Good morning, good afternoon, everyone. My name is Kimberly Sack. Thank you for joining us for today. I'm going to be the moderator, host, and one of the presenters as well. I want to thank you for joining us for today's Cutover Live session. Cutover Live is our series of short informative sessions on improving IT operations. So things like IT disaster recovery, cloud migration, major incident management are our focus. Today's session is super charge your Doctor or disaster recovery plans from creation to optimization with Cutover AI. And before we kick it off, I just have a couple housekeeping items to discuss. So for those joining us via Zoom, you will find a q and a box on your screen. Please feel free to submit any questions you have during this session, and myself or another presenter will answer them at the end of the session if we have time. If we run out of time, we will absolutely follow-up with you after. And for those joining us on LinkedIn or LinkedIn Live rather, please comment any questions into LinkedIn, and one of our moderators will get it to us, and then we'll answer it again at the end. This session is being recorded and will be available at the end on LinkedIn. So over the next twenty to thirty minutes, we're gonna cover some key challenges in IT disaster recovery and major incident management and opportunities that AI can help solve. Then we're gonna share a short live demonstration of CutOverAI, so you CutOverAI so you can see some capabilities and benefits, and then we'll wrap it up and open it up for questions and answers. So, again, I'll be starting us off today. My name is Kimberly Sack. I'm a principal product marketing manager at Cutover. I've been in the b to b technology space for about twenty years, and at least a decade of that has been focused on product marketing. I'm excited to be at Cutover for almost four years now. I'm excited to start us off today, but joining me is Marcus Cowles, senior customer success manager at Cutover. Marcus, just gonna hand it to you for a quick intro as well. Thanks, Kim. Yeah. Hi, everybody. So Marcus Cowles. I've been with Cutover for about a year, but I came across Cutover about three years ago when I was with AWS, where I was a resilience and operations consultant and very much focused on operations and management governance. So working with Cutover now, absolute match made in heaven. I'm very excited to be with you today. Great. Thank you, Marcus. So let's start by looking at some of the common day to day challenges in managing IT disaster recovery, incident management. You know, we've gathered some of this from conversations with customers, prospects, partners, you know, people we work with and and communicate with in the market, and some of them may sound very familiar to you. So it often starts for IT disaster recovery. There's a a blank page problem. Right? There's a a wealth of knowledge in your organization, but it's scattered across wikis or Confluence pages. And the idea of manually translating all of that unstructured information into an actionable plan can feel overwhelming. So it's oftentimes very hard to even get started. Then during an actual event, you know, especially incident management, the problem becomes one of chaos and complexity. Right? So teams are constantly switching between different systems, what we call swivel chair. The critical updates and discussions get lost in chat tools, makes it impossible for leadership to get a clear view of what's happening or for teams to effectively coordinate. And then finally, once the dust settles, you know, the event has passed, you're recovered and responded to everything, The challenges shift to analysis and improvement. Right? So you look at a runbook or your plan, and you can't tell what its original purpose was or where things fell flat or where there was a a weakness spot. So you're left wondering, where can I improve my plan to better meet my objectives or to recover faster or meet an RTO if you missed it? And trying to piece together the sequence of events for a post incident review or an audit becomes really painful and can be a very manual process. So these are precisely the the persistent frustrating problems that have plagued operations teams, and it's exactly where automation and AI, when applied correctly, can fundamentally change the game. So the industry is recognizing this potential, and Cutover executed a our third annual IT disaster recovery survey this year. We also recently did one on major incident management, and we're gonna be sharing those results in a couple weeks, so stay tuned for that. But when it comes to considering AI and how it solves some of the problems, you know, the data shows a very clear and exciting trend. Right? So we're seeing a strong consensus from enterprise leaders that they're not just optimistic, they're truly excited about the opportunities that AI presents, with over eighty percent believing it'll fundamentally transform both the design and execution of their Doctor plans. And it's not just vague excitement. So leaders have specific practical expectations for where AI can deliver the most value. So as you can see, more than half expect AI to analyze post event metrics to find improvements, analyze plans while they're in progress, and and also create templates and runbooks right from the start. You know? However, there are some considerations or barriers. And as you can see in the chart on the screen, enterprises have concerns about incorporating AI into Doctor. And from that goes from the people side being the leading concern, lacking AI skills internally, to finding appropriate vendors and partners, to compliance, to data quality. So as you can see, there's not one true outlier. You know, there's a lot of concerns and considerations for incorporating AI into Doctor, but the appetite is there. Right? The use cases are there. There's excitement about it, and it sets the stage for a new approach. And moving beyond, like, the traditional manual time consuming methods. But how can you harness that and bring that potential into practical use? So that's where CutOverAI can help. So we've designed our platform to directly address those core IT operations challenges and meet the expectations the industry has for AI. So remember the feedback. It's hard to get started from knowledge in Wikis and Confluence pages that I mentioned. You know, our AI can help solve that. Right? Taking existing unstructured data, and that can be in a multitude of formats, and Marcus will talk more about this, and instantly convert it into a fully formed executable runbook. Right? So reducing creation time from days or weeks or longer into into minutes. Then to address the problem of, well, what does the runbook even do? Cutover AI automatically generates clear, concise summaries. So leadership, executives, key stakeholders, people that need to understand what's going on but aren't actually in the weeds and executing the the tactics and tasks, you know, can have that at a glance. And then finally, when it comes to how can I improve my runbook for this next for the next time around, you know, our AI can look at performance and suggest improvements to it, right, and to keep that cycle of continuous improvement going? And delivering all of these powerful capabilities requires a really secure and trustworthy technical foundation. So, you know, we believe in using the best tools for the jobs and not reinventing the wheel. So we're built on best in class foundational foundational large learning models, LLMs, from Anthropic, Amazon, and Meta. And we partner very closely with AWS to give it so we get access to the the latest and greatest enterprise grade innovations like AWS Bedrock and Amazon Q. So given all of that, I'm excited to hand it over to Marcus who can show this in action. Amazing. Thank you so much. Okay. So as Kim Kim has kindly introduced us with, there is a lot of complexity and a lot of details to be thinking about when we want to leverage AI in order to create runbooks. At Cutover, our goal is to give customers the the right tools and versatility to be able to develop the content that they need. But the biggest barrier to entry and challenge that we find customers having is how do I get started. So here on my screen I've got a number of pre canned templates that customer cutover comes with. The top one being a template on prem to on prem backup and restore. And we find that these help customers springboard forward so that they have something to start with and develop from, but it's always a one size fits all fits none. It never specifically solves a particular customer problem. So what we wanted to be able to do is give them to give our customers the ability to take this big first step, but in a more specific way, and that's what I'm gonna demonstrate today. So fundamentally, if I want to take my process and I wanna get it into a runbook, I've got three approaches. I can start from scratch, the blank page approach. I can do an import from CSV or some other document, or I can use the AI create feature, which I'm gonna demonstrate today. So let's get started. I'm gonna go over into my runbooks interface here, and you'll see I've got a number runbooks already created. And if I go into create runbook at the top and select normal runbook, here's where I get some of those options. So I mentioned we've got a number of templates. Some of these come pre canned, provided by CutOver. Others can be developed by the customer. I can have a completely blank runbook where I just start with a blank page, or as I'm gonna demonstrate today, I can use the AI. Now fundamentally, what I'm doing here when I prompt to the AI is I'm gonna give it an instruction to take details that I want, and then it's gonna return me a runbook straight into CutOver that I can take forward. So I've got a prompt that I've prepared here already. I'm just gonna bring this up on the screen. So I'm gonna be doing a multi region DRS failover in this case. So DRS being elastic disaster recovery from AWS and my use case here is gonna be an AWS application. So you can see at the top of my screen, I've got a description statement and underneath that I've got a purpose. And in this case, providing the CutOver AI with two text statements, I'm gonna be giving it these details of what I want it to produce. So the description statement, really high level. I'm telling it that it's gonna be an AWS application, so immediately knows it's cloud. I'm gonna be failing it from US East one to US East two, so I'm telling it that it's a multi region, not single region application. And I'm gonna be using some AWS services, ARC, application recovery controller, DRS, elastic disaster recovery and route fifty three for DNS routing. So what's really important here is I'm being very specific about the things that I know about my application. I'm not gonna leave the AI to guess what services or what architecture I'm using. What I am gonna ask the AI to do is use the various cutover features in the right way to achieve the outcome that I want, which might be something that as a new cutover customer, I'm not necessarily an expert in. Moving on to the purpose statement, we can see that I'm asking it to create a recovery runbook. Really straightforward, but I'm being specific about what I want. It's gonna fail over my AWS application. So I don't wanna do replication in real time. I'm telling it I'm gonna do a fail over scenario. In the blue text, I've also given it some tips around the operational and governance elements that I want to include. So I want to use parallel execution where possible, which is really important because otherwise the AI might interpret my request as being a necessary sequential series of of tasks or events. I've also told it that I want to add integration task placeholders. So I'm not necessarily saying to it, I want you to develop these integrations for me. That's gonna be very specific to my technology needs. But I do want it to put in place tasks where I could add integration later. So it's gonna be scalable and iterative from this base runbook that