Two weeks ago, OpenAI released ChatGPT to the public and the internet lost its mind. By December 5th, the system had hit 1 million users, breaking records for the fastest consumer adoption of any software in history. If you work in IT, you’ve probably already tried it, or you’ve watched someone else generate perfectly functional Python scripts in seconds while they complained it was too slow. The reactions in Slack channels and forums have been split somewhere between “this is the most important tool since Unix” and “this is the beginning of the end for my career.” The truth, as it usually does, lives in the middle.

What ChatGPT actually is right now matters more than what people are saying it is. We’re looking at GPT-3.5, a large language model trained on text up to April 2022. It can complete patterns, follow instructions, and generate text that sounds coherent on first read. It’s impressive. But it’s not conscious, it’s not superintelligent, and it’s not coming for your job next week. What it is, very specifically, is an autocomplete engine that got really, really good at being helpful. The distinction matters because it changes what you should actually do about it.

What ChatGPT Can Do Right Now

If you’ve been in IT for more than a few years, you know the feeling of writing the hundredth regex pattern to extract email addresses from a log file, or spending forty-five minutes drafting documentation for something you’ve done a hundred times before. This is where ChatGPT shines. It can generate boilerplate code, offer a starting point for scripts you’d normally copy-paste from Stack Overflow, and do it fast enough that you’re not losing time. When a junior admin asks you what’s wrong with their PowerShell script, you can paste the error into ChatGPT and see a reasonable explanation of what went sideways, usually with a fix worth considering. The system is excellent at explaining things that already exist in its training data.

Documentation is another win. ChatGPT can draft runbooks, create formatted procedures, and build out templates for incident response. You still have to fact-check every line and make sure it matches your actual infrastructure, but it handles the shape and structure in a way that saves real time. For regex, for Kubernetes manifests, for bash one-liners, for explaining why your certificate chain isn’t validating, ChatGPT is useful right now. It’s a tool that genuinely improves productivity on things that are pattern-heavy and have clear right answers.

What It Cannot Do

Where ChatGPT breaks down matters as much as where it works. The system cannot understand your specific infrastructure. It has no idea whether your environment is built on a hybrid cloud model or running entirely on-premises, whether your network is segmented by department or by security zone, or whether you’re running workloads on hardware from 2008. It can’t reason about your business constraints. When a security recommendation conflicts with availability requirements, when uptime matters more than some other property, when the “secure way” costs fifty thousand dollars but the “secure enough way” costs five hundred, ChatGPT can’t make that call because it doesn’t understand the judgment call itself.

More critically, it cannot handle nuance in incident response or security decisions. IT security is fundamentally about tradeoffs. You secure against some threats by accepting risk on others. You run MFA on everything, or you accept that some accounts aren’t protected that way because the business use case justifies it. ChatGPT sees policies, not context. It can explain what zero trust is, but it cannot tell you whether zero trust is worth the operational overhead in a small team with limited budget and stable workforce, because it doesn’t understand your team, your budget, or your stability. That judgment belongs to humans.

The system also cannot replace the relationships that make IT teams function. When you’re debugging a problem at 2am and the database team comes offline to help because you’ve built trust with them over years, that’s not something a language model generates. When you can get a business stakeholder to accept a longer incident resolution because they trust you to explain the tradeoff clearly, that’s a skill ChatGPT can’t teach you, only you can build it. The human element in IT is the thing that breaks down first under stress, and it’s the thing that holds things together when systems fail.

What This Actually Means for Your Career

The real threat isn’t that AI will replace IT workers. The threat is that workers who understand how to use AI fluently will outperform those who ignore it entirely, and also those who over-rely on it. The gap isn’t between humans and machines. The gap is between people who learn to work with these tools and people who don’t.

If you’re a systems administrator and you spend the next year ignoring ChatGPT while your colleagues use it to automate documentation, draft runbooks, and speed up routine troubleshooting, you will fall behind. The colleagues who learn to use it will ship more work in less time. They’ll have cleaner documentation because they had time to clean it. They’ll have better incident response playbooks because they could draft them faster and iterate. The productivity gain is real, and your competitors for promotions, raises, and better positions are going to use it.

