Key Points
- The first questions about AI in IT operations are about control: what it can see, what it can change, and who decides.
- NinjaOne Katana AI recommends next steps but doesn’t act on its own. A technician approves every script or fix before it runs.
- Katana AI works with the same permissions as the technician or service account using it, so it only sees what that person can already see.
- Per-AI controls let IT leaders decide where Katana AI is active, down to the individual technician, and a global kill switch turns it off org-wide.
- Every recommendation and action is logged, including who made the decision.
- Katana AI is currently in Early Access.
Before an IT team brings AI into daily operations, a few questions come up. What can it see? What can it change? Who decides where it runs? And if something goes wrong, how quickly can you shut it off?
Those are the right questions, and they need specific answers. In part one of this series, we introduced NinjaOne Katana AI and how it helps technicians resolve issues faster. This post covers the controls behind it.
Access that matches your permissions
Katana AI works with the same permissions as the technician or service account using it. If a technician can’t see a device, an organization, or a client environment, Katana AI can’t surface it for them. The role-based access controls (RBAC) you’ve already set in NinjaOne carry through. For MSPs, a technician scoped to specific clients gets summaries and recommendations from those clients only.
Katana AI also stays within your data. It draws on your own NinjaOne environment, never a pool of data from other customers, and it won’t use non-validated web-based resources to answer questions about your specific environment. Your data isn’t used to train models outside the NinjaOne ecosystem either.
Technician approval on every action
Katana AI recommends next steps and doesn’t act on its own. Every suggested script or fix needs a technician’s explicit approval before anything runs.
AI can be wrong, which is why that approval step matters. A technician reviews each recommendation against what they know about the environment and the business before it reaches an endpoint. The technician stays the decision-maker, and Katana AI shortens the path to the decision.
Bring order to IT asset chaos
Choose where AI is active
Per-AI controls let IT leaders decide where Katana AI is active, down to the individual technician. You choose who gets access and can widen it as your team gets comfortable. For MSP owners, that means deciding which technicians use Katana AI across client work, with each one still operating inside the access they already have.
A record of every decision
Every recommendation and every action is logged. When a technician acts on a recommendation, there’s a record of that decision and the person who made it. If a client or leadership asks what the AI did and who approved it, you can show them.
A global kill switch
If you ever need to stop AI access, a global kill switch shuts off Katana AI instantly, platform wide.
How the controls fit together
Each control answers one of the questions from the start of this post.
- Permissions define what Katana AI can see
- Per-AI controls define where it’s active
- Technician approval defines what runs
- Audit logging records what happened
- The kill switch lets you stop access instantly
Useful AI for IT operations has to be built right, with controls in place before capability. That’s the order we followed with Katana AI, and it’s why IT leaders can adopt AI assistance while keeping control of access, approvals, and oversight.
Simplify remote IT management
AI that works within your rules
NinjaOne Katana AI brings your best technician’s judgement to everyone on your team, inside the guardrails you set.
NinjaOne Katana AI is currently in Early Access.
