Is AI turning IT teams into power users, or is it making them more complacent?

It’s a fair question, and a growing number of IT leaders are wondering aloud. The concern is that as AI takes over more of the routine work inside IT operations, engineers are quietly losing the critical thinking abilities they need when things go wrong at 2 a.m., and there is no algorithm to bail them out.

Here’s the thing: AI isn’t the problem. It’s leaning into over-automation without guardrails.

Consider the calculator. When calculators became standard tools for accountants, nobody argued that math professionals would become obsolete. In fact, the opposite happened.

With arithmetic offloaded to a machine, accountants were able to focus more on the work that actually required a human, like analysis and strategy. AI in IT is the same dynamic, just at a larger scale.

AI is not an existential threat to IT expertise. Think of it more like a filter that separates the strategic thinkers from those who simply process tickets.

The Cognitive Offloading Trap

What happens when workers stop treating AI as a tool and start treating it as a substitute for thinking? Three recent studies looked at this and the findings:

An ANSI survey found that frequent reliance on AI tools may negatively affect critical thinking abilities. This is in large part due to a mechanism called cognitive offloading, the tendency to delegate mental effort to a tool rather than doing the thinking yourself. The study also found that younger workers between 17 and 25 showed the highest dependence on AI, while higher education seemed to provide a buffer against this effect.

As someone with a career in product management, I still manually build prioritized lists of epics for my roadmap. I know Claude can do this faster and prettier, but the three-hour exercise of gathering data, asking questions and moving things around forces my brain to learn what is happening across the teams. This is a great example of task management that on the surface feels remedial but is adding a deeper value.

A Microsoft study observed that the more confidence a worker placed in an AI tool’s ability to handle a task, the less critical thinking effort they applied to it. In other words, the smarter you think the tool is, the less you use your own judgment. This becomes a liability for IT teams managing complex, high-stakes infrastructure.

A KPMG study found that nearly 60% of employees admitted to making mistakes due to AI errors. Roughly half use AI at work without knowing whether it’s even permitted, while 40% knowingly use it improperly.

These findings should be a warning for IT management still defining their AI governance posture that intentional leadership is necessary. When no one sets the standard for how AI should be used, the culture will set it for them.

AI Does the Work, but Humans Do the Thinking

None of the research above suggests that AI is harmful to IT teams, but it does show that how teams use AI makes all the difference. The same AI tool can either make a strong engineer even more effective or reveal the limitations of a complacent one. The outcome depends on whether leadership is setting clear expectations for how AI is used. AI handles repetitive work that involves large amounts of data, while humans handle decisions that require context, experience and accountability.

Here are two examples from IT service management (ITSM) to illustrate what an effective use of AI looks like in practice:

1. AI that automatically corrects typos and formats knowledge base articles.

This might seem like a minor convenience, but it adds up to real impact. IT specialists can document solutions faster and communicate them more clearly, which means the next person who encounters the same problem can resolve it without opening a new ticket. This saves time and improves the quality of the knowledge base over time.

2. Sentiment analysis built directly into support ticket interactions.

As a conversation progresses, the tool is able to flag signals that a customer is frustrated. Some IT professionals were initially skeptical of this feature, but over time, many have found that it helped them catch their own tone before a situation escalated. This is a simple way to keep the customer experience front and center even during high-pressure incidents.

Both examples show how AI can help experienced professionals do their job with less friction.

The concern is what happens when teams skip past this kind of intentional adoption and start leaning on AI to make decisions it should not be making. Consider a team that auto-generates every response and defers every priority call to an algorithm.

On the surface, this can appear productive. But when an incident occurs that falls outside the patterns the model has seen, there is no substitute for an experienced engineer who knows what to do.

How Much Can AI Really Do?

Most of the AI features in ITSM today are aimed at helping engineers work faster and with better information. This can look like summarizing tickets, suggesting routing, flagging sentiment or formatting documentation. These are tools that support human decision-making without replacing it.

There’s a more significant shift underway toward agentic AI. Earlier AI features could surface recommendations for a human to act on, but agentic AI systems can take action directly. In an IT operations context, that means an AI agent could be responsible for things like restarting a service, scaling cloud services or executing the steps of a remediation playbook, all on its own without waiting for someone to click a button.

As we evolve from the LLM/ChatGPT era of the last two years to this new ‘Agentic’ era, the challenges will compound. With the rise of MCPs, team members will be able to hand off their personal access tokens to AI and then give them an instruction. “Claude - Restart that MS Server for me.” We’re well beyond the creation of content at this point and moving into actual agentic action. The need for governance and control on top of AI usage goes from nice to have to critical.

This is a major change to what AI can do, and engineering teams are right to approach it carefully. Some IT leaders aren’t comfortable with the idea of virtual agents handling critical infrastructure tasks. I spoke with a customer who questioned why they would want to hand over these responsibilities, because their IT operations professionals are the most critical, highest-paid, and hardest to hire roles in the company.

That reaction reflects reasonable concern about accountability. When something goes wrong in a fully automated remediation sequence, the organization still needs to understand what happened, why and how to prevent it next time.

Support Human Expertise

In the coming months, the issue will come down to how deliberately IT teams are using AI. If teams simply treat it as a shortcut, they will find themselves less capable over time. Teams that treat it as a productivity tool with defined boundaries will get faster, more accurate and better able to focus on the work that actually requires their expertise.

What does that look like in practical terms?

1. Using AI to consolidate and filter the flood of alerts that modern monitoring systems generate.

2. Automating the structured, predictable parts of workflows, like onboarding and access provisioning.

3. Allowing sentiment tools to stay attuned to the customer experience, even when the ticket volume is high.

4. Treating AI-generated outputs as a starting point to be reviewed and acted on by a human expert.

AI is simply reflecting something that has always been true: There is a big difference between IT professionals who solve problems and those who complacently push buttons. That has nothing to do with AI, it’s a preexisting dynamic.

Leaders must set clear expectations and guidelines for their IT teams' performance. Much like a calculator is not a crutch, AI shouldn’t be one either.

We ranked the best help desk software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit