Transcript
I'm Rick Vanover. The Rickertron and I have really called upon one of the smartest and sharpest folks here at Veeam to join me here today. Ian Engelbrecht, welcome to Tech Bytes. Thanks, Rick. We're going to have an awesome session today. Hopefully, we're going to dive deeply into what we can do with the powerful AI that we've added to the product. That's right. We are going to talk about Veeam Intelligence and we've got an action-packed show. But first of all, I need to know, where are you tuning in from? True story, we're live is live. Here I am. I got three people already checking in. Irina, I got your slides for you. Mark, Ella, Marco, got some fans already showing up. So we're going to jump into Veeam Intelligence. I want to give a shout out. First of all, I've got producers on the line from Texas, from Ohio, from Virginia, and you're joining from South Africa, Ian. So we've got a global situation going on, and we're going to jump right into it because we've got so much content to cover. Let's start with introducing Veeam Intelligence to the audience. This is a technology that we brought to market in 2023 as basically a chatbot really powered off of product documentation. But holy buckets, Ian, it's really come a long way to where it is today in the products. Yeah, 100 percent. I think to your point, it started off as a chatbot. It was just taking the huge amount of technical write-up data that we have on the help center, and basically answering customers' questions, or the user's question. It was nicely fit into the UI inside of Veeam One and VBR. But in this latest release, adding actions takes it a step further, becomes a co-pilot to the backup admin, right? Absolutely. I think what's really important to think about here is, you mentioned a few products, and it's truly going across the portfolio of products here at Veeam. Veeam Intelligence, high-level, what we're going to talk about today, things like job session analysis, proactive integration and support, as well as capacity advisory. Let's see if we can get to all of it here today. I don't know, but we got to first start off and introduce one of the big milestones that really have, I think this came in either later in 12 or in 13, but this notion of basic and advanced mode for Veeam Intelligence. Because let me be super transparent, everyone's on their own journey of AI. I'm convinced that everyone can benefit from operational excellence. Then there's a choice. Do you want it with the bottom advanced mode to really assess what's going on locally in your environment, including calling things to happen? So it's a little bit more directive and authoritative in what it can do like retry jobs, automate fixes, and stuff like that. But when we think about where this sits in how people use these different modes, Ian, I think the tinkerers amongst us get really excited for advanced mode. Yeah, 100 percent. I mean, the basic mode is going to purely just speak to Veeam documentation. It's not going to take any of your data and feed it into the model. But if you really want the power of AI, you need to give it context. To give it context, you put it in advanced mode, it has access to tools, it's able to pull data out of your environment. Some logs, some session history, some stats, some reporting metrics, and then build out a very cool response, and then feed you all that information. So with that additional context, you get very accurate responses rather than just a response based on some technical documentation. It's actually a response based on your information that you're seeing live inside of the Veeam product. Yeah, and I think it's super important when you mentioned some of that reporting information. Because through this generative AI technology, and Veeam Intelligence is generative AI technology with actionable things that come from it, but you can build dynamic custom reports. We've had the ability with custom reports, for example, in Veeam One, but now it's as easy as prompting. I think what's really super important to understand is that folks, first of all, need to turn it on. Across the portfolio where Veeam Intelligence is implemented, we have these control elements, where basic mode and advanced mode are selected. This particular screenshot is from Veeam One, and then I'll counter with Veeam Backup and Replication, for example. There is a option to even take full administrative control to actually do things. So I think it's super important, Ian, to illustrate that where folks are on their journey, setting up how Veeam Intelligence will operate is a very important first step. Yeah, I agree. It actually goes beyond just the advanced mode and adding additional context. So if you think about the advanced mode, it was using an API to get your metadata that's inside of the Veeam product. In the latest release, we actually added put or post API tools, so it has the ability to actually do actions which we're going to show you a little bit later on. Exactly. I'm going to pick up from Michael's question here. I was already trying to type an answer, but I'm just going to go into it. So Michael, we've got a link going to be pushed out here to you in a second. I want you to be properly informed before you start prompting away. But Michael, stay tuned. His question was, do you have some examples of the context that I can use conceptually? So stay tuned, Michael. Thanks for tuning in. By the way, if you have a question, drop it in. Myself, Ian, and the team, we will get through it and trust me, there's not a question that Ian can't handle. So you have the right show to bring the questions to it. So let's start, first of all, with some main concepts and architecture in the AI agents. We'll walk through a sequencing here in a bit. But the private hosted Azure OpenAI LLM that powers Veeam Intelligence, some of those external learnings, that's on Veeam up in the top right. But walk us through some of these other elements that really power Veeam Intelligence for the market, for the products that people are using. Yes. So first of all, we can't just use a single agent to do everything. We can have specific agents and they have specific tasks. We call this an agent swarm. So we have these specific roles that will handle different tasks based on the prompt that you give the co-pilot or the VI inside of the Veeam console. From that, we're going to either in advanced mode, we're going to fetch data using a tool. So we need more context. If you're reporting a backup job failure, we're going to go and search for that backup job. We're going to grab the latest session history about that job. We're going to attach that to the prompt and send it up into our agent swarm. It's going to be allocated to the correct agent, the agent that actually does log parsing to understand that, and then pass around to the backup admin agent, and then it will send back down a response. But before it even gets there, it's going to use our trusted knowledge library. So we've trained the model and we've attached the knowledge base to the models of all of our technical documentation, which is extensive across Help Center, across KB articles that we post continuously, and then match that against what you've asked. So it's giving you context with your data, with our technical data, and then responding back to you. Yeah, that's pretty much the flow. Yeah, and the best part is about this, Ian, you mentioned agent swarm, like from an architectural standpoint of what our customers need to do, it's literally just turn it on and ensure that they have outbound connectivity for this. None of this actually has to be built in your Veeam environment, so that from a accessibility standpoint, it's really easy. So let's take a look and walk through the sequencing of how a actual flow happens. So this does have some builds here in advanced mode. Let's start with the first thing. I ask a question in Veeam Intelligence, walk us through the first step and then give me the clues and I'll click us through the rest of the way. Yeah, so I'm going to ask a question. So I open the Veeam console web UI or the thick client console. I find the little tab for the AI assistant, ask it a question. What do my jobs look like for today? And immediately it's going to reroute that to the correct agent. So it's going to try and understand your context. In this specific case, it's running inside of Veeam One. So we asked it the question, what is my latest SLA across my backup environment? It then goes and finds the latest information inside of Veeam One. So it calls the correct API to fetch the correct data inside of your Veeam One, takes that, passes that back up to the agent swarm. The correct agent is assigned to the task, then searches through our custom knowledge base to see if there's any specific additional detail it needs in order to explain the response. And then packages that nicely and sends that back down to the user. So it could sound complicated, but the flow works perfectly. Again, to Greg's point, I think the most important part is you're not burning tokens. We're running the LLM inside of a private environment powered by OpenAI. And you're using all of the tokens that are sitting inside of our private AI. One thing I'll add to this, Ian, is that this part of it, like I said, it's simply turning it on for the clients and then using it. When we have big milestone releases like 13.1, introduced the deep malware analysis and the backup admin agents. But like this part in the green box, we tune it daily. Now, I'm saying that, so first of all, just to get expectations right, your experience may change in the form of an improvement if you come back and use it tomorrow or even weeks down the road. And I can give you evidence through like the 12, version 12 era, we would make changes to let it do new things dynamically without even having to do product updates and product installations and patching, etc. So this is definitely something that our teams are working very hard on. And I think they're doing a fantastic job. Shout out Jorge, Pavel, so many other folks. But sometimes people want to do a little bit more. And this one, I'm going to need adult supervision to walk through this one. And this one's a little bit above my level. But walk us through kind of some next level stuff with model context protocol. Yeah, so to your point, maybe you want to use your own AI, or maybe you don't like the response coming back from our model, or you want a specific domain specific model that you want to use, or you're hosting or inferencing a model already in your infrastructure. Then you use MCP, which is basically the USB-C of the AI infrastructure environment. It basically allows you to plug in your own model into Veeam's, Veeam intelligence platform inside of VBR or Veeam One. And you can then relay those requests directly into the product, but it get inferenced on your infrastructure, the data staying inside of your model. And then you get the guided responses that you want based on your AI infrastructure setup, right? The nice thing about MCP is this became very popular and became the standard on the best way to connect third party data to an AI model. And you get a lot of different MCP servers. So you could connect Veeam MCP, you could connect ServiceNow MCP, you could