal governance? I would say as soon as you start. Right? I mean, here's the the IT owns the infrastructure, that's on prem or cloud. Right? Finance owns the budget, and now you've got a business unit owning the outcome. Nobody knows that owns the cost. Nobody understands the full visibility and that's why it's the importance of governance. Your sales team never was gonna go spin up an EC2 instance, Right? They didn't have credentials. They didn't have access. They didn't understand it. Right? They might go buy a SaaS tool off the shelf and you had shadow IT. But I mean, from a a standpoint of increasing your bill without consent, they didn't have those capabilities. Every organization does now. And that's what's creating this challenge is the business unit owns a business outcome, can go and create these tools. But who owns a budget? Well, everybody at first, when Microsoft when AI was copilot for thirty dollars a month, it was that's IT's problem. IT's gotta have the budget for it, which didn't make any sense. So I think as soon as you start have you should have a governance in place now because there's not an organization that doesn't have some level of AI even if it's, you know, partner delivered onboard AI. It exists today. The other thing when I talk with customers is it doesn't need to be some heavy framework today out of the gate. Just start with a a center of excellence for use case, some tollgate for approval, and then start with, you know, an internal use policy that you can at least hold employees accountable for throwing your data up into chat GPT that you can come back to. And, you know, just by starting that conversation will be a huge first step. It can be lightweight governance, but understand your organization's using it today. You need to have some type of policy in place, and you can evolve it over time so that you don't, you know, smother your AI use because you do want as AJ was saying, you want your employees to use it. You want we have right now a big campaign around AI literacy inside of our organization to use it correctly, to use it the right way, to make yourself a five x employee. We want that, but governance doesn't need to be, you know, some heavy thing, but it's got to exist Yeah. That idea of starting lightweight and then evolving over time. AJ, you mentioned something very similar to that when we were having our planning talks for this session. Do you wanna expand on that? Yeah. I think you need to put into place the you have to start with governance from the beginning. Like, before you roll out anything, how are you going to govern this? And not just in the framework way, but how are you keeping track of who's using what? Because you can have somebody who uses a billion tokens in a week, and you go, hey. What where did you who did who why is our bill so high? And you have no visibility into that. With the right tooling, you can do that. And that's where when you get into this is why I I find it very interesting how many companies are saying, well, we'll just pay for Claude individually because Claude Enterprise is a seat plus token thing, and we don't wanna think about tokens. We're just gonna hand everybody the, like, very expensive Claude accounts, and we'll just let them all happen individually. You have zero visibility. None. You have no idea how much anybody's using. You have no control over what's going out, and it it it's there. When you get into the enterprise tiers, they give you some of the visibility, but not all of it. And so what we actually built from the start with our toolset was here's full visibility. What prompts, what responses, what models, how many tokens, which users, and when they all happen so that you have that level of observability. Observability is a key piece of the governance because I can, you know, put all the rules in place I want to, but if I can see that nobody's following them, I haven't I can't see that anybody's following them. I don't really have a choice there. I think the other side of this is John, you mentioned the whole, you know, using the tools and five x employees. There's also a flip side to this. And I can tell you with great certainty in my personal life that I do not have anybody who loves AI. None. They don't want it. They don't want the tools. They don't like Copilot being shoved in their faces. They don't like saying you have to use the AI tools or else. They don't like that. So also consider the employees who are good at what they do, but they don't want to use this tooling or they feel very, you know, confident about their opinions on that. Have that conversation and understand, like, does your user base want that? I can tell you the younger generations actually don't. They would like to be able to come into the workforce and they see this as a a threat to them. So show them that it either a, is not a threat or b, it is, hey. Here's how this tool is going to make you coming on board easier so you can do the thing that we're paying you to do. I think those are some of the other pieces of the governance and and onboarding to AI that need to be understood because it's not the panacea. Everybody loves this tool, we're just going to do it. I think there's a lot of people who are using it begrudgingly because they feel like they have to. Otherwise, they're gonna get fired. So also point that out as a we don't wanna make that a condition of, you know, your performance. I think that's a really great point. And I'm gonna hand it over here to Kenny. You mentioned it in kind of your answer to our very first question, how important how important governance is. What do you what do you think about I think the the importance of, you know, in in in our focus on the contact center environment, when, what would be the cost to an organization if an AI was doing something that was off brand or created a problem? What would be the cost to the business and organization if something was to go terribly wrong? Which as we know AI, if not fully controlled or fully adopted and putting guardrails around it can hallucinate or or take us in different directions or directions that we are not we're we're not wanting to have, have it go. So in in that scenario, the importance of it is just imagine how much it