You are a few semesters from graduating, your feed is full of people saying AI has eaten entry level jobs, and your university is still grading you on things nobody asks about in an interview. That gap is genuinely stressful, and most advice about it is useless, because it either tells you not to worry or tells you to learn thirty skills at once.
Here is the more useful truth. Hiring has not collapsed. It has changed what it screens for, and the change is actually in your favour if you know about it early. Employers have moved away from credentials as the first filter and toward evidence that you can do the work. Evidence is something you can build on purpose, starting this month. Grades and job titles are things you mostly cannot.
This guide is deliberately specific. Where it tells you to do something, it also tells you who to approach, what to say to them, what to build, and what the finished thing looks like.
Quick answer
Stop optimising for your transcript and start collecting proof. Employers increasingly screen on demonstrated skills rather than GPA, they expect you to use AI competently in your field rather than avoid it, and internship or real project experience is still the strongest tiebreaker between two similar candidates. Target four things: one deep skill, three pieces of real proof, honest AI fluency, and one real world placement. A plan for the next 12 months matters more than picking the perfect career.
On this page
Is this you?
Early years
You have time and want to avoid wasting it
Final year
Slightly panicked, want the highest leverage move left
Graduated
Applications stalling, and it is not your CV formatting
You do not need a technical background. Nothing here assumes you study computer science, and the worked examples deliberately cover a commerce student, a design student, and a software student.
What actually changed in hiring, and what did not
Start with what shifted, because most student career advice is still written for the hiring process of ten years ago.
The old process filtered first and looked second. A recruiter sorted by degree and grade, cut the pile down, and only then read anything. If your GPA was below the line, nothing else on the page mattered. The current process increasingly looks first: employers describe screening on skills, ask candidates to demonstrate them during interviews, and lean less on grade thresholds as the initial cut.
The old filter
Sort by degree and grade, cut the pile, then read what is left. Your transcript decided whether anyone saw the rest of you.
The current filter
Screen for demonstrated skill, then test it in the interview. Evidence decides, and evidence is something you can build.
That single shift is why this article exists. If the filter is evidence rather than credentials, then the highest value thing you can do as a student is manufacture evidence deliberately, and most students never think to.
70%
of employers report using skills based hiring, up from 65% the year before.
Source: NACE Job Outlook 2026
73% to 42%
the fall in employers screening graduates by GPA, from 2019 to 2026.
Source: NACE Job Outlook 2026
1 in 3
entry level jobs now require AI skills, nearly triple the share in fall 2025.
Source: NACE Spring Update 2026
Three things did not change. Employers still care a great deal about communication and teamwork: NACE’s Spring Update found that what they most want to see on a graduate CV is the ability to work in a team, solve problems, and communicate, not a list of tools.
Internships still matter enormously, and the data here is blunter than most students expect. When NACE asked employers what actually influences the choice between two equally qualified candidates, the top two attributes were both internships: having interned at that organisation scored 4.5 out of 5, and having interned in that industry scored 4.3. A high GPA came sixth, at 3.3, below leadership experience and general work experience. That single chart is the clearest argument in this article for everything that follows.
And being reliable and easy to work with still beats being brilliant and difficult. No survey needed for that one.
The genuinely new ingredient is AI, and the shift has been fast. More than a third of entry level jobs now require AI skills, nearly triple the share six months earlier, and close to 60% of employers say they are giving interns projects that use AI tools. Globally, the World Economic Forum names AI and big data, networks and cybersecurity, and technological literacy as the three fastest growing skills through 2030.
The reassuring part of the same NACE research is easy to miss. Just 11% of employers said they were discussing whether AI might replace positions, and more than half said AI is not reducing the tasks entry level workers perform. Their own summary is that AI is reshaping early career work rather than replacing it.
Key takeaway: None of this says “become an AI engineer.” It says AI is becoming a baseline expectation in ordinary roles, the way spreadsheet competence quietly became one decades ago.
That reframes the question from “which job is safe” to “what can I show.” So let us look at what an employer actually checks.
The three things an employer is actually checking
Underneath every application form, interview question, and take home task, a hiring manager is trying to answer three questions. Everything you build should feed one of them.
1. Can you do the work?
Answered by evidence
2. Can people work with you?
Answered by references and stories
3. Will you keep up?
Answered by a visible learning habit
Notice that a degree only partially answers the first question and barely touches the other two. That is not an argument against your degree. It is an argument for adding things around it that your degree does not produce on its own. Which brings us to the five things worth building.
