The case that made every law firm in America nervous started out pretty mundane. A man sued an airline, Avianca, saying a metal serving cart had injured his knee. His lawyers filed a brief. The brief cited a stack of earlier court decisions, and roughly half a dozen of them didn’t exist.
ChatGPT had invented them. It even supplied quotes and docket numbers. When the judge asked for copies, one of the lawyers went back to ChatGPT and asked whether the cases were real. It said yes. In June 2023 the court fined the two lawyers and their firm $5,000.
What gets me about this story is how ordinary the people involved were. These weren’t reckless amateurs, just experienced attorneys who treated a text predictor like a legal database because nobody had ever explained the difference to them.
That kind of misunderstanding is everywhere now, and it’s not limited to law. You don’t have to build models to work next to one. You do need a rough sense of how Generative AI produces its answers. It isn’t looking anything up. It predicts what a plausible response should look like, based on patterns in its training data. Plausible and true often overlap. Sometimes they don’t, and the tool won’t warn you.
“I use ChatGPT” doesn’t count anymore
Everyone types prompts these days. My neighbour uses one to write birthday messages.
What employers increasingly want is something harder to see: knowing what an answer is good for. Can you rely on it, or treat it as a rough first pass? Where is it likely to go wrong? And who checks it before it lands in front of a client, a patient or the board?
The hiring side is moving quickly. In Microsoft and LinkedIn’s 2024 Work Trend Index, 66% of business leaders said they wouldn’t hire someone without AI skills. That’s two out of three.
Regulators got there too, at least in Europe. Since February 2025, Article 4 of the EU AI Act has required organisations that provide or deploy AI systems to make sure their people have a sufficient level of AI literacy. It’s a legal duty now, sitting with the employer, not a nice-to-have for whoever runs IT.
So how do you actually get there? For most people, it’s unglamorous: a few short courses, some internal training, and plenty of hands-on use where you deliberately try to catch the tool out. Some people need more than that, though. If you already have a technical background and expect to lead AI work (setting direction, owning projects, being the one who answers for them), a formal qualification can help. An Executive Master’s in Artificial Intelligence like New England College’s combines machine learning, deep learning and natural language processing with the ethics and practicalities of running AI projects inside real organizations. One caveat: it expects a technology degree or a tech role going in, and it includes in-person residencies, so it isn’t a weekend course.
No one is asking managers to learn Python. The point is that people who make decisions about AI should understand what they’re deciding.
Same habit, different jobs
The habit itself is simple. Before you act on what the machine says, ask whether it deserves to be believed. What that looks like varies a lot from job to job.
In healthcare, the stakes are obvious. AI already drafts clinical notes, handles scheduling, helps with research and sorts through data. If a system flags a patient as high-risk, somebody should be asking what data trained it, what it was really designed to predict, and whether this particular patient looks anything like the population it learned from. Bias and validation sound like ethics-seminar words. On a hospital ward they’re practical questions, and administrators run into them as often as doctors.
Finance has lived with algorithms for decades, and that history can breed complacency. “We’ve always used models.” Sure, but a model that summarizes 200 pages of filings in thirty seconds can bury a bad assumption in very confident prose. The analyst worth keeping is the one who asks what’s underneath the forecast instead of admiring how fast it arrived.
Marketing changed almost overnight. Copy, images and endless ad variants now cost close to nothing to produce, which sounds like a gift until you realize your competitors have the same tools. When everyone can make more content, more content stops being an advantage. What still counts is taste. Know the audience, keep the brand sounding like itself, check the claims and the rights to every image, and cut most of what the machine produced. The first draft is a draft.
Manufacturing is where AI stops looking like a chatbot. It sits inside vision systems on inspection lines, maintenance schedules, robot cells and supply plans. Picture a predictive-maintenance model telling a plant to delay servicing a press for another month. The data scientist can explain why the model thinks that. The shift supervisor who has run that press for fifteen years might know it always starts running hot in August. You want both of them in the room.
Education has the oddest problem of the lot, because AI changes how students learn and what they’ll need once they leave. An essay a chatbot can write in nine seconds isn’t much of an assessment any more. The better question for teachers is what’s worth teaching when routine writing is free, rather than whether a student cheated. The likely answers are reasoning, weighing sources, building an argument, and being able to say “that’s wrong, and here’s why” when a polished answer is off. Teachers who understand the tools can build lessons around those skills instead of spending the year playing detective.
And then there’s law, consulting and accounting, which brings us back to the airline case. AI drafts and summarises at a ridiculous pace, and it gets things confidently wrong. The professional’s name is still on the work. In practice, AI literacy here means knowing which outputs must be verified line by line, keeping client data out of tools that shouldn’t have it, and having a review process for AI drafts that actually works. A memo that takes five minutes to draft and an hour to check can still be a win. A memo that takes five minutes and nobody checks is the airline case waiting to happen.
A word to the people signing the cheques
Executives don’t need to follow the maths. They can’t hand the whole subject to IT, either.
AI decisions involve budgets, headcount, privacy, security, legal risk, customer experience and, eventually, the company’s reputation. Those decisions belong to leadership whether leadership likes it or not. A leader with even basic AI literacy can tell a real use case from a slick vendor demo, mostly by asking a handful of uncomfortable questions before signing off.
What problem does this actually solve? Whose data does it run on, and are we allowed to use it that way? How will we know in six months whether it worked? What happens on the day it gets something badly wrong? And when that day comes, whose name is on it?
Skip those questions and “AI strategy” usually turns into a drawer full of subscriptions nobody can justify at the next budget review.
The twist: your experience matters more now
There’s a popular fear that AI makes years of expertise pointless. Mostly, the opposite is happening.
The machine produces answers cheaply. Judging those answers is the expensive part, and it takes the experience people were afraid of losing. An architect spots the AI-suggested cantilever that couldn’t possibly stand. A supply-chain manager notices the plan assumes ships never wait at port. A physician catches the context the model never saw. A decent editor reads three elegant paragraphs and realises they say nothing at all.
Look at the Avianca lawyers again. They knew the law. What they lacked was enough understanding of the tool to realise they needed to apply that knowledge to its output. Their expertise sat idle because they didn’t think it was needed.
So if you’re wondering what to do about all this, don’t abandon your field to chase every shiny release. Go deeper in what you already know, and add enough AI understanding to see how it’s reshaping that work. My bet is that the people who do best over the next ten years won’t be the pure AI specialists. They’ll be the nurses, engineers, accountants and teachers who know their field inside out and can tell, on sight, where the machine is helping and where it’s quietly making things up.
Remember when “good with computers” was something people put on a CV? Nobody writes that now. It’s simply assumed. AI literacy is heading the same way, and fairly fast. If you understand the systems moving into your profession, you get a say in how they’re used. If you don’t, other people decide and you live with the results.
Related: Will AI Replace Executive Assistants? What Changes in 2026
