NinjaOne has introduced NinjaOne Katana AI, an assistant built into its IT management platform that summarises device problems and suggests fixes, with every action gated behind three layers of control before a technician can approve it. For small firms that rely on an outside IT provider or a lean in-house team, that combination of speed and oversight is the part worth paying attention to.
NinjaOne has introduced Katana AI, an assistant built directly into its IT management platform, designed to summarise device problems and suggest fixes before a technician has to go digging through alerts and logs. The tool is reasoning over a customer’s own NinjaOne environment rather than generic data, is currently in Early Access, and NinjaOne has not yet announced general pricing or a full release date, according to NinjaOne’s own announcement.
The problem it is aimed at is a familiar one for any business that depends on outside IT support. An alert fires, and before anyone can fix anything, a technician has to work out whether it is a one-off glitch or part of something bigger. Small IT teams already juggling the basics often lose more time working that out than they spend on the actual fix, and resolution quality can depend heavily on who happens to pick up the ticket.
Katana AI is built to close that gap. It pulls together device health, recent activity and related past incidents into one summary, then recommends a next step, whether that is a script, a known fix, or a similar incident to check against. Technicians can also search their own environment in plain language, rather than building filters or knowing exactly where to look.
NinjaOne has built in three specific controls to keep that assistance in check: role-based access controls, per-AI permissions that decide exactly where Katana AI can operate, and a global kill switch that an IT leader can use to shut off access instantly across the whole organisation. Nothing it suggests runs automatically. A technician has to approve a fix or script before it executes.
That governance layer matters more than it might first appear. As agentic AI tools spread into day-to-day operations, deciding who controls what an assistant is allowed to see and do is becoming its own discipline, as seen in KnowBe4’s recent launch of Agent Risk Manager for governing Claude AI use. A kill switch and tiered permissions give an owner something concrete to ask their IT provider about, rather than taking reassurances on trust, particularly at a time when attackers are increasingly targeting the data sitting behind these systems rather than the systems themselves.
The clearest winners here are managed service providers (MSPs) handling several clients at once, and any small business whose IT support runs on NinjaOne already. For a newer or less experienced technician, getting the same starting context as a veteran could shorten the time a ticket sits open. For an owner who has never heard of NinjaOne, this is one more sign that AI assistants are becoming a standard layer in business software, alongside moves such as Google Workspace’s recent addition of Gemini across its apps.
If your business uses NinjaOne or a provider that does, it is worth asking two things on Monday morning: whether Katana AI is switched on for your account, and who holds the kill switch. Early Access means features and pricing can still change, so there is no need to act today, only to know the question is coming.