connect an architectural drawing diagram MCP. And through a single prompt, the AI can use multiple tools to generate whatever it is that you want to generate. So I could get it to speak to Veeam, capture a bunch of data, create an internal ServiceNow ticket to the support team or the storage team or the networking team, whatever it be. I could get it to generate a diagram based on the data pulled from Veeam just by connecting multiple MCPs. The nice thing about it is this all runs inside of your private AI. If you're going to run on-prem AI, you get to plug that into MCP, use MCP connected to connect into Veeam's infrastructure or any other infrastructure. And we've already made the decision when Ian and I were preparing, we're going to have to do a dedicated show just on the model context protocol capabilities and dynamically build visualization technologies off of the backup environment infrastructure. This is cool stuff, stay tuned. You just got to come to watch Tech Bytes. That's how you're going to learn the good stuff. So before we get into the show and tell of some of these really cool capabilities, I want to first walk everyone through kind of some of the large primitives, the large areas that you can get out of Veeam Service Provider Console. So remember at the start, I indicated that Veeam Intelligence really traverses many parts of the portfolio. We're going to focus on the Service Provider Console, Veeam Backup and Replication, Veeam One, I love a good service provider. We're going to start with that one. And some of these endpoints really can help build some really good information. So if you were a practitioner and use some of these things, this is where the ideas should really start to click in. Yeah, I mean, if you look at the depth of data that you could get as a service provider, you log into the Service Provider Console, speak to the co-pilot or the VI inside of Service Provider Console, you could get information about your resellers, discover computers, backup servers that are connected in the different environments and customer environments. You could get all the different policies. So think about how you could tune a prompt to get specific information. So I want the specific policies for a specific reseller, for a specific date, specific session. I mean, that is so simple, so easy. The AI is going to do it within seconds to minutes rather than you having to search around the console trying to gather this information. Also, a lot of times information might live in two different places, two different menus or two different navigations. AI is going to consume this as a single data source, give you a single response in the output that you wanted. If you want it in a table, you'll get it in a table. If you want it in pure text, you'll get it in pure text. That's the wonder of it. That's absolutely fascinating here for a second. And I want to pick up a question that I got from Michael. So Michael had a question. He's bringing the good ones here. Good thing I got Ian. I wouldn't be able to do this on my own. So thank you, Felicia, for relaying this. So Ian, here's a question. Is the MCP path for connecting a tool to leverage Veeam Intelligence plus MCP for another vendor, such as API endpoints versus MCP, what's the difference? So I think Michael's trying to put all types of things together. And I think by having that common tool that you bring it all into, I think that's very possible, but I kind of want your take on it, Ian. So before MCP, it was tools, right? We had to create tools and connect tools to our agents. And then you had to tell the AI, please use this tool and run this and here's the command and X, Y, Z. And effectively the tool was either an AI endpoints or some kind of SDK or whatever it is. MCP allowed you to orchestrate all of your tools, abstract them and your tools are effectively API endpoints and they're self-describing. So once the AI connects to an MCP server, you don't have to tell it what to do. You just ask a question. It determines, hey, in order for me to answer that question, I need to go fetch this tool. This tool is inside of that MCP connection. The MCP server is describing which tool to use for that specific request and how to use it. And then the AI can run the command. Before, when you just had API endpoints as tools, you had to tell the AI, hey, use this tool. This is how you have to use it. This is the data. This is how the parameters X, Y, Z, where that is all fully described inside of MCP. So MCP simplifies the way that AI can use third-party tools by abstracting all of that noise and all of that configuration. So do that make sense? Oh, it does make sense. And it makes it absolutely clear that we need an MCP show. And I'm gonna ask Jeffrey at some point for an hour on that one because the questions will keep coming in. But I did just get notified that the map is ready. So let's take a quick look. And there's still time to let us know where you're tuning in from. Added a few states, Wisconsin, Georgia, New York, looking good. Let's go to another geo. What have we got? We got EMEA. We got UK. We've got Ireland, both of those, nice. And we've got Germany, Italy, Spain. And oh, is that Romania? I think so. Thailand, how about you? Thank you for joining at Bizarre Nariman. Thank you for that. And of course, South Africa. So now, let us know in the chat where you're tuning in from. We always love seeing how this goes. And let's talk briefly. VeeamOne as well has a bunch of really good endpoints. Any of these really stick out to you, Ian? Yeah, I think it makes the most sense inside of VeeamOne when you're thinking of being overwhelmed by