The paradox: AI can do the work, but you need the people
IBM is tripling the number of entry-level positions in the US in 2026. Software developers, HR staff, precisely the roles everyone says AI can do.
CHRO Nickle LaMoreaux is clear about it: “The companies that will be most successful in three to five years are the companies that invested in entry-level positions now.”
She is not alone. AWS CEO Matt Garman called replacing juniors with AI “one of the dumbest ideas I have ever heard.” His argument: “How does that work if, in ten years, you have no one who has learned anything?”
| Company | Action in 2026 | Reason |
|---|---|---|
| IBM | 3x more entry-level positions | Retain talent pipeline |
| Cognizant | 25,000 starters (+20%) | Gen Z more productive with AI tools |
| Dropbox | Internship programs +25% | Leverage young people’s AI fluency |
This is not charity. It is strategy. Bringing in experienced people externally is more expensive and takes more onboarding time than letting your own talent grow. And those entry-level positions are not the same as two years ago, they have been rewritten. Less routine coding, more customer contact. Less answering standard questions, more checking and adjusting chatbot output.
What goes wrong when you replace juniors with AI
The theory sounds logical: AI does the junior work, you save salaries. In practice it goes differently.
Klarna is the best-known example. CEO Sebastian Siemiatkowski proudly claimed that AI did the work of 700 employees. The result: declining customer satisfaction, generic answers, more complaints. His conclusion in mid-2025? “We went too far.” Klarna started hiring people again.
Deloitte had to repay €290,000 to the Australian government. An AI-generated report turned out to contain fabricated academic references and a fake judicial quote. Not coincidentally, Deloitte had shortly before reduced the number of graduate positions by 18%.
The broader numbers are just as telling:
- 55% of companies that laid off employees because of AI regret it (Forrester Predictions 2026)
- Employees spend on average 4.5 extra hours per week correcting AI output (Asana)
- 77% of employees call AI tools unreliable
- 28% report that work was rejected by clients, 27% report security incidents due to AI errors
Forrester predicts that half of all AI-related layoffs will lead to “quiet rehiring” of the same roles. The message: AI without people who check the output doesn’t work.
Gen Z is already AI-native, and learns faster
This is where it gets interesting for your recruitment. Gen Z doesn’t use AI as an experiment. It is their standard way of working.
- 76-80% regularly use AI tools, versus 36% of Gen X and 20% of boomers (Deloitte/Gallup 2025)
- 93% of Gen Z knowledge workers use two or more AI tools per week (Google)
- 55.5% of early-career developers use AI tools daily (Stack Overflow 2025)
- On average 14 AI sessions per week, about 4.2 hours total
But it goes further than their own use. 62% of Gen Z employees actively coach older colleagues in AI tools (IWG 2025). That coaching works: 77% of directors say younger colleagues’ AI knowledge has improved their department’s performance. 80% say it has created new business opportunities.
That reverse knowledge transfer is worth gold. You pay a junior salary and get an AI trainer thrown in.
And Gen Z knows that AI skills are their trump card. The Class of 2026 lists AI skills on their CV twice as often as the Class of 2022 (Handshake). 59% see AI skills as required for career growth. 35% of applicants name AI adoption as the most important criterion when choosing an employer.
They choose you just as much as you choose them.
Hard learning is still hard learning
And here comes the nuance you won’t read in the headlines.
We know this at JumpScale firsthand. We have a Gen Z’er as co-owner. That speed of adoption is real, new tools, new workflows, new possibilities, it goes incredibly fast. But that doesn’t mean it happens by itself.
You have to develop critical thinking about AI output. You don’t build domain knowledge with prompts alone. The difference is that Gen Z isn’t afraid to start. They experiment, make mistakes, and learn from them. Where older generations want to read the manual first, Gen Z opens the tool and starts. That insight, what works and what doesn’t, is what makes the difference between vibe coding and responsible AI use.
But ultimately that applies to everyone who works with this technology, or has to start. The learning curve is there for a 22-year-old and for a 52-year-old. Gen Z has a lead in adoption, not automatically in understanding. That lead is valuable, but only if you give it structure.
Interestingly, Gen Z worries about this too. 79% fear that AI makes people lazy. 62% worry that it reduces intelligence. They want mentoring and structure, not just access to tools (HBR/Gallup 2026).