The five things worth building while you are still studying
These are ordered by leverage, not by difficulty. If you only do the first three, you will still be ahead of most of your graduating class. Each one is something you can start without permission, funding, or a job.
1. Honest AI fluency in your own field
The mistake almost everyone makes here is treating AI fluency as a collection of prompt tricks. What employers are describing when they ask for AI skills is much more ordinary: can you use these tools to do real work in your discipline, and do you know when not to trust them.
Here is what that actually looks like in three different degrees, so you can copy the shape into yours.
| If you study | A real task to use AI on | The check that makes it a skill |
|---|---|---|
| Business or marketing | Draft five campaign angles for a real local shop, then have AI argue against each one. | Ask the shop owner which angle matches a customer they actually get. Usually two of the five are fantasy. |
| Biology, medicine, or law | Summarise a paper or a judgment you have already read, then compare the summary against your own notes. | Open the source and check every specific claim, number, and citation. Fabricated references are the classic failure. |
| Design | Generate twenty layout variations for a poster brief, then throw out eighteen. | Show three people the two survivors and ask what the poster is telling them to do. If they cannot say, the design failed. |
| Computer science or IT | Have AI write a function you could have written yourself, then break it deliberately. | Feed it empty input, huge input, and wrong types. Code that looks right and fails on an edge case is the normal outcome. |
| Accounting or finance | Ask it to explain a real statement line by line in plain language. | Recompute two figures by hand. Arithmetic delivered confidently and wrongly is common. |
The pattern in that right hand column is the whole skill. Anyone can produce a confident paragraph. The employable part is catching the paragraph that is confidently wrong, and being able to say how you caught it. In an interview, “I use ChatGPT” is worth nothing. “I use it to draft, then I verify numbers by hand and check every citation resolves, because it invented two references on me” is worth a great deal, and it is a sentence you can only say if you have actually done it.
Do
Use AI on work you already understand, so you can tell when the output is wrong. Keep a note of the errors you catch. That note becomes an interview answer.
Don’t
Collect prompt tricks for tasks you cannot judge. An impressive prompt on a topic you do not understand produces confident nonsense you have no way to catch.
If you want a structured way to build the checking habit, our guide on how to check AI answers before you trust them walks through it, and learning prompting for free covers the basics without paid courses.
Warning: Your university almost certainly has rules about AI use in assessed work, and they vary by institution and even by module. Read your own module handbook before you experiment. Being flagged for academic misconduct is a far bigger setback than being slightly slower at essays. Our explainer on whether AI content can be detected covers what those checks actually do.
2. One skill you are genuinely good at
Broad familiarity with ten things is not memorable. Being the person who is properly good at one thing is.
The test is simple: could a classmate describe you in one sentence without using your degree name? “She is the one who is good with data.” “He is the one who can actually run an event.” “She builds the working prototype while everyone else is still arguing.” If the only sentence available is “he is doing a business degree,” you have not picked yet.
Pick by intersection. Write down three things your field values, three things you do not resent doing for four hours straight, and three things you can produce visible output in. Something usually appears in all three columns. If nothing does, pick whatever you are already slightly better at than your classmates and push it for one semester. You can change later. Drifting for two years is the expensive option, not choosing wrong.
Tip: The skill only counts once someone outside your head can see it. “Good at data” becomes real the first time you hand somebody a chart that changed what they did next.
3. Proof of work, which beats a polished CV
This is the highest leverage item on the list and the most neglected. A CV is a claim. Proof is a thing somebody can look at.
Most advice stops at “build a project,” which is why most students end up building a to-do app nobody opens. Something only counts as proof when three things are true of it at once: a real person wanted it, you actually finished it, and you can explain the decisions you made. Miss any one of the three and it stops being evidence and goes back to being a claim.
Build this
Something one named person will use next week, small enough to finish, with a written page about what went wrong.
Not this
Another to-do app, a portfolio site about yourself, a clone of a famous app, or an AI wrapper with no user. All four are invisible because everyone has one.
Three finished artifacts is the target. One reads as luck, two as coincidence, three as a pattern, and a pattern is what a hiring manager is actually buying. They do not need to be big. Two weekends each is enough, and smaller is usually better, because the common failure here is not building something unimpressive, it is never finishing at all.
What none of that tells you is where to find the real problem in the first place, or what to say to the person who has it. That is where most students stall, so there is a full walkthrough further down: who to approach, the exact words to use, three worked examples, and the one page write-up that matters more than the thing you built.
4. One real contact with the working world
An internship is the strongest version of this, and if you can get one, prioritise it over almost anything else on this list. The evidence on internships as a hiring tiebreaker is unusually consistent.
But internships are competitive and unevenly available, so treat the underlying goal as the point: get one experience where someone outside your university depended on your work and would answer the phone about you afterwards.
Ranked by how hard they are to get, these all count:
| Option | How to get it | What it gives you |
|---|---|---|
| Formal internship | Apply 4 to 8 months early. Deadlines close far sooner than students expect. | The strongest signal, plus a reference recruiters recognise. |
| Paid freelance, even once | Turn one of your three artifacts into a small paid job for a second, similar client. | Proof someone valued the work enough to pay. Unusually persuasive. |
| Working for a lecturer | Ask the one whose subject you did best in whether they need help on anything. | The easiest reference to obtain, and often the most detailed. |
| Real responsibility in a society | Volunteer for the job nobody wants: the budget, the sponsors, the logistics. | Genuine stories about deadlines, money, and difficult people. |
| Volunteering with a small charity | Offer a specific skill, not general help. Specific offers get accepted. | Real users, real constraints, and a reference outside academia. |
Notice the pattern in the right hand column of that “how to get it” list: the specific offer wins. “Let me know if you need any help” gets ignored, because it makes the other person do the work of inventing a task. Compare the two:
Ignored
“Hello, I am a student looking for an internship. Please let me know if you have any openings.”
Answered
“I built a credit ledger for a hardware shop that cut month end reconciliation from forty minutes to five. Here is the one page write up. I would like to do the same kind of work with you for six weeks, paid or unpaid.”
The second message is only available to you after you have done item 3. That is why proof of work sits above this one on the list.
Tip: Start smaller than feels impressive. A two week placement with a reference who remembers your name outperforms a prestigious internship you did not get. And ask for the reference while you are still there, not six months later.
5. The ability to explain your work in plain language
You can do everything above and still lose the offer here, because everything you built has to survive the moment somebody asks you to explain it.
This is trainable and almost nobody trains it. Take one thing you have built and practise explaining it three ways. Using Example 1 again, so you can see the difference in length and depth:
One sentence, for the corridor
“I built a credit ledger for a hardware shop so the owner stops losing track of who owes him money.”
Two minutes, for the interview
The problem, what you built, the one number (forty minutes to five), and the one thing that went wrong. Nothing else.
Ten minutes, with the tradeoffs
Why a spreadsheet and not an app. What you did not build and why. What breaks if he gets fifty more customers. What you would do differently.
The question to rehearse against is not “tell me about your project.” It is “why did you do it that way, and what would you do differently.” That question separates people who built something from people who followed a tutorial, and it is almost always asked.
Practical drill: record yourself on your phone doing the two minute version. Watch it once. You will immediately hear the parts where you are describing features instead of explaining decisions. Redo it. Twenty minutes of this beats a week of CV formatting.
How to find and build your first real project
This is the section to come back to when you are ready to start. It covers the part every other guide skips: not what a portfolio should contain, but how you get from an empty week to one finished thing somebody actually uses. Work through it in order.
Who “someone else” actually is
You already know five or six people with an unsolved, repetitive problem. You have just never looked at them that way. Read this list and put a real name against at least three rows before you continue.
A teacher or lecturer
Rebuilds the same attendance sheet every term, chases the same missing submissions by hand, or sends forty near-identical emails about one deadline.
A parent, uncle, or cousin with a small business
A shop, a tuition centre, a pharmacy, a tailoring business, a delivery service. Almost always tracking money, stock, or customers in a notebook or in their head.
A campus society or student body
Collects event signups on paper, loses track of who paid, and remakes posters from scratch every single time.
A small shop or clinic near campus
No price list customers can read, no way to answer “do you have this in stock” without walking to the back.
A local charity or community group
Volunteer rotas by phone call, donation records across three notebooks, no simple way to show donors where money went.
Your own classmates
If forty of you reformat lab reports by hand every week, that is forty users with one shared problem, sitting in the same room as you.
If you genuinely know nobody: use your own course. Pick the task every student on it does manually and dreads, and solve that. Your classmates are real users, their complaints are real feedback, and nobody can accuse you of inventing the problem.
How to tell a problem is worth your two weekends
Not every annoyance is a project. Before you commit, check it against these five signals. Three or more and it is worth doing.
The five signals
✓ It repeats. Weekly or monthly, not once a year.
✓ They complain about it unprompted. You did not have to convince them it was a problem.
✓ It is done by hand right now. Paper, memory, or copy and paste.
✓ Getting it wrong costs something. Money, a missed deadline, an annoyed customer.