would cost if something went terribly wrong, and that number that cost could be astronomical. So, with that in mind, really, as you begin you know, companies I I I look in in my, initial preparation for this at seventy six percent of organizations now have an AI chief a chief AI officer. It's a and that was twenty five percent in in twenty twenty five. So those the as companies are seeing the importance of this, we need somebody that is overseeing and and providing value for those reasons. Not only costs, but also hidden costs that are potentially risks that can that can pop up, that can cost your company or your brand, tremendous value. Something goes sideways and it gets reported, goes viral, all of a sudden, business is is, is on the fritz because of of of some of these challenges. So absolutely, critical for for businesses who are leveraging this resource or these resources in all their forms to, have a a governing body that, to ensure that those types of things are are less of a risk. Yeah. I think all three of you bring up, like, very different but equally important ways to think about it that we're not hearing across the industry as often. You know, AJ, to your point of, like, are are you making your employees happy? That's an important part of governing technology. Technology enables our employees to be happier, to work better. And, also, to your points, John and Kenny, you know, keeping control of this is so important. I'm gonna hand this one over to you, Martin, to finish up this section, and then we'll hop into the next. Yeah. I think what what I'm seeing too now in the marketplace is is most companies are doing annual pen tests. Right? And and the evolution of that is changing now, and they're starting to ask the cybersecurity companies that are doing it, like, can you do AI penetration tests? And that's twofold. It's not just using AI as a cybersecurity company to do the testing, but it's testing AI that I may have deployed. For example, maybe I deployed a bot on my public website that tied into an LLM on the backside, right? It's like, can that be penetrated? And those conversations then lead into further behind the firewall, say, how far can I go? What can I get to? But then what is your overall governance strategy towards AI? Because maybe your development team who's motivated by their productivity of deploying these different things, but not necessarily thinking about it from a governance perspective. So I would suggest if you're not thinking about that and you do annual pen tests, right, you should be thinking about that. And even we're seeing the evolution of those kinds of tests actually, in some people's opinion, say you're going to displace vulnerability scanning and become continuous pen tests, both in internal and external with regards to your firewall. So that's something that I think everybody should be thinking about in their strategy for AI. All right. We have one more section to go here, and we're at the forty ish minute mark. So we're gonna try and do this one lightning round style so we can, hear from everybody, but we're gonna transition into this section about guardrails. And this is not necessarily the same guardrails that we've all just been talking about on the governance side. This is around how you measure and define value, kind of back to John's point of if the value is not there, you walk away. So how do you think about these guardrails? And our questions here is just how do you determine if you're seeing value, and how long do you wait? So let's do this lightning round style. I'm gonna I'm out of order on my round robin here, but I'm gonna hand this one to AJ first. Sure. The first part of generating the value is you'll know pretty quickly whether you're getting a value or not. That's actually one of the beautiful parts is this agent that we spun up to do this thing, it either worked really well or it didn't work. You'll get that fairly quickly, but I would not look at it as a pure, we implemented this agent and it saved us x dollars. It's going to save you time. It's that time is obviously money. But if we consider it as pure dollars, there's not going to be well, we implemented this agent, and now the bottom line is higher. It might be different. It might have avoided some hires that you thought you were going to have, but that doesn't directly impact bottom line. I think the other side of this is from a, you know, understanding the value proposition is what value do you think you have in your head? Because there's other times where you say, hey. This actually turned this thing that we deployed now, hey. I can actually do that for, like, six other things. So you don't have to have five other projects or six other projects that go along with that. It's the same flow. It's just different data sources for different use cases, but it's all generally doing the same thing. We're processing this thing and sending out some notification. I can use that for, like, nine different things inside of our organization, but that flow, that agent structure that we invested in upfront, now I can build that out from there. So I think the value proposition needs to be, is there a time savings? Is there a people being freed up? You know, how many how many things are you now able to do? Are you providing a better communication experience for your customers? Those are all the things that are there beyond just, hey. Here's this bottom line revenue number going up. How about you, Kenny? I can just use an an example. Like, the the companies that we've worked with that have had the most success implementing AI were companies that knew very, very clearly where they were at the time of the start of of leveraging these different solutions. I have a a global financial company. I think I have fifteen hundred agents contact center agents. They had very clear reporting structure. They knew exactly what how many calls were being handled, how what their, first call resolution was, what their, average handle time was, how many calls were transferred. They had a very clear picture. When they set out to to leverage AI, they knew this is what we're trying to solve for. Our our average handle time is ten minutes or our first call resolution is only forty percent. We