And only 9% of working Gen Z’ers feel fully prepared for AI in their work (Gallup 2025). They adopt quickly, but they also know they still have a lot to learn. That combination of daring and self-awareness is exactly what you’re looking for.
How to select for AI fluency when recruiting
This is where it gets concrete. If you hire juniors in 2026, select for AI fluency. 70% of employers already do this, usually indirectly (The Interview Guys 2026).
Meta goes furthest: there, candidates may use AI tools like Claude and GPT during coding interviews. The question is not whether they know AI, but how they work with it.
Five things to ask and check
1. Ask about their AI toolkit
Not “do you know ChatGPT?” but: “Which AI tools do you use daily? For what? What works well and what doesn’t?”
A strong candidate names specific tools for specific tasks. A weak candidate says “I use ChatGPT for everything.”
2. Ask about their AI-generated portfolio
Have they built a website with AI? Created social media content? Made a presentation? Let them show it.
It’s not about whether they used AI, that’s normal in 2026. It’s about how. Did they check the result, improve it, add their own style? Or is it copy-paste from ChatGPT?
3. Ask when they do NOT use AI
This is the real test. An AI-fluent candidate knows when AI is unsuitable. They recognize hallucinations, understand privacy risks, and know when human judgment is needed. We wrote earlier about the risks of AI-generated code; a candidate who knows those risks is worth gold.
4. Give a practical assignment with AI tools
Have them perform a realistic task with AI support. Don’t just look at the result, but at the process. How do they formulate prompts? How do they validate the output? How do they iterate?
5. Watch for these signals
| Green flag | Red flag |
|---|---|
| Names specific tools for specific tasks | ”I use AI for everything” |
| Knows limitations and risks | Trusts AI output blindly |
| Shows their own AI projects | Has no portfolio or examples |
| Asks about your AI policy | Hasn’t thought about it |
| Knows when AI doesn’t fit | Can’t name any limitations |
The lesson for organizations
You don’t have to be IBM to learn from this. The core is simple.
Stop thinking in terms of replacement. AI doesn’t replace people, it changes their work. The question is not “can AI do this?” but “who steers the AI and checks the output?” We wrote earlier about how to become AI-native as an organization; step 3 is “give your team permission to experiment” for a reason.
Start with recruitment. Add AI fluency to your selection criteria. Ask about tools, portfolios, and limitations. It costs you nothing extra and tells you a lot about a candidate.
Invest in training. McKinsey showed that one hour of prompt training is already enough to remove “prompt anxiety” and increase adoption. That is not a big investment. Accenture is investing $865 million in AI upskilling, PwC $1 billion, but for organizations an afternoon is enough to start.
Pair young with old. Let Gen Z bring in the AI knowledge, let experienced colleagues share the domain knowledge. That combination is more powerful than both apart. The AI Periodic Table helps you assess which AI complexity your team can handle.
Rewrite roles, don’t cut them. IBM did it: software engineers do less boilerplate and more customer contact. HR staff monitor chatbots instead of handling every standard ticket themselves. Look where AI takes over routine work and shift your people to work where AI still falls short.
The AI skills gap is estimated to cost the world economy $5.5 trillion in 2026 (IDC). 82% of leaders consider AI skills essential, but 60% of employees lack them. Closing that gap doesn’t start with technology. It starts with people.
Need help?
Want to know how to factor AI fluency into your recruitment and team development? We think along, from our own experience, not from theory. No sales pitch, just an honest conversation.
Sources: Fortune, IBM Tripling Gen Z Entry-Level Hiring (Feb 2026), Fortune, AWS CEO: Replacing Juniors “Dumbest Idea” (Dec 2025), Forrester Predictions 2026, AI Layoff Regret, Fortune, Klarna Rehiring After AI (May 2025), Deloitte, Gen Z & Millennial Survey 2025, CNBC/IWG, Gen Z Coaching Older Colleagues (Sep 2025), HBR/Gallup, How Gen Z Uses Gen AI (Jan 2026), IDC/Workera, The $5.5 Trillion Skills Gap, Handshake, Gen Z Hiring Trends 2026, Stack Overflow, AI vs Gen Z Developer Survey (Dec 2025)