a bunch of data or metrics and alarms. VeeamOne is a beast on its own. We're collecting so much data. You can create custom dashboards. You can get everything visually nice together. There's so many reports you could run. But now, if you think about all the out-of-the-box reports inside of VeeamOne that have different metrics, maybe there's a specific report you're looking for, and you probably don't wanna go and click and create this report yourself, you could quickly just ask AI to do this for you. So prompt the AI, ask it to fetch data, collect that, and give it out to you as a HTML report or in a CSV or a table, whatever that be. It collects all of the different metrics, the alarms, the reporting, monitoring, put that all together, and then you get it instantly, right? Rather than you having to go search for something, rather than you having to go create and customize your own report, still good to do that, by the way, because you're gonna reuse that report. But if you're looking for one source of information on the fly, so easy to just ask AI Assistant that's built in to just do that for you. Yeah, and same goes for Veeam Backup and Replication. You know, it all starts here. So many of these posture, scheduling, these elements going into these endpoints really, truly will allow you to do incredible things. And so we're gonna move quickly through here because we got a surprise for you. We do. For one, this is, I think, one of the first questions. And Felicia, if you could push the first link. On many of the streams, we're gonna get some links pushed out. It's super important, I think, that folks start with Veeam Intelligence. First of all, we've overviewed what it can do. We're about ready to show you it in action. But some of these different resources we have here on the screen can help you kind of enable yourself and then share what you have to learn and what kind of interesting prompts you've done. You definitely wanna check these out. So we got some links coming out your way here momentarily. So let's check these out. And yeah, I'll let those links get pushed briefly here for a second. And Jeffrey, hold that one more time, and we'll go two up just in a moment because I wanna tell you all about something that's going on around the world. So we were talking a lot about VeeamON earlier in the year. So we had the New York, the London, the Sydney events, and May, June, and July. But guess what? It's VeeamON Tour season around the world. In fact, Ian, you just had the VeeamON Tour in Johannesburg last night or yesterday, I guess it was. So it's fresh. We're getting ready to launch some. We have three in North America. I'm gonna be in LATAM in September. Let's take a look at what VeeamON Tour can be for you. So when we look at VeeamON Tour, this has a lot of good opportunity. We've got networking, you got product elements, you've got expert bars, you've got our sponsors. We have global sponsors, got keynotes, all kinds of good stuff. And I think one of the best parts about the VeeamON event today is that you can really learn a lot about the new profile, the convergence of data and AI, as it were. So it's really the time to check it out. We also should push out a link that will allow you to find a VeeamON Tour near you. I'll be in LATAM, we got some more in EMEA, and there is that. So VeeamON Tours, coming to a market near you. All right, it's time for a demo. So Ian, tell me what we're looking at and show us some amazing Veeam Intelligence. Yeah, so we're looking at the VSA, so the software appliance logged in. This is the dashboard. We can see all the different metrics. I'm not gonna spend too much time here. I just wanna show you, here's where you can access Veeam Intelligence, right at the top. That opens up and you have your history. You can create new chats. But I wanna show you something even cooler. If we go and I go and look at my jobs, I can see I have a few jobs here, some for Proxmox, one for VMware. I'm gonna go look at this job specifically. I wanna see it's run session. I can see here's a warning. If I click on this little warning symbol, I get this new pop-up that says explain the error. And it's explaining the error using Veeam Intelligence. So it automatically is taking the session data of that error, passing it into Veeam Intelligence. And Veeam Intelligence is now searching that against knowledge base and gonna give me a response. So right there, it's responded. What happened in the latest session? It wasn't a failure, it's a warning. Proxmox worker doesn't have a DNS entry. So it's not resolving our update repository, repository.veeam.com, where you get all the Linux updates from, right? So not a big thing, but great that we can, straight out of the UI, we can immediately access VI through our navigation just by clicking on different elements in the UI and asking VI to get involved and help us with that problem. So the next thing I'll do is I'm gonna show you some actions. And here I've already pre-staged one. So I asked, please collect log bundle for my web server job that I have running. This point it went and it generated a log bundle. So if you think of, you have a support issue, you would have to go into different menus, go and click, go through a wizard to generate a log bundle, export it to a specific disk location, then go and log into your Veeam support portal and attach that. What we're gonna do here is that entire process, hopefully through the Veeam Intelligence right here in the console. So I told it to generate this log bundle. It did for both runs on the 27th of August, that's today. You can see it here attached at the bottom. And the next thing