Send us Fan Mail Can AI safely troubleshoot and resolve IT issues without putting your entire endpoint environment at risk? NinjaOne Field CTO Jeff Hunter returns to ClearTech Loop to join host Jo Peterson for a conversation about Katana AI, the company's new approach to AI-powered endpoint management, IT automation, and human-centered decision-making. Modern IT environments generate enormous amounts of data. Katana AI is designed to turn that information into actionable context, helping technicians identify potential root causes, share institutional knowledge, and recommend remediation while maintaining human oversight. In this episode, Jo and Jeff discuss: Endpoint intelligence: How Katana AI connects alerts, device history, and operational data to help technicians understand and troubleshoot issues. Knowledge sharing: Turning experienced technicians' solutions into reusable knowledge that helps the entire IT team. AI-assisted remediation: Recommending solutions and scripts while keeping technicians responsible for approving actions. AI governance: Role-based access controls, permissions, data boundaries, and MCP access. Human-centered AI: Why the goal is making technicians more effective rather than replacing their judgment. As Jeff explains, "The very last thing you want to have happen in IT is making a bad decision fast." For CIOs, CISOs, IT leaders, and managed service providers exploring agentic AI and intelligent endpoint management, this conversation examines how organizations can balance automation, operational efficiency, and accountability. Resources & Links NinjaOne Katana AI: https://see.ninjaone.com/uueAH Jeff's Previous ClearTech Loop Interview: https://www.youtube.com/watch?v=UKxb1Bxs4z4 Current Episode Video: https://youtu.be/BIx5plv1OK0 ▶ Watch on YouTube: https://www.youtube.com/@ClearTechResearch/videos 📰 Subscribe to the Newsletter: https://www.linkedin.com/newsletters/7346174860760416256/
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
SPEAKER_01Hey everyone, thank you so much for joining this edition of Clear Tech Loop. I am here with Jeff Hunter, field CTO Ninja One. Hi, Jeff.
SPEAKER_00Hi, Jeff. Pleasure to be on the show again.
SPEAKER_01Thanks for coming. So super excited to have you here and talking about the new product that you're rolling out, and that is Ninja One Katana AI. And so, you know, in our normal hot and take format, we're just gonna start by asking you questions and you can fill us in on what's new and exciting about Katana AI. So you and I've had a chance to talk, and you showed me a product demo and got me thinking about some things.
SPEAKER_01Every endpoint throws off signals. Alerts, patch events, performance data, activity logs. Individually, none of it's hard to read. But here you are, you know, as an analyst trying to look at what's really going on across your environment and at scale across a growing fleet of devices, it becomes noise that you gotta dig through before you can find anything that actually matters or kind of piece together a picture of what's going on.
SPEAKER_01How does Ninja One Katana AI help prioritize that noise and what actually matters?
SPEAKER_00Yeah, and you bring up a great point, which is that an endpoint management platform generates a massive amount of data, a massive amount of information. And at a certain point, uh you start to reach this bottleneck where context gathering, information gathering is a key part of most tickets, because it's unrealistic for a technician to immediately know every bit of information they need to know in order to solve whatever ticket that they're presently faced with. And so with Katana, our goal here is to shortcut that process.
SPEAKER_00Our goal is to shortcut that information gathering, that context gathering that is a bottleneck in so many tickets. Uh and so, with our tool, because we're integrated into the platform already, we have access to the data points. But more than that, we know what the data points actually mean. So that when you ask a question to Katana, it is not just a general AI model that's trying to figure out, okay, what does this data point mean? What does this data point mean? And then sort of stitching together some sort of context that may or may not be accurate. Uh, that context is already preloaded into Katana while you're in the platform and you're actively using it.
SPEAKER_00Uh, and so you're able to get that quick summary. Uh, you're able to have those alerts surfaced and you have the confidence that Ninja One is going to provide a more accurate summary because we have all that contextual data already in the platform. So we're going to be able to give you that answer of, okay, we have these alerts that are triggering. This means that these services are then down, which then means that you're going to have critical disruption for this particular server here. Uh, that is uh something that I think could take you 10 taps uh that you need to open if you're a lower-level technician before you sort of realize the full import of what's going on here.
SPEAKER_00And then, of course, how does it actually get fixed? Uh, and this is, I think, again, where Katana is going to excel because we're able to take all that massive amount of data, we're able to show you what really matters there, and then we're able to allow you to quickly uh act on it while keeping that human being in the loop so that we're not substituting uh a human's judgment for something that uh an LLM is going to spit out.
SPEAKER_01Yeah, and that's important. And I think as we all get further in our learning journey with IA, AI, it feels like the thing that is happening right now that's bubbling up is just what you're talking about. It's context and it's trust. So if you've been around for a while in the IT field, you know the traditional monitoring tools show you the data. And to your point, they don't tell you what it means or what to do about it. So the burden falls on the technician to correlate an alert with it with a device's history, cross-reference it against past issues, decide whether it's worth escalating all the things, right?
