AI Can Do 40% of the Job. Here’s How You Win the Other 60%. How you should work with AI in 2027

AI Strategy · Careers

AI Career Advice for Students: How to Win the Other 60%

By Jane Chew, AI Strategy Coach 8 min read For students & fresh graduates
AI takes the routine 40%. The other 60% is where you get hired.

In this article: what AI is doing to entry-level jobs, what employers now look for, and the Human Margin Plan, three moves you can start this semester.

The research comes from Andrew Ng, LinkedIn’s Ryan Roslansky, the CEOs of HubSpot and Box, and three TEDx speakers who work with AI every day.

Quick answer

AI can do about 30-40% of the tasks in many jobs, according to Andrew Ng. The rest still needs a person, and that part is worth more now. Students who use AI for routine work, practise the human skills AI lacks, and show real projects in public will get hired ahead of those who only hold a good transcript.

A final-year student sends out forty internship applications and hears back from three. Her friends say AI is taking the junior jobs. Her lecturers say study harder. Both are half right.

The job market for new graduates is tougher than it has been in years. The people who hire, though, are saying something specific about who still gets in. This article puts their answers in one place.

What AI can do in your future job

Andrew Ng co-founded Google Brain and Coursera. He points to economists at Stanford and MIT who broke jobs down into individual tasks. Their finding: AI could do around 30-40% of the tasks in many jobs.

Ng draws the opposite conclusion from the headlines. When AI makes 30-40% of a job cheaper, the 60-70% a person still does becomes worth more. He doesn’t expect a “job apocalypse”. He does expect trouble for anyone who still works the way they did in 2022.

“Don’t do stuff that 30-40% of AI can automate. Gain new skills to do the other 60-70% that AI cannot do.” — Andrew Ng (paraphrased)

For a student, that 60% is the target. It includes judgement, client conversations, deciding what matters, and knowing your field well enough to spot when an answer is wrong.

What’s happening to entry-level jobs

The pressure on graduates is real. Some numbers from US hiring research:

  • The average open role gets 244 applicants, and a machine often reads your resume before a person does.
  • 43% of recent US graduates work in jobs that don’t need the degree they paid for.

Ryan Roslansky ran LinkedIn for six years. He says the skills needed for a given role have shifted by more than 25% in the last few years, and LinkedIn expects a 70% shift by 2030.

70% of the skills in a role are expected to change by 2030.

His advice is to ask a shorter-range question. Instead of “where will I be in five years?”, ask “what new skill will I learn in the next few months?” That question you can act on this semester.

What this means for Malaysian students: the junior role as it used to be defined is shrinking. Tasks like formatting reports, first-draft research and data cleaning are moving to AI. Employers still hire juniors who can do the thinking around those tasks.

The Human Margin Plan

Svetlana Makarova built AI systems at Mayo Clinic. Her line at TEDxJacksonville: “You can’t out-robot a robot.” AI learns from data. People learn from lived experience, and some human abilities have no training data at all. She calls that space the human margin.

The plan below has three moves that work together.

MoveWhat you doWhy it gets you hired
1. Hand AI the routineUse AI to do the task. Do your own thinking.You work faster and can still answer the hard questions.
2. Build human skillsPractise communication, trust, intuition and judgement.Five of the top skills in job ads are human skills.
3. Prove it in publicRebuild one process with AI and share what you learned.Employers can see how you think before the interview.

Move 1: Hand AI the routine work

Sol Rashidi was the world’s first Chief AI Officer. At TEDxStanford she told a story about two analysts. The junior one used AI and delivered a polished report in one week. The senior one took three weeks. Both reports looked excellent.

Then the leadership team asked questions about customer behaviour and market share. The junior analyst couldn’t answer. The senior analyst could, because she had done the research herself.

Rashidi’s team rule after that: use AI to execute your work, and do the thinking yourself. She calls the risk intellectual atrophy, the slow loss of your ability to think deeply when AI does it for you. Her fix is to treat AI output as a first draft. Editing is easier than authoring, and editing still makes you think.

Move 2: Build the human skills AI can’t copy

Emily Gill left a top law firm to teach these skills. She calls them human skills, because “soft skills” makes them sound optional. Her evidence:

  • In 2022, five of the top skills across 21 million job ads were human skills.
  • Targeted practice improved teamwork by up to 23% and communication by up to 15%.
  • These skills can be measured and trained like a sport.

Makarova names five human-margin capabilities you can practise:

Five skills AI has no training data for.
  1. Invention: ask “what if we did this completely differently?”
  2. Intuition: make decisions with incomplete data.
  3. Social acuity: notice what people leave unsaid in a meeting.
  4. Emotional courage: stay authentic when others play it safe.
  5. Sensory imagination: design how an experience feels for a customer, patient or guest.

Campus life gives you practice for all five. Lead a society event. Handle a difficult group member. Present to a panel. Work a part-time job serving real customers. Each one builds skills a model can’t learn from a dataset.