I'm gonna do is let's show me what support cases I have. So I'm going to ask, basically what it's gonna do now is contact Veeam support and try and log me in. And there we go, changes the screen. So it's trying to log into Veeam support. It's logged in and it should pull back all the support cases I have. So it has, there's two support cases. One's for web server with DNS issue. One's a demo that we were trying out earlier. So I'm gonna tell it to please attach these logs to the first case, and we're gonna give it a description. So please upload, and excuse my spelling, upload log. Well, no, it's good. It's good when you spell it wrong because this is AI. It'll pick it up and heck, you can even use multiple languages too. It has built-in multi-language capabilities. Yeah, I think AI made me lazy when it comes to spelling. Well, one thing I'll say while you're typing here, Ian, it's also important to note, we have a kind of a lab license. And so it had some interesting behavior because we don't really, we're not a real customer in our internal lab usage, but that's just a unique anomaly for our use case. But yeah, I love this, putting the log files directly on the case. Yeah, giving it a support ID that it must use, of course, because I don't have a real support ID. And then I'm gonna tell it to add a description. And it's gonna be Rick and Ian were here. So hopefully support will actually ignore it if they see that comment in the case. And let's attempt that. So it's now got the logs that it's exported. They were attached in the prompt, knows what it has. It connected to my Veeam support account. It populated what existing cases I have that are open. And there, it's successfully uploaded logs. So to check that I'm not lying, let's swap screens. And let's jump in. Yeah, and while you do that, I wanna highlight, it shows those quick prompts at the bottom. So it's very contextually aware of what you're prompting and what you're doing. And it has some kind of top suggested actions, which really makes it easy on, okay, what do I do next? And then, but in this case, we're going in here and let's see if we've got it. Oh yeah, there it is, real time. And we have a support engineer that's already responding to our case. That is fantastic, that's awesome, that's awesome. Wow, that's it. Let me jump back. Yeah, and that just, I would love to spend way more time showing you all the different actions. We can get it to start jobs, retry jobs. But if you think about what we've just done and how much time that could save a backup admin on a day-to-day basis, you can ask the VI agent to please investigate the job, try and diagnose it for me. You could try fix it yourself. If you can't, ask it to collect logs, upload it into support. And once it's there, we can check the status of our cases. We can ask for feedback. How's the case going? Any updates? What can we do? If there is a response from support, we could potentially then ask the AI agent to implement some change if it has access to what we're asking it to do, or we could manually do it. So there's a update to my case. Support engineer has said that he has taken ownership, so on and so on. Yeah. Yeah, I'm gonna go ahead. Well, one quick thing to add also is that the multi-language capability is really a big one. And then if you think about larger environments where you have multiple engineers, like from a customer standpoint, like multiple Veeam admins, like, hey, another person, another site or something like that, this is a really good way to have transparency. And then what's been going on, I think just having that visibility on larger teams, having it right in product is really helpful. Yeah, so I'm gonna quickly jump to, once you find a great prompt, because everything's about a prompt. So once you've discovered a great prompt, share it with the community, share it with the rest of the Veeamers. You can find it on Veeam's prompt library. If you go to the sites, it speaks about basic and advanced mode, just reminds you that there's different ways that we run the Veeam intelligence. And specifically, there'll probably be different prompts based on if you're running it in basic mode. And if you're running an advanced, you can filter then in basic and advanced mode, what types of prompts you want, the platform, and then the topic. Maybe we want code and automation type prompts. And there we have a prompt itself. So you can view the prompts, copy the prompts. So we view it, and potentially it's gonna have the prompt in writing, in text, sample output. So what you should expect back from the VI itself. And that basically gives you an idea if this is the correct prompt for you. So you can come and access prompts here. You can share prompts here, but it's as easy as copying that, jumping back to your backup server, and pasting the prompt in, and you could get a similar result. Yeah, and note, you switched out a backup admin agent, specific context here. And this is just the start here. Everyone's gonna be on a journey for these prompt opportunities. Veeam intelligence is where it's at. I would be remiss. Shout out, California, Canada. Hey, Matt. Another South Africa. We also had California. We had Armenia. A couple other folks joined in India. Thank you all so much for watching. And Ian, fantastic information. Gotta have you back, and we're gonna have to talk MCP for sure. Sure. We have to do an MCP session. Absolutely. Well, thank you all for watching, and a big shout out to Jeffrey in the Barncade, as well as Hazel and Felicia bringing in the content for you here today. Thanks for watching, everyone. We'll see you next time on Tech Bytes. ♪♪ ♪♪