SPEAKER_01Before even touching a fix. Or, you know, so why is katana AI different and how does it help every technician build expertise faster?
SPEAKER_00Yeah, and that's I think a very good point there, which is that we've explicitly, you know, designed katana with the idea that it is going to be a human in the loop, that the goal of the uh tool is to make the human being who's using it better at their job, more effective at their job. Uh, and so you know, we're able to go ahead and provide uh what the root causes are, uh, recommend what the likely next step is. Uh, but how do we get to that point? How do we know what the likely next step is?
SPEAKER_00And I think this is where it's the power of taking your best technician and allowing them to be sort of a force multiplier for all your other technicians. Uh, if you imagine that a senior level tech is going and they solve two of the same type of ticket, same sort of problem in one day. And they go, Well, we should probably go ahead and create a KB article for this. Uh, with Katana, they just go to Katana and say, Hey, this ticket that I'm currently on right now, and Katana has the context for whatever you're looking at within the platform, uh, hey, Katana, go ahead and create a KB article for this ticket.
SPEAKER_00And Katana just takes what's inside the ticket and it creates a KB article that goes into the Ninja One knowledge base in the documentation section. Uh, and so now uh the key thing about Katana is that it does not have internet access. It is not going to go out and try to figure something out from the general internet. Uh, it's going to use what's inside the Ninja One platform. And so now you have Ninja One consuming its own knowledge base, preferring that, or Katana preferring its own knowledge base. Uh so that's where the information is coming from.
SPEAKER_00And so now when a level one technician, somebody who's brand new, comes in and they see this issue, they go to Katana and they say, How do I fix this, you know, this ticket that I'm on right now? Katana is going to prefer the knowledge base. It'll go and find that article that was earlier created by a senior level technician and it will spit back out to the level one technician, the brand new guy. Uh, they're going to say, Hey, this is how you fix the issue. Here's all the relevant information that you need. And as it turns out, the senior level technician actually created a script that goes through and automates the fix for this particular problem.
SPEAKER_00Do you want to go ahead and run it? And at that point, it is a simple matter of going and just running the script, clicking the button to actually execute it, confirming that it actually fixed the issue. And then from there, Katana could even summarize a response indicating that the issue has been solved back to the requester who is having the issue. And so right there, it's that, you know, uh one to many. You know, we're going one to many here. Uh, from, you know, maybe one high-level technician who knows everything, that knowledge can now be easily consolidated inside of Ninja One.
SPEAKER_00And then Katana is going to make sure uh that that information is easily accessible to all the technicians who are using it. And that over time is gonna naturally uplift those technicians. It makes it easier for them to find the answer. It makes it easier for them to find the answer independently, uh, which I think um is going to improve not just speed, but I think is also more effective for somebody's own learning, their own education when they're doing it independently. I think that that sits better in their brain. Uh, and so uh again, we're going from your best technician.
SPEAKER_00It's no longer just your best technician as one person. That best technician is now somebody whose knowledge can be consolidated inside of Ninja One, can be accessed by Katana, and can then flow to all of your other technicians just uh as a rising tide lifting all boats.
SPEAKER_01I love that analogy, the rising tide, because it is, it feels like that. It feels like everybody's getting a benefit here. The seasoned techs are able to share their knowledge. The newer, fresher techs on the scene are able to just, you know, take that knowledge when they need it. That's so important. Um, so along with context and trust, it feels like access is also one of the AI tools' top topics right now. Um, every CIO in CISO that I'm speaking with is concerned about, and rightly so, AI access.
SPEAKER_01How does Ninja One control katana AI access?
SPEAKER_00A great question, and one that we wrestled with, I think, quite a bit before we started to enter the realm of actually putting hands on keyboard to actually write code. Uh, and it's because, you know, we we have seen in all our conversations that AI is not going to be used if AI cannot be governed. You know, the control has to come first before the raw capabilities of the AI model or the AI tool actually matter. Uh, and I think that's the correct tack to take. Uh, you know, this massive amount of information that uh AI is able to consume, you know, it means nothing if you're not able to effectively govern it.
SPEAKER_00Uh and so we thought deep and hard about governance uh, you know, for a very long time before we actually started building code. Uh and so with Katana, it's going to respect the role-based access controls that you've already established within Ninja One. Uh and so if you have access to workstations, but not servers, because you're not on the server management team, you're not going to be able to use Katana AI to do something potentially dangerous with the servers. You're not even going to know that they exist. Uh, and so right there, that's one element of uh role-based access controls so that you don't have somebody with kind of broad access to the entire system.