✓ They would notice if you stopped. The strongest signal of all.
How to actually ask, without it being awkward
This is where most students freeze, so here are the exact words. The ask that works is small, time boxed, free, and easy to refuse. The ask that fails is “can I do a project for you,” which sounds like you want a favour and puts the burden of inventing work on them.
Notice what all three scripts below have in common: you name a specific thing you have already observed, you ask for twenty minutes rather than a commitment, and you give them a clean exit.
To a teacher or lecturer
“Sir, I noticed you rebuild the attendance sheet by hand every term. I am trying to build one real thing this semester instead of another practice assignment. Could I sit with you for twenty minutes while you do it once, just to watch? If I can make it faster I will build it for free, and if it is not useful you can throw it away, no problem.”
To a relative with a small business
“Uncle, you keep saying you lose track of who still owes you money. I want to build one proper thing for my portfolio this month. Can I come on Sunday and just watch how you record it now? I will not change anything. If I can make something that helps, it is free, and if you do not like it you keep using the notebook.”
To a shop or small organisation you do not know
“Hi, I am a student at [university] and I am building one real project this semester. I am not selling anything and I do not want payment. I noticed customers keep asking whether something is in stock. Could I ask you two questions about how you handle that? Two minutes, and if you are busy I will come back another day.”
Expect some no’s. Roughly speaking, ask three people and one will say yes. That is a normal hit rate, not a verdict on you. Ask three in the same week so a single no does not stall you for a month.
The first conversation: watch, do not propose
When they say yes, resist the urge to arrive with a solution. Ask them to do the task while you watch, and time it. Then ask exactly four questions.
“How often do you do this, and how long does it take?”
This gives you the number you will quote later. “Forty minutes every Sunday” is a fact. “It saves time” is not.
“What goes wrong most often?”
The failure they name is usually the actual project. It is rarely the part you assumed.
“What have you already tried?”
Saves you from rebuilding something they abandoned, and tells you why they abandoned it.
“If I fixed only one part of this, which part?”
This is your scope. Build that one part and nothing else.
The two weekend rule: if your plan cannot be finished in two weekends, it is too big. Cut it until it fits. An unfinished ambitious project is worth zero, and half-finished work is the single most common outcome of student projects.
Three worked examples, start to finish
These are illustrative examples rather than case studies, written to show the shape of a finished piece of proof. Notice how small each one is, and notice that none of them required permission, money, or a job.
Example 1: a commerce student and her uncle’s hardware shop
The problem: he sells on credit to regular customers and records it in a notebook. By month end he cannot remember who has paid, so he stops chasing and writes off money he is owed.
What she built: one spreadsheet, deliberately laid out in the same order as his notebook so he did not have to change how he writes. One button-free monthly sheet that prints a per-customer statement he can photograph and send on WhatsApp.
Time: two weekends, plus one evening fixing it after he actually used it.
What she learned the hard way: her first version sorted customers alphabetically, and he could not find anyone, because he thinks in the order they visit. She had to rebuild the layout around his habit rather than her logic. That sentence is the most interview-useful thing in the whole project.
Example 2: a design student and the campus literary society
The problem: a different member makes each event poster in Word, so nothing looks related, and people keep turning up on the wrong day.
What he built: four editable poster templates plus a one page rule sheet covering which fonts, which two colours, and where the date goes. Handed over as a shared folder anyone in the society can use without asking him.
Time: one weekend for the templates, one week of small fixes as people used them.
What he learned the hard way: he showed early drafts to five people and asked what the poster wanted them to do. Three could not find the date. So the date became the second largest element on every template, above the society’s own name, which the committee initially hated.
Example 3: a software student and a family friend’s tuition centre
The problem: fee reminders go out manually over WhatsApp, always late, and the owner sometimes messages people who already paid, which embarrasses everyone.
What she built: a small script over a shared sheet that flags who is unpaid after the fifth of the month and drafts the message text for the owner to send herself. Deliberately no automatic sending, because the owner did not want a machine messaging her students’ parents.
Time: two weekends.
What she learned the hard way: the code was the easy part. The real problem was that payment dates were recorded inconsistently, sometimes as a date and sometimes as “paid,” so the first version flagged half the paid students. Most of the work was cleaning the data and agreeing one way to record a payment.
In all three, the interesting part was not the thing built. It was discovering the real problem was different from the stated one.
The write-up, which is worth more than the artifact
An artifact with no explanation is a screenshot. An artifact with a write-up is evidence of thinking, and thinking is what you are actually being hired for. Keep it to one page and use these five headings.