it's gonna create for me. And then in my final statement, communication and validation tasks, completely nontechnical, but governance and process elements that I want it to include. So I've got a screenshot below of the text that I'm gonna enter into the prompt. You can see that the title is my description at the top, and then that whole paragraph purpose goes in there. Critically, what I wanna draw your attention to is that in the supporting information for this example, I'm not gonna add anything else. We're just gonna give the AI those two statements above where I've been specific about services and process points that I want, and I'm gonna let the AI decide which tasks, how many tasks, what structure it's gonna create for the runbook. In the next example, we're gonna give it an architecture diagram, and we'll be able to compare the two. So let's do that. So here I am again in the CutHover interface, create normal runbook using the AI. So I'm just gonna grab my description statement from my other screen and pop that into the title ending with Route fifty three. Yeah, that's correct. Then I'm gonna grab the purpose statement and pop that into the prompt. And it starts with create and finishes with validation tasks. So all of that has gone in there. As I mentioned, on this example, I'm not gonna add any supporting information and I'm gonna make sure that I put this in my MC folder. So I'm being nice and clean with my creations. Now what's gonna happen in the background is CutOver is gonna send this information to the AI in the form of a prompt along with some other enhancements and specific training models. And then it's gonna return me a runbook. These other metadata elements below I could add. We can have all sorts of additional data against the runbook, but for the purpose of this example, I'm gonna leave these as they are and click create runbook. Now it's not instant. The AI goes away and churns and processes this, and it's gonna give me a prompt once it's created. But what I can do is go off platform and it will send me this nice tidy email once my runbook is created with a link go to that runbook. So I can sequence these, schedule them, go off and do some other work or take a break and then come back and it will notify me once that runbook is completed. Now for the purpose of today's demo, I've created one yesterday which we can jump straight into. So if I filter on my folder, here it is at the top, failover of an AWS application, etcetera. And this is one that I created yesterday with exactly the prompt I've shown you. So let's have a look at what it generated for me. Okay, so we can see straight away that it's given me a series of tasks, which is perfect. They all have intuitive names and we can see that the different task types have been set. I'll talk about that in just a moment. On the left hand side, really nice feature here. It's already created a number of streams for me around the types of activities that I'm gonna be executing in this process. And this is fantastic from a high level based on the prompt that I gave, I can see how the AI has structured the process into these various phases. Naturally, can edit the names of these streams. I can add a new stream, delete one, etcetera. But the AI is showing me for the prompt you've given me, I think you need these phases and this is the number of tasks I've put into each phase. If I want to filter on one of those streams, I can see just the tasks in that stream. So as I'm assessing the output that I've been given or want to enhance it, I can filter down just on the elements that I want to look at. We can also see, as I mentioned, that CutOver has used a series of different task types. So the diamond at the top is a milestone. So that's a governance checkpoint to say that something has started. The triangle is a decision point. So it's gonna ask me what I want to do at this stage. And then the green with the brackets is a communications task where I'm gonna send an email out. And that is an automated task within CutOver based on some inputs that I can provide it. Further down, we can see the circular task. That's just a normal task. So here it's saying you can customize that yourself. Perhaps I could change that to an integration task and hit an API end point. And if we look further down, we can see that as the streams change, we've got colored denotion of that and various other task types taking place. So intuitively, this looks great. It's really useful. If I go over onto the right hand side and click the AI suggestions, sorry, the description, we'll see that the AI has also generated a description of my runbook. And this is really good because it cuts down on the manual effort that ISD user have to put in here. So not only has it taken my prompt, but it has also developed the runbook and then reverse engineered a description of what is just produced. Really useful for anybody else coming into the environment saying, oh, Marcus has created runbook. Instead of trying to interpret what he's trying to do from the tasks, they can just read the description. It's something that I didn't have to put effort into. Now I mentioned earlier that in the prompt we specified parallel execution. So if I go over to the node map, we'll see what's happened there. And we can see that the AI engine has really taken that on board. We've got a few serial tasks at the beginning, initiation, assess the scenario communications. And then in this pre failover preparation, we've got this bubble of things where the AI is saying you could do all of these tasks at once. They can happen in a parallel execution mode. However, it's also paid attention to the fact that they all have a dependency upon a validation task. So this validate task will only become startable or enabled once all the prereq tasks have been completed. So immediately I can see that I've got governance and dependency mapping built into that AI created runbook straight out of the box. I haven't edited this since the AI originally created it. If we look a little bit further down, we've also got another bubble of parallel execution around the execute failover task whereby we can have the DRS recovery initiating and updating Route fifty two records happening at the same time. Now perhaps I might want to edit this sequence. I might want to change the dependencies, but it gives me a really good starting point as to where I can develop from. Looking further down, we've got a post failover validation step, which makes perfect sense, taking me into another milestone and then the failover success decision where I can record whether the failover was successful or not for my governance and auditing purposes. And then it takes us through to a final milestone. So overall, I think this is pretty good. I think this is not far off what I would come up myself manually, but as I say, the AI created it in a matter of minutes and gives me the opportunity to iterate and develop from here. Additionally, if I go into an individual task, let's say the failover decision and I edit that task and go into the description, it's given me some text about what happens at that step. So even as the operator, I'm getting instructions on how to execute this particular step from the AI and these descriptions will be present for every single task in the runbook. So a really impressive output from a relatively simple prompt. And if you've ever done any prompt engineering or AI testing and development, you'll know that from here, you might go back to your prompt and tweak it and say, also do this, or actually, I might also want to use that service. Create a new runbook with that prompt, see what the output looks like and go forwards from there. So a really versatile and very functional way of putting in a relatively simple input and getting quite a complicated runbook output at the end. Now I mentioned the other AI features earlier on. So as I'm developing and testing this runbook, I can also make use of the AI assist features which exist within the runbook. And we can see here that we've got a few really useful ones suggesting detail improvements which might be okay Marcus, you've got these various steps here but actually you need to be more articulate about what this task is doing or perhaps add an extra task. Stream improvements is a fantastic one that we make use of if I import just a flat list of a hundred tasks and they don't have streams, they don't have structure, the stream improvements can help break those tasks up. Redundant tasks, really useful as we develop an iterate and tasks are no longer required. The AI can suggest maybe this one's not necessary, etcetera. And dependency improvements, really nice where we want to make sure that we've got our interconnectivity of tasks as tight and as clean as possible. Now, because this runbook has just been generated by the AI, I wouldn't go and run the AI assist straight away. I would use this as my co pilot as I'm iterating and developing this run further, rehearsing it and testing it, and then using these various assistance to help me tweak my process. As with all of the cutover features, the suggestions are for me to consume and then make a decision on. The AI isn't gonna run off and go and do things without my without my control or without my consent. Okay. So that was creating a runbook from a simple text prompt. And as you can see, it gives us a fairly complex, comprehensive, but easy to interpret outputs. And if I go back to my list of runbooks, we'll see this one at the top is the one that is actually just created from my prompt. So that one has come through. So it only took a couple of minutes. So now let's ask the AI to do a little bit more heavy lifting. Now I'm gonna give it a different prompt in the form of an image. So we're gonna have a very similar structure with our prompt. I'm gonna have the description multi region failover, but this time from an architecture diagram. My purpose statement, I'm gonna say using the architecture diagram attached. So I'm being specific about what I want, make me a runbook, which will help me failover. Again, I'm giving it the hint that I want it to use parallel execution. I want integration task placeholders, and I'm telling it to include communication and validation. Now it might be that the antithesis is of this is true, and I might say, I don't want any validation tasks. I want a clean run which wouldn't require any human intervention. So I can be really versatile here in my request. You can see the diagram is a fairly straightforward pilot light failover. I've got US East one on the left, and in this case, I've got US West two on the right. Now I want you to note that in the prompt, I'm not telling it which services I'm using. I'm not even telling it how many regions I've got. I'm gonna let it interpret all of that from the diagram, whether I'm using EC2 ASGs, I've got the VPCs naturally, we've got some network load balances and so on. I'm gonna ask the AI to interpret all of that from my image. So let's see that happen. So as before I'm gonna go to create runbook, normal runbook and using the AI, I'm gonna take my image prompt, which I just showed you, pop that into the title, multi region failover from a diagram. I'm gonna take my purpose statement again, the one that I just showed you. So using the architecture diagram and validation tests. So that's all gone in. Now this time in my supporting information, we're gonna go with file upload. And I just so happen to have an image that I already have prepared here. The AWS pilot light architecture diagram, which I showed you just a moment ago. Once again, I'm gonna stash this in my AI create demos to keep everything neat. And as before, I'm gonna leave the other fields as they are and click create. So exactly as with the text prompt, the AI is gonna churn in the background and let me know once this is completed. But for the purpose of today's demo, you'll notice if I filter down on my folder, I've already got one here that I created yesterday from exactly that prompt and exactly that image. So let's see what it gave me. So straight away, we