A marketing coordinator wants to move into data work. A developer wants to try product management. A few years ago, both would have needed months of courses and side projects before anyone let them near the real thing. Now they can open a chat window, ask for a plan, and produce something that looks like it came from someone who has done the job for years.
That feeling is real, and it’s also where people get into trouble. “AI makes you feel like you can do anything,” Harvard Business School professor Iavor Bojinov told HBS Working Knowledge. His team set out to test whether the results match the feeling. So, can AI make you an expert? The best recent evidence says no, at least not by itself, and the gap they found says a lot about how to use these tools when you’re trying to move into a new field.
What the IG Group experiment tested
The study, described in a March 2026 Harvard Business Review article called “Gen AI Won’t Make Your Employees Experts,” ran inside IG Group, a UK trading and fintech company. Researchers from Stanford and Harvard Business School worked with 78 employees. Twelve were web analysts whose actual job is producing investing articles for the company’s site. Twenty-six were marketing specialists, people close to that work but not doing it. Forty were technology specialists, such as software developers and data scientists, who were far from it.
Everyone got the same AI model and the same assignment: plan an article of the kind the analysts publish, then write it. IG’s head of research supplied the scoring rubric for clarity and competence. The underlying paper is a Harvard Business School working paper, so treat it as solid early evidence rather than settled science.
Planning was easy, finishing was not
At the planning stage, AI closed the gap almost completely. Outlines and research notes from all three groups scored about the same, a little above 4 out of 5. The technologists held their own here, which the researchers put down to planning being structured, template-like work that engineers already think in.
The drafting stage told a different story. Marketers using AI scored 3.92, close to the analysts’ 3.96. The technologists scored 3.42, roughly 13% lower. Same tool, same access, same assignment. What they lacked was a feel for what a good investing article sounds like, which claims need a caveat, and when a sentence that reads fine is quietly off for this audience.
The researchers describe AI as giving outsiders the map. Knowing the route on paper is one thing. Walking it, with the mud and the missing signposts, still takes someone who has been there.
Distance from the job decides how far AI carries you
The most useful idea in the study is what the authors call knowledge distance. Marketing and web analysis sit next to each other. Both groups think about audience, tone, and what keeps a reader going. So when AI filled in the gaps, marketers knew which suggestions to keep and which to throw out. The developers had no such filter. They could take the draft, but they couldn’t reliably judge it.
That’s an important limit on something we’ve covered before, that beginners gain the most from AI. Both things hold. AI lifts newcomers quickly inside a job they already partly understand, because the tool hands them the habits of strong performers. It lifts them much less when the job itself is foreign, because there’s nothing underneath to check the output against.
Faster still counts
None of this makes AI useless for career moves. The speed gains in the study were large and applied to everyone. With AI, the planning phase took about 23 minutes on average instead of about 63. Drafting took 22 minutes instead of 87.
That time is worth a lot if you spend it well. The trap is filling it with more polished-looking output you can’t evaluate, which is exactly how an outsider ends up confidently shipping something a practitioner would catch in a minute. The better use is to put the saved hours into the part AI can’t give you.
How to build the terrain knowledge yourself
If you’re using AI to move into a new role, a few habits make the difference between looking capable and being capable.
Pick moves that are one step over, not three. The marketers did well because their existing experience overlapped with the target job. If you want to reach a distant field, plan a route through an adjacent role first, where AI can carry you most of the way while you pick up the rest.
Get your output graded by insiders. The technologists in the study had the same tool as everyone else and lacked the context to see what their drafts were missing. Find one person who does the job you want and ask them to mark up something you made with AI. Their red ink is the terrain knowledge the model doesn’t have.
Do some of the execution by hand. Let AI help you plan, then build a portion yourself before comparing it with what the tool produces. The comparison teaches you what good looks like in that field much faster than reading AI drafts ever will.
Study finished work. Read twenty real examples of the thing you want to produce, whether that’s product specs, investment notes, or incident reports. Notice what they leave out, since a practitioner’s judgment often shows up in the omissions.
Keep a short log of where AI was wrong. Every time a reviewer catches something the tool missed, write it down. After a month you’ll have a personal list of that domain’s blind spots, which is close to what experience actually is.
What a career change looks like to a hiring manager
Hiring managers know AI can make a portfolio look strong. What they probe for now is whether you can explain the choices behind it: why this structure, what you cut, what you checked and how. A career changer who can say “the tool drafted this, an analyst flagged two problems, and this is what those problems taught me about the field” sounds like someone climbing the curve rather than faking it.
Telling that story clearly takes practice. Openskill runs the most realistic AI voice interviews, with follow-up questions based on your answers, so a career changer can practice explaining what a reviewer caught, and what it taught them, before a hiring manager asks. Start from the job posting URL for the role you want, and the practice stays in the new field instead of drifting back to your old one.
To find the new-field roles worth aiming at, search current openings in Jobs on openskill and filter by country, work arrangement, employment type or date posted. Click “Prepare for this interview” on a role and the practice is built around that exact job. openskill is a complete career platform, and you apply on the employer’s own site.
AI will get you into the room faster. It won’t give you the years. The encouraging part of the study is that the gap shrinks with proximity and feedback, and both are things you can go and get on purpose this week.
Questions, answered.
Can AI make you an expert in a new field?+
Not on its own, according to a Stanford and Harvard Business School experiment at IG Group. AI helped every group plan the work equally well, but software and data specialists writing outside their field still scored about 13% below the analysts who do that job.
Why does AI help some career changers more than others?+
The researchers point to knowledge distance. Marketers, whose work sits next to the analysts' work, nearly matched them with AI because they could judge what the tool produced. People from distant fields could take the draft but could not reliably tell what was wrong with it.
How can I build real experience when I use AI to switch roles?+
Choose moves that are one step away from what you already do, and ask insiders to grade your AI-assisted work. Do part of the execution by hand, study real finished examples, and keep a log of every mistake a reviewer catches that the tool missed.
How can I prepare for interviews when I'm changing careers with AI help?+
Practice explaining the choices behind your AI-assisted work, including what an insider flagged and what you learned from it. Openskill runs the most realistic AI voice interviews and can start from a job posting URL, so you can practice with follow-up questions tied to the role in your new field.
Compare tools
Choosing where to practice?
See how Openskill compares with other interview practice tools, and how its job search and AI match score fit alongside realistic voice practice.