But if you’re the person who uses ChatGPT to generate code without reading it, deploy configurations without understanding them, or make security decisions based on what the system suggests without thinking them through, you’ll fail in different ways. You’ll ship bugs that ChatGPT hallucinated into plausible-looking syntax. You’ll create security vulnerabilities because the system suggested something that works in isolation but doesn’t work in your environment. You’ll make yourself replaceable by someone who can do what you do, faster, without needing your judgment, because you’ve trained yourself out of the habit of thinking.

The winning move is the one that’s always worked: stay curious, learn the tool, use it where it helps, know where its limits are, and keep the parts of IT work that actually require you. You’re in an industry where the ability to learn new tools quickly has always separated the senior people from the junior people. This is just another tool.

The Broader Lesson

This moment also sits next to something else that happened in November. FTX collapsed in spectacular fashion, and we’re watching a company that had sophisticated technology and brilliant engineers go up in smoke because nobody was paying attention to judgment calls about risk and ethics. Technology without judgment doesn’t just fail; it fails at scale. That’s true for financial systems and it’s true for your infrastructure.

The people who will thrive over the next year are not the ones who decide ChatGPT changes everything about how they work. They’re also not the ones who decide it changes nothing. They’re the people who understand what changed, what stayed the same, and how to be more effective in a world where this tool exists. That person will understand when automation is safe and when it requires human reasoning. They’ll use AI to handle the pattern-heavy, low-risk work, and they’ll keep their hands on the lever for the decisions that matter.

If you want to think about what this means for your position in 2023, it’s worth circling back to some things we talked about when the industry was reacting to 2021. As we discussed in the specialist versus generalist post, the question of whether to focus deep or broad has always shaped career strategy. ChatGPT changes that calculus slightly. Generalists who can understand systems holistically will be better positioned to judge when an AI suggestion makes sense. Specialists will need to stay sharp on the judgment calls specific to their domain, because that’s where the tool can’t help. The hybrid approach, the T-shaped professional who goes deep in one area and understands the shape of many others, suddenly looks even smarter.

The broader industry is also in the middle of a reckoning about what 2022 taught us all. The 2021 recap covered the Great Resignation and the shift in worker power that happened across tech and IT. ChatGPT doesn’t erase that shift. Workers who learn to use AI fluently have power because they’re more productive. But workers who can’t be replaced by an AI assistant because they understand their infrastructure, their team, and how to think in ambiguity, also have power. The people who lose ground are the ones doing purely pattern-matching work that AI is now good at, who decide to compete with machines by trying to be more machine-like, and who choose not to understand the tools reshaping their industry.

What to Do Now

If you’re in IT and you haven’t tried ChatGPT yet, spend an hour with it this week. Paste in an error message and ask what it means. Give it a script you wrote three months ago and ask it to explain what it does. Ask it to generate documentation for something boring that you’ve always meant to write up. Get a feel for where it’s actually helpful and where it’s confusing or wrong. Assume it will make mistakes and learn to spot them quickly. This is how you get fluent with any tool.

If you’re in management, understand that your team is probably already experimenting with this, and they should be. The ones who feel nervous are the ones who are quietly trying it without telling you. The open conversations are better than the silent ones. Ask your team what they’re finding useful. Understand the security implications for your organization. Understand where code generated by ChatGPT makes sense (boilerplate, documentation, explanations) and where it doesn’t (core business logic, security-critical code, anything that handles sensitive data without human review).

Most importantly, don’t panic. Panic makes people choose between two equally bad strategies: ignoring the tool entirely, or trusting it completely. The tool is real, the productivity gains are real, the risks are real, and your career is long. You have time to get good at this. The people who are going to struggle in six months are the ones who made a strong ideological choice in December. Stay flexible, stay curious, and stay critical. That’s been the winning strategy for IT careers for decades. ChatGPT doesn’t change that rule.