SPEAKER_00But there's also the individual tools that Katana is going to have access to. Uh, case in point, I mentioned earlier Katana recommending a script and saying, hey, do you want to go ahead and run this script? And you just need to click click a button for that. If you're not comfortable with that, you know, you want to be able to turn that feature off. You might not want an AI tool to recommend a script that a low-level technician can run. And so we've got multiple kill switches here. We've got sort of an overall kill switch at the uh over you know the top level, where you can say, I don't want katana to use information or use these features at all.
SPEAKER_00I don't want katana to recommend a script at all. I don't want katana to recommend any knowledge base articles to solve problems at all. Uh, but then we go down to the RBAC level. And so we can say, all right, for a level one technician, uh, they can't use the script search feature to run a recommended script. Uh, they can't create KB articles because they're a lower level technician. I don't want them to go and end up creating a KB article that's going to be incorrect. And so then Katana starts uh consuming information that is incorrect from this knowledge base article that was created by somebody who shouldn't have had permission to do that.
SPEAKER_00And so that's why having these effective permissions over what somebody is able to do, not just in terms of the devices or device types that they can access, uh, but in terms of what features of the AI tool can they use. Uh we're talking about Katana today, but you know, do we not want them to have the ability to use this as an MCP? Uh and so we can go ahead and turn that off so that way they're restricted to only being able to use the AI tool actually inside of the platform, as opposed to it being a little bit more portable if they have access to the MCP.
SPEAKER_01Yeah, that's really important. And when you and I were talking offline, you know, we were talking about what Katana can access as training data. And I think it's really important to note kind of the hard line in the sand that Ninja One has put in here to protect users. Can you talk about that a minute?
SPEAKER_00Yeah, 100%. Uh there is no internet access. I think we're all very used to whatever LLM is your favorite. You can ask it a question and it'll go and search web pages and it'll come back with information. And that's great, but uh that can also lead to hallucinations. That can lead to things being made up. Uh and so at Ninja One, that's a hard firewall. There is no internet access for Katana. Uh, Katana has access to a limited set of information. Uh, it's great, for example, at searching our dojo, which is the Ninja One Knowledge Base, so that you can find out where you need to go next inside the platform and uh how to take certain actions, like integrating Microsoft 365 into Ninja One.
SPEAKER_00Uh, but then it also is going to prefer information from the knowledge base. Uh, and so again, this is a way to customize Katana to your own processes and procedures. Uh, you go and you put in all of your knowledge base articles that dictate how help desk technicians should operate. That information is preferred first. That is the thing that Katana is going to reach for first when providing an answer. It's going to look at all those knowledge base articles, it's going to try to find the right one. Uh, and so when it is providing that information, that is not something that is uh, you know, just pulled from the internet out of a hat.
SPEAKER_00Uh, it is not going to be just general inference. It is going to be pulled from the information that you store within Ninja One that you know is good and valid and it works for your organization.
SPEAKER_01And that is so important. But the other side of that coin is that Ninja One is not using some other company's information, nor sharing my information outside of my organization, right? The firewall is works on both sides.
SPEAKER_00100%.
SPEAKER_01You know, so um you had alluded to it earlier, the training aspect of Katana AI. Um, but let's talk about the workforce aspect of it. So as organizations grow, the volume of endpoints and service desk, service desk tickets, well, you say that three times fast, usually demands this proportional increase in IT staff. Makes sense. How does Katana AI act kind of as an operational force multiplier?
SPEAKER_00Yeah, and and to the point that we discussed a little bit earlier, Joe, uh, we we discussed you know the information gathering stage, the context gathering stage. Uh, you know, any technician is going to need to gather information at some point during their day, maybe not for every single ticket, but at some point throughout their day, they're going to need, oh, okay, I don't remember how what this procedure is. I need to go and look this up. And so we find that that ends up being a bottleneck. Uh, and organizations that deal with it well, uh, they have much more streamlined IT departments and help desks.
SPEAKER_00Uh, and organizations that don't deal with it well, it's typically because they're not able to shortcut this process. They're not able to effectively shortcut the context gathering and information gathering stage so that they can operate effectively. This is what Katana does very well. To me, this is the force multiplier of Katana is the ability to go and make it so that this uh knowledge gathering stage is uh is shortcut. Uh, so that you don't spend an hour and 10 or 15 tabs of doing research to find the information you need to be effective at the task at hand.
SPEAKER_00Uh, you're able to get that in a much shorter time frame right inside the console where you're already using Katana AI. Uh, and so being able to shortcut that process right there, uh I think is tremendously useful. And again, speaks to the point I made earlier about you know, uplifting all of your technicians. Uh, you have your your you know, your rock stars, you know, you have your S-tier technicians uh who are going to always do a great job, uh, but being able to take all your other technicians and make them better through this tool, that's absolutely key.