can see we've got a mixture of task types. Fantastic. So CutOver has interpreted the fact that we've got a process which requires various types of activity. I can also see on the left hand side that it's broken the process up into streams. Now you may recall that in my prompt, I didn't tell it to do that and I didn't give it what the stream names needed to be. So I might say, oh, actually, I need a minimum of five streams or one of the streams needs to be governance or whatever that might be. So I could go back and make my prompts more specific very easily in order to tweak the output that I want. Of course, I could alternatively just create the thing that I want without going back and recreating, so I've got both options. As before, if I go into the description of a task, we'll see that the AI has given me some guidance about what happens in that task, what I should be expected to do. And if I jump into the node map, we'll see that it's taken on board the parallel execution. So verification tasks, they can all happen at the same time. We can see that it's picked up that we've got a database replication involved. We've got EC2 auto scaling groups and we've got NLBs network load balances present. So it's gleaned a lot of detail from that image that I gave it. And if I wanted to be more specific, I absolutely could be. If we look a little bit further down, we can see that it's grasped the idea that we're gonna be promoting standby clusters to primary and scaling a PC to auto scaling groups. So it's even dived into the detail that we don't have an active active situation and we don't have our ASGs on the pilot side already scaled up. It's recognized that we need a task to do that. And these two tasks are really good examples as to where I might plug in an AWS integration at a later date. So scaling up an auto scaling group, that's a really simple activity for AWS to do. For me as a human to log on to the console and click on the ASG, I can do that, but we can further enhance our runbook by swapping this normal task out with an integration task, which goes and does that automatically. If I look a little bit further down, we've got a great process bubble here of post validation, post failover validation. So again, it's not necessarily a technical step, it's governance and auditing, and we've got some communication tasks, fantastic. And then we've got the post failover configuration if I want it. So again, a pretty comprehensive process from a relatively straightforward prompt and a fairly simple image. So I think for today, in the interest of time, that's what I would like to show you. Kim, if I could hand back to you for the wrap up, if that's okay. Great. Thank you so much, Marcus. Okay. Great. So, you know, thank you, Marcus, for that twenty minute demo of CutOver dot ai. Hope everyone found it informative. So just a quick recap of our session today. Right? So we started off by discussing some of the key challenges of the manual nature of disaster recovery, incident management, creating plans. Then we talked about how AI is posed to help solve these problems, how organizations are excited about it with some data from our recent survey and report, and how organizations can kinda move beyond the traditional manual nature of them and move towards using and embracing AI and finally showed in real time a live demo of how Cutover AI can help you instantly create these executable runbooks from multiple formats, summarize them, and make them suggest improvements as well. So while we open it up for questions and answers, I just wanna highlight we have a couple upcoming events. Approximately every two weeks or so, we do have our cut over live sessions, and our next one in on November fifth, same time, will showcase the findings of our major incident management survey that I had mentioned. So we're really excited to showcase that. It's the first year we've done this one. It complements the IT IT disaster recovery one we've done for the past three years. We also have a live, what we call an immersion day or a live workshop that we are gonna have on November sixth. It is our it's on AWS Workshop Studio, so you can always search it and find it. We are having it. It's a two hour free workshop that if you wanna get hands on, build CutOver Runbooks, see how the automation works, we have integrations with AWS Fault Injection Service or Fizz. So if you wanna be able to test in a cloud environment in an AWS cloud environment rather, test your your recovery, inject chaos, you can do that live. So it's a great way to kinda take the next step in evaluating runbooks and kind of the way that automated runbooks and automated Doctor plans work. So we hope you can join us there. Please reach out. And then last but certainly not least, we will be not only attending but also sponsoring AWS re:Invent, which is the first week of December. We have a great executive dinner. We have a booth there, booth two three three. Please feel free to stop by. You can scan the QR code and check out our page with all of our all of our options. I apologize. We did get a few questions in, but we don't have time. We have a hard stop because of Cutover Live, but we will capture these questions and follow-up with you after. So I wanna thank all of our attendees. I wanna thank Marcus as always. Great presentation. You know, find us on LinkedIn. Follow our CutOver page. Reach out to myself or Marcus. Reply to one of our emails. You know, we'd love to discuss how CutOver can help you kinda move into more automated and AI generated recovery and response. So with that, thank you everyone, and have a wonderful rest of your day and week.
Building and maintaining disaster recovery runbooks is a time-consuming, manual process. How can you ensure your plans are created quickly and continuously optimized for success? Watch this 30-minute session to see how Cutover AI instantly generates comprehensive DR runbooks and intelligently analyzes them for critical enhancements to uncover bottlenecks. Build and refine your DR plans faster than ever!