need it to be seventy five or eighty percent. We want to deflect some of these calls that are repeat calls. We wanted to be able to, deflect those calls, to self-service AI. And and what was really cool over the years that I've worked we've worked with them is to see their reporting change and have clear direction. My last call that I had with this customer, I said, I wanna understand how how this is working for you. Tell me where you guys are at in your reporting. How many calls are now deflected? And they provided clear value that they they saw. Fifty percent of their calls were now being handled by self-service AI agents, which was a dramatic for a company of that size globally, fifteen hundred. See see the the clear value that they saw in their customer experience and also the reduction of the repeat calls, transfers of calls, all of which required humans at the time and provided the customer not a good experience, that changed. So I would I would I would say that the customers that fully understand their business have a clear clear picture of where they are and what they're trying to accomplish and the improvements that they're trying to make in very specific areas. If they have that use case of we're trying to improve this because this is too high or this is our CSAT scores aren't as as good as they should be because of these things. That's where we've seen the most success and the dramatic improvement that, adopting AI is, the most successful. Understanding your business at the outcome or at the outcome at the onset of the of the, adoption. How about you, John? So I can go on for hours on this specific topic, but I won't in our interest of time. I'll focus on one thing. AI never is going to drive value in isolation. Where we see customers go, this AI project is gonna cost five hundred thousand dollars They get approval and then they have to go back two years, twelve months, two years later to the business to go, well, the AI thing costs five hundred thousand. What we didn't realize was if you have a broken process that your human employees are following, that same broken process being followed by AI is still an issue. So if you have bad data, disjointed data, applications not talking to one another. And so I think the big thing when we talk about guardrails and measuring value is being honest with the entire estate, looking at everything. AI is going to do this function, but in order for it to do that function effectively, what are the other elements? Where is it gonna live? What's it gonna connect to? What does it need access to? Because I think there's a ton of cost that ends up creeping up. And then, again, back to the governance, well, that was IT's infrastructure cost. It wasn't part of the AI presenting that entire unified business case because not once have I seen somebody go, we're gonna make this agent and it just runs and everything's perfect and it doesn't need middleware or integration to something else or increased infrastructure costs. There are other elements and I think a lot of organizations don't look that. Yeah, that's a great point. Martin, round us out here. Yeah, I think the biggest guardrail or value I think that I would suggest is to not try to do it by yourself, right? Because there is a lot of excellent providers, including the three we have today, right? And we have access to. And I think the biggest value that I see is being able to take those decisions and we run into CIOs, CFOs, and they're struggling with not just costs, but like, how do I I don't even know where to start with this project. And you can engage some of these providers, but they don't know who they are, they don't know how they're performing, what the cost should be. And so that's where we fit in in the marketplace. We know all of these gentlemen and their organizations and what are good fits for them. We align ourselves with the customers, help them understand. Everyone has talked about knowing what their outcomes are. To be quite honest with you, many people don't know what outcomes they want. They just know they want AI, right? And so to help you take those decisions, which could take you months or years down to minutes or days, right? Engage someone who can help you. It's okay to say, I don't know, because not all of us do. This is moving far faster than I think than what we saw with the cloud initiative when everybody jumped into the cloud. AI is accelerating AI, right? So that would be my biggest piece of advice to the audience is don't be afraid to ask for help, because there is help out there. And it's a journey. And like any other project, the requirements will probably change as you go through it. And and rely on the people who have been there and done that just like all three of these gentlemen have done. Awesome. That is such a great way to round us out here. Took the words right out of my mouth, Martin. At Bluewave here, we help connect you with solutions like CBTS, Expedia, and Booker and Souter after we help you define what those outcomes are with a vendor neutral lens. So if you need help with figuring out any of the things that we discussed today, please reach out. We can help you and then connect you with a solution provider who can help you get where you need to go. So with that, everybody, thank you for joining us. John, AJ, Kenny, and Martin, thank you for sharing your expertise. So many good pieces of knowledge today, and I wish we could just keep talking forever because I know there's a lot of things that we could expand on and keep getting into, but we are gonna let people go here today. There's a QR code if you wanna view some content. We are going to send an email with the recording for today's session, as well as some resources from all of our partners here so you can learn more about what they do and what their thoughts are on some of these topics, some of those details we didn't get to. And we are a few minutes over here, so we don't have time to get to the Q and A. But if you did submit a question in there, we will get it to our panel here and we will have them get you an answer. So with that, thank you for joining us. Thank you if you've stayed on these extra couple of minutes. We hope to see you on the next one. Thanks all. Thanks.