SPEAKER_00And then there's the other portion that we talked about earlier, which is taking that best technician or those best five technicians that you have. Uh, and now you have the capacity to go ahead and take their knowledge and insight and put it into Ninja One, where Katana is going to process it and then convey that information downstream to all of our other technicians. Uh, so it's not just that Katana is going to uh you know make your level one technicians operate that much better. Uh, it's going to give you the ability so that your best technicians can be uh, you know, mavens, to borrow an old term about sort of uh a pre-influencer term.
SPEAKER_00Uh, but but to be a Maven and to uh you know provide all this information downstream to lower level technicians, that uh can, you know, right there, that's that's a huge problem, especially if you're working in sort of uh geographically disparate regions where your best technician is in the middle of nowhere, and then uh either they don't have a ton of communication with your lower level technicians. Now you end up in this space where your best technicians are putting all their knowledge in the Ninja One, and then Katana is relaying all that information to your uh your entire help desk.
SPEAKER_00And overall, that is just going to be something uh, you know, a tired metaphor, I know, but a rising tide that lifts all boats, uh, it's going to make your tech, all your technicians better. And whether that's your technicians, uh your level one technicians, it'll make them better at their job, and it'll make your most effective technicians better teachers and better communicators uh to the rest of your organization, which I think is a tremendous value.
SPEAKER_01I love that. You know, in sales and marketing, they have this term, it's called win-wire. This feels like an IT win-wire, right? For the whole IT staff to share. So I read through some of the technical documentation that um was sent to me before we sit down to chat, and we sat down to chat, and I they came across a couple terms and really liked them, and I wanted you to share what the thinking was behind them. So, Katana AI is what you folks are calling human-centered AI, and it's focused on pragmatic force multipliers rather than completely replacing human technicians.
SPEAKER_01Talk to me about what those terms mean to you organizationally, and what would you like the folks watching to know about them?
SPEAKER_00Yeah, I think uh the fundamental piece for me to communicate here would be that uh, you know, we don't believe that in endpoint management, in IT operations, that judgment can be substituted. AI can be great at surfacing information, uh, providing information to a technician. Uh, it can be great at making decisions. Uh, it can be great at implementing those decisions very quickly. Uh, I think where we see the difficulty here is about the very last thing you want to have happen in IT is making a bad decision fast.
SPEAKER_00And that is the recipe of uh AI in conjunction with the raw power of an endpoint management platform. There's so much you can do with the endpoint management platform. You have this AI that can make you know very rapid decisions, things can get out of hand very fast. And so we see that a human being being in the loop uh from the judgment layer, that's absolutely key towards effective and safe IT operations. Uh, and so that's why we talk about human-centered AI. We fundamentally, at our core level, believe that the human being is inseparable from endpoint management.
SPEAKER_00The day-to-day tasks may change, and we expect that they are going to change. They, in fact, have already changed rapidly in the last, I'd say, uh, three to three to four years. Um, but we expect that human beings still need to be involved there. There is still an intrinsic element of a human being being required to provide that level of judgment. Despite all the very impressive capabilities of AI, uh, I think we still recognize that there is an element of judgment there that is missing. And again, if we take that bad judgment uh or, you know, a failure in judgment during one critical moment, you end up in a situation where an AI thinks that the solution uh to a problem is to format the C drive.
SPEAKER_00You know, and without any data, there can be no problems whatsoever, right? Uh and so this is um, you know, we just fundamentally believe that human beings and endpoint management, those are not things that are going to be separated. The nature of the work may change. Uh, you may go from being more of a musician who plays an individual instrument to more of a conductor who kind of plays the entire orchestra. Uh, you know, the the role may change, but fundamentally, the judgment of human beings in the decision-making process uh for IT operations, I think, is is uh integral to being successful and will continue to be so.
SPEAKER_01Yeah, really important stuff. Um, it's great to keep learning with you. So thank you for taking time to visit today. And audience, thank you for joining. I hope you enjoyed this. I learned a lot, and we did as well. We'll see you next time.
Send us Fan Mail What if the next insider threat to your organization isn't a person at all? AI agents are moving beyond answering questions. They're accessing systems, executing tasks, requesting credentials, and...
Send us Fan Mail Your engineers may be using AI agents your security team doesn't even know exist. Those agents could be accessing sensitive data, handling company code, or operating with credentials nobody is...
Send us Fan Mail What happens when AI becomes capable of making decisions its creators can't fully anticipate or control? AI agents are gaining autonomy, accessing enterprise systems, managing credentials, and making...
Send us Fan Mail Your company has an AI governance committee. Great. But when an AI agent makes a decision that harms your business, who actually has the authority to shut it down? And here's a bigger question: Would...
Send us Fan Mail Your company wouldn't hire an employee without onboarding, supervision, or defined access. So why are AI agents being given responsibilities without those same safeguards? In this episode of ClearTech...