TL;DR: Choosing the right skills to learn matters more than ever, because AI is already making some skills worthless. Most skills don't vanish completely, though. Each one has a routine layer that AI now handles and a judgment layer that's gaining value. Here are seven skills losing value right now, what to learn instead, and a quick test for your own.
Remember when the person who could unjam the office fax machine was basically a local hero? Nobody put "fax whisperer" on a résumé, but everyone knew who to call.
Technology has always slyly retired skills. What's different now is the speed at which this is happening. When I started this piece, the working title said these skills would be worthless in five years. Then I thought: Five years? That's generous. A few of them are close to being on life support.
If you’ll allow it, I’m going to be direct in this article. Some of what you’re good at is losing value, and pretending otherwise won’t help you. The good news is that nothing you’ve learned has to go to waste. The real challenge is figuring out how to adapt it.
Companies have been doing this for a long time. Nintendo made playing cards for most of a century before it ever made a video game. When digital cameras gutted the film business, Kodak ended up in bankruptcy court, while Fujifilm took what it knew about film chemistry and turned it into skincare and medical products. Both started with the same expertise. Fujifilm built a bridge from what it already knew to what the market needed next. That’s the goal for your skills too, so let’s look at what the research says, why some skills fade, and how to build that bridge.
What the research actually says
PwC's 2025 Global AI Jobs Barometer found that the skills employers want are changing 66% faster in the jobs most exposed to AI, up from 25% just a year earlier. Jobs that require AI skills also paid an average of 56% more than similar roles without them.
The shift shows up in hiring, too. Stanford researchers tracking payroll data found that employment of 22- to 25-year-olds in the most AI-exposed jobs is now 19% below where it would be if it had kept pace with less-exposed jobs. Experienced workers show no similar gap, so the entry-level tasks people used to learn on are going first.
Freelance markets moved even faster. A study in the Journal of Economic Behavior & Organization found that after ChatGPT launched, demand for freelance skills it could substitute, like writing and translation, dropped 20% to 50% compared with the expected trend. Demand for chatbot development, meanwhile, nearly tripled.
For what this means for your specific role, see our guide to will AI take my job or take the quiz below. For your skills, the next question is why some fade while others hold up.
Why skills lose value: the routine layer vs. the judgment layer
Here's the part most future skills lists skip. According to McKinsey Global Institute, more than 70% of the skills employers want today are used in both work AI can automate and work it can't. So most skills don't actually disappear. Instead, they split into two.
Every skill has a routine layer, the repeatable part AI can now do in seconds, and a judgment layer: deciding what's worth doing and whether the result is right. The routine layer is losing value fast, and the judgment layer is gaining it.
A skill's routine layer is most exposed when it fits one of these patterns:
- It follows the same steps every time. If you could write it as a checklist, software can follow it. Think invoice processing or formatting citations.
- "Pretty good" is good enough. When a decent result is acceptable and easy to check, companies automate first and fix later. Think routine email replies or summarizing a meeting.
- It's mostly remembering things. Recall is cheap now, so checking answers beats knowing them.
- It converts one format into another. Audio to text, one language to another, data to a chart. Conversion is exactly what AI is built for.
- It's tied to one tool's menus. When you can just tell software what you want, memorizing menus matters less.
McKinsey's analysis lines up with these patterns. It flags accounting, coding, invoicing, and SQL among the most exposed skills, while negotiation, coaching, and caregiving are likely to change the least. Writing and customer relations sit in the middle, where the way you use them is changing fast.
Okay, now let’s get into what you came here for. Below are seven skills where the routine layer is already losing value, plus what to do to combat losing job relevance.
7 skills AI is already making worthless (and the skills to learn instead)
1. Writing first drafts
Why it's fading: AI can produce a passable blog post, email, or product description in seconds. McKinsey found that writing and editing showed up less often in job postings across 134 occupations.
Learn this instead: Briefing, editing, and owning a voice. Someone has to decide what a piece should say, who it's for, and whether the draft actually sounds like the brand. Good editors also catch the confident nonsense AI loves to produce.
Where to start: Study a public voice guide like the Mailchimp Content Style Guide, then edit an AI draft until it matches.
2. Word-for-word translation
Why it's fading: Machine translation handles routine text well enough for many companies, one reason freelance translation demand dropped after ChatGPT arrived.
Learn this instead: Localization and machine translation post-editing (MTPE). Localization means adapting content for a culture, including idioms, humor, and what counts as polite. Post-editing means fixing machine output so it reads like a person wrote it. As a Brazilian who writes in English for a living, I can tell you that idioms are where word-for-word translation falls apart. Translate “break a leg” or “piece of cake” literally into Portuguese, and you get a threat and a dessert. A good localizer knows to say “boa sorte” or “moleza” instead.
Where to start: RWS offers a paid Linguistic AI certification through Trados that covers post-editing and AI translation.