Move 3: Prove it in public

Roslansky sees your LinkedIn profile as more than a list of schools and jobs. Posting what you’ve learned shows the knowledge in your head. Hiring managers interviewed for the Silicon Valley Girl channel gave a three-step version:

  1. Take one process you already do and rebuild it with AI.
  2. Write about it publicly.
  3. Go one level deeper technically than your role strictly requires.

Want help with your first AI project?

Join the free DigitalAI community on WhatsApp. You’ll get practical AI tips, prompts and session invites from Jane Chew.

Join the free WhatsApp group

What employers say they hire for

HubSpot’s CEO names three things she looks for in every hire:

  • A scientist mindset: form a hypothesis, test it, and accept being wrong.
  • Curiosity to go deep in the work.
  • Customer orientation.

Box’s CEO adds a balance point. Domain skills such as product sense, customer interviews and market analysis still matter, and they outlast any one technology. AI fluency now sits on top of them in every role, including marketing, sales and finance. He wants people to understand what an AI agent is doing and how it works, even if they don’t code.

One hiring manager described the interview answer she wants to hear: “I’m not sure, but here’s how I tested it, and here are the two things I tried first.” In ten seconds she can see how the candidate thinks.

Your 30-day plan

Start small. One process, one skill, one post. Here’s a four-week plan you can run alongside your classes.

Four weeks, one process, shared in public.
WeekDo thisDone when
Week 1Pick one process from your studies, society or part-time job. Map its steps.You have a written list of every step.
Week 2Rebuild it with an AI assistant. Edit every output yourself.The new version saves time and you can explain each step.
Week 3Practise one human skill: present the result to a lecturer, team or manager and ask for feedback.You’ve had one honest feedback conversation.
Week 4Write a LinkedIn post: the problem, what you built, what you learned.The post is live and you can talk about it in an interview.

Two prompts to help. Both keep you doing the thinking.

Prompt 1: Find a process worth rebuilding
Prompt 1 · Process Finder
Act as a career coach for a university student in Malaysia.

Context: I study [YOUR COURSE]. Outside class I [YOUR SOCIETY / PART-TIME JOB / PROJECT].

Task: Ask me 5 questions, one at a time, to find a repetitive process I do every week. After my answers, suggest 3 processes I could rebuild with AI.

Structure: For each process, give the current steps, which steps AI could do, which steps I must still do myself, and the time I could save.

Goal: Pick one process I can rebuild in 2 weeks and talk about in a job interview.

Answer the questions honestly. The more specific your answers, the better the suggestions.

Prompt 2: Practise thinking out loud for interviews
Prompt 2 · Interview Coach
Act as a hiring manager interviewing me for a graduate role in [FIELD].

Ask me one hard question about [THE PROCESS I REBUILT WITH AI]. Don't give me the answer.

After I reply, score my answer from 1 to 5 on:
- clarity of my thinking
- whether I explained what I tested and why
- whether I showed what I did myself and what AI did

Then ask one follow-up question that goes deeper. Repeat 5 times.

The AI asks the questions, you give the answers. This trains the thinking that Sol Rashidi’s junior analyst was missing.

Frequently asked questions

Will AI replace graduate jobs?

AI will replace some tasks inside graduate jobs, mostly routine drafting, summarising and data work. Andrew Ng estimates AI can do about 30-40% of the tasks in many jobs, which makes the remaining human part worth more. Graduates who can do that human part, and use AI for the rest, will be the ones hired.

Should I still study computer science?

Yes, if you enjoy it. Andrew Ng’s advice is to work hard in class and also learn to build with AI through online courses, because universities take a year or two to update their syllabus. The same applies to any degree: master your field and add AI skills on top.

Which AI tool should I learn first?

Start with one general AI assistant such as Claude, ChatGPT or Gemini and use it every day on real coursework or part-time work. Box’s CEO says any one of the top AI productivity tools gets you most of the way there. Depth with one tool beats a shallow tour of ten.

How do I show AI skills without work experience?

Pick one process you already do, such as a club event, a group assignment or a part-time job task, and rebuild it with AI. Write a short LinkedIn post on what you changed and what you learned. Bring that example to interviews and walk the interviewer through your thinking.

What skills do employers want in the AI era?

Employers want domain knowledge, AI fluency and human skills together. HubSpot’s CEO hires for a scientist mindset, curiosity to go deep and customer orientation. In 2022, five of the top skills in 21 million job ads were human skills such as communication and teamwork.

Build your AI skills with people who use them at work

DigitalAI Business Club gives you practical AI playbooks, prompt libraries and live sessions on using AI for real business outcomes. Start with the free community, or join the Club for the full Strategy Vault.

Join the Club Follow Jane Chew on LinkedIn
AI strategy AI adoption AI readiness productivity DigitalAI Business Club

Jane Chew / AI Strategy Coach, HRDC Corporate Trainer / Founder of digitalaibusinessclub.com