I've worked in IT for over 20 years, and I've seen plenty of changes. But AI is the biggest shift I've faced yet. Like any new technology, it brings real benefits and real challenges.
That's especially true in training. Most of the students I work with are already using AI, and the rest plan to start soon. It's making them faster and more efficient. But it's also introducing a new risk, the quality of the answers their AI tools hand back.
This is where I think Jamf has gotten something right. Their own AI Assistant, built into Jamf Pro, takes a different approach than the general AI tools my students reach for.
AI Assistant can read, analyse, explain, and surface information from your Jamf environment. But it can't modify configurations, push policies, enrol devices, or take any action that changes your fleet's state. It's built to help you understand, not to act on your behalf.
What makes it useful is where its answers come from. When you ask it a question, it pulls from two places before responding:
Jamf's verified product knowledge base, and, if you've enabled it, your organisation's actual configuration data. Ask it why a device is out of compliance, and it looks at that device. Ask what a configuration profile does, and it reads that profile, not a generic one from a training document somewhere online.
It's also off by default. You choose which tool groups to turn on, for which products, and an admin can flip that switch at any time. Nothing happens without someone deciding it should.
Compare that to what I see in my training sessions. Most of the AI tools my students turn to have no idea what a Jamf environment actually looks like. They're trained on the internet at large, not on any specific instance, so they can only guess based on patterns they've seen before.
That guessing shows up in a few ways. Jamf Pro moves fast, new features ship, old workflows get retired, the interface changes and a general AI tool's knowledge is only as current as its training data. Ask it how to do something today, and it might describe how things worked a year ago.
The bigger risk is confidence without understanding. A student copies a script an AI tool suggests, it looks right, and they move on without knowing why it works or whether it still applies to their setup. A wrong answer that sounds plausible is often worse than no answer at all, because nobody stops to question it.
So where does that leave a student who's already relying on AI? Not by telling them to stop, that's not realistic, and it's not the point of training anyway. The real goal is making sure they have the skills to check what AI hands them, instead of taking it on faith.
That's actually what good Jamf training has always done, even before AI entered the picture. When a student learns how Smart Groups actually evaluate criteria, or how a policy's scope and triggers interact, they're not just learning to build one, they're learning to read one. That skill matters just as much when the thing in front of them came from an AI tool instead of a colleague or a support article.
I try to teach it as a habit, not a one-off check. If an AI tool suggests a script, can a student explain what each line does? If it recommends a scope, do they know how to verify it against their own environment before deploying it? That's the same critical thinking we'd want from any admin reviewing someone else's work. AI just means they're reviewing it more often.
The students who get the most out of AI tools aren't the ones who trust them blindly, and they're not the ones who ignore them either. They're the ones who've built enough hands-on Jamf knowledge to know when an answer is right, when it's outdated, and when it's confidently wrong.
To finish off, we should mention Jamf's AI Governance.
AI Governance gives your IT team visibility and control over the AI tools running across your Mac fleet. Discover which AI applications, agents, and MCP servers are active on your devices, deploy vendor-correct configurations for tools like Claude Code, Claude Desktop, and OpenAI Codex, and generate the audit-ready evidence your organisation needs, all from one place. Standardise model access, enforce authentication requirements, and scope what each agent can reach. AI Governance puts AI management right where you already manage everything else: the same Jamf workflow, applied to AI.
More details can be found here: https://www.jamf.com/solutions/ai-governance/