3. Writing code from scratch
Why it's fading: AI coding assistants now write much of the routine code developers used to type by hand. McKinsey lists coding among the skills facing the greatest disruption, and the Stanford payroll research points to early-career software developers as one of the groups most affected.
Learn this instead: Code review, testing, and system design. AI-written code still needs someone to check it for bugs, security gaps, and shaky architecture. Knowing why code should work a certain way beats remembering syntax.
Where to start: Google's engineering practices guide to code review is free and shows how a giant engineering team does it.
4. Summarizing research
Why it's fading: Ask an AI tool to summarize 10 articles and you'll have something in a minute. McKinsey found that general science and research skills showed up less often in job postings across 140 occupations.
Learn this instead: Verification and question framing. AI summaries can mix real sources with invented ones, so the valuable skill is checking claims against the original evidence. Even better is knowing which question to ask in the first place.
Where to start: Digital Inquiry Group, formerly the Stanford History Education Group, has a free lesson on lateral reading, the technique professional fact-checkers use. It's built for classrooms but works for anyone.
5. Bookkeeping and invoicing
Why it's fading: Matching receipts, categorizing expenses, and sending invoices are classic same-steps-every-time tasks. McKinsey ranks invoicing and accounting among the most exposed skills.
Learn this instead: Financial analysis and advising. Clients want to know what their numbers mean. Can they afford to hire? Which product is losing money? What happens if costs go up 10%?
Where to start: The Corporate Finance Institute's Financial Analysis Fundamentals course on Coursera covers the basics.
6. Writing basic SQL and routine reports
Why it's fading: You can now ask a database a question in plain English and get a working query back. McKinsey lists SQL (structured query language) among the higher-exposure skills.
Learn this instead: Data interpretation. AI can pull the numbers. Someone still has to notice the strange spike, ask whether the sample makes sense, and explain the result to a room of people who would rather be at lunch.
Where to start: Khan Academy's free statistics and probability course covers the fundamentals.
7. Following a support script
Why it's fading: Scripted answers to common questions are exactly what chatbots handle best. Even so, Gartner predicts that by 2027, half of the organizations that planned major customer service cuts because of AI will abandon those plans, since the hard cases still need people.
Learn this instead: Complex problem resolution and conversation design. Conversation design means planning how a chatbot talks: what it says, when it asks a follow-up, and when it hands the customer to a human. Former support reps excel at it because they know where customers get confused.
Where to start: Google Skills offers a 90-minute Conversation Design Fundamentals course.
Will your skills become obsolete? A 5-question check
Want to check your own skills? Ask these five questions about any task you do regularly:
- Could you write the steps down as a checklist?
- Would a "pretty good" result be acceptable?
- Are both the input and the output digital?
- Does a person have to be accountable if it goes wrong?
- Does it depend on knowing specific people, places, or context?
If you answered yes to most of the first three, that task's routine layer is exposed, so start shifting your time toward the judgment work around it. If you answered yes to questions four and five, you're on firmer ground, and that's where to invest.
You don't need to start over
Look at the seven "learn this instead" skills again. None of them asks you to throw away what you know. The editor builds on years of writing, and the code reviewer builds on everything they learned writing code by hand.
Remember Fujifilm? When the film market collapsed, it kept its chemistry know-how and asked a better question: where else could this be useful? You can ask the same thing about your own skills. Everything you’ve learned so far is the bridge, and the new layer you add decides where it leads.
The professionals who adapt fastest treat their expertise as the foundation and add one layer on top, usually AI fluency plus sharper judgment. McKinsey found that demand for AI fluency grew sevenfold in two years, faster than any other skill in U.S. job postings. For a free, low-pressure start, try the University of Helsinki's Elements of AI course.
And if all of this feels like a lot, a bit of AI fatigue is completely normal. You don't have to learn everything at once. Pick one layer and add it to what you already know. For a look at which roles are most at risk, check out our list of jobs that are being replaced by AI.
FAQ
What skills will be obsolete in 5 years?
Few skills will disappear entirely, but the routine parts of many are losing value now, including writing first drafts, word-for-word translation, writing code from scratch, summarizing research, bookkeeping, basic SQL, and scripted customer support.
What skills should I learn for the future?
Focus on skills that sit on top of AI, like editing and briefing, verification, data interpretation, code review, financial analysis, and conversation design. AI fluency matters too, since demand for it grew sevenfold in two years, according to McKinsey. The best pick builds on expertise you already have.
What jobs will AI replace first?
Jobs made mostly of routine, rule-based digital tasks are most at risk. The World Economic Forum's Future of Jobs Report 2025 lists cashiers and ticket clerks and administrative assistants among the fastest-declining roles. Take a look at our guide to jobs that are being replaced by AI.
Your expertise is the foundation for what comes next. Explore how PowerToFly connects domain experts with AI work that fits what they already know.