The one page template, with Example 1 filled in
The problem, in their words. “I lose track of who still owes me. Every month I give up chasing about ten customers.”
What I built. A one sheet credit ledger laid out in his notebook’s order, producing a printable monthly statement per customer.
Why I built it that way. He is not going to change a fifteen year habit for my spreadsheet, so the tool had to match the habit rather than the other way round.
What went wrong. My first version sorted alphabetically and he could not use it at all. I watched him fail with it for five minutes before I understood why.
What I would do differently. Watch him work before designing anything, not after. And ask what he does when a customer pays only part of the amount, which I did not handle and had to add later.
That “what went wrong” line is the one interviewers remember. A student who says “my first version failed and here is what I learned from watching a real person struggle with it” is demonstrating something a flawless tutorial project never can.
High signal proof versus low signal proof
Not all evidence carries the same weight. This is where a lot of student effort gets spent badly, so it is worth being blunt about the difference.
| What you have | Signal | Why |
|---|---|---|
| A finished project with a real user and a written explanation | High | Proves you can ship and reflect. Hardest to fake, so most trusted. |
| An internship with a reference who will speak specifically about you | High | Answers “can people work with you” better than anything else. |
| A GitHub, portfolio, or writing archive updated over time | High | Shows a learning habit, which answers “will you keep up.” |
| A tutorial project everyone on your course also built | Low | Proves you can follow instructions. Every applicant has one. |
| A long list of certificates with no output attached | Low | Certificates prove attendance, not capability. Fine as support, weak as evidence. |
| A CV listing tools and skills with nothing to look at | Low | An unverified claim. This is the default state most graduates are in. |
The pattern is straightforward. Anything that shows a real outcome, a real relationship, or a sustained habit is high signal. Anything that can be acquired by simply enrolling is low signal. That gives you a filter for the next twelve months.
Your next 12 months, depending on where you are
The plan changes a lot with how much time you have left, so find your situation rather than reading all four.
Which plan is yours?
Two or more years left: you have the luxury of depth. Months 1 to 6, pick your one skill and use AI inside your actual coursework, with nothing published. Months 7 to 12, do the three-people ask, build artifact one, and put internship deadlines in your calendar now, because they close far earlier than students expect.
Middle year: internships are the priority and they are time sensitive. Work backwards from application deadlines, not your exam calendar. Alongside that, finish exactly one artifact and write it up properly. One finished thing beats three in progress.
Final year: do not start a portfolio from scratch. Mine what you already have. Take your group project, your dissertation, your society role, or your part time job, and write two of them up using the one page template above. Then line up one reference and rehearse the two minute version. This is the highest return available to you now.
Already graduated, applications stalling: the problem is usually evidence, not formatting. Do the three-people ask this week, pick the smallest real problem you find, finish it within a month, and write it up. That single move changes what your applications contain. A CV rewrite does not.
If you are in your final year or already applying, our guides on using AI for your job search and optimising your LinkedIn profile cover the application side once your evidence is in order.
If you want this turned into something dated and specific, the free AI Learning Plan Generator builds a day by day path around your goal and the hours you actually have. For a longer structured start on the AI side, Learn AI in 30 Days is a reasonable on ramp.
The mistakes that cost students the most
A few failure patterns show up again and again, and each one is avoidable.
Collecting courses instead of finishing things. Courses feel productive because they are structured and end with a certificate. But a finished course is low signal and a finished project is high signal. If you have completed more courses than projects this year, the ratio is wrong.
Building for an imaginary user. This is the specific reason most student projects are worthless as evidence. If you cannot name the person who will use it and say when they will next open it, you are building a tutorial with extra steps.
Either avoiding AI entirely or outsourcing your thinking to it. Both are damaging, in opposite directions. Refusing to use it leaves you behind a baseline expectation. Letting it do the thinking leaves you unable to answer follow up questions about your own work, which is exactly where interviews probe. Also be careful what you paste in: unpublished work, personal data, and a shop owner’s customer list all deserve real caution. See whether your data is safe with AI tools.
Waiting until final year to start. Almost everything here compounds. Two artifacts built slowly across two years beat four rushed in a final semester, because the earlier ones let you apply for better things sooner.
Chasing whichever job title is trending. Title forecasts are unreliable, and a title that is hot in your second year may be crowded by the time you graduate. Skills and evidence transfer between titles. Titles do not transfer between themselves.
Optimising the CV instead of what the CV describes. Formatting matters at the margin. It cannot rescue an application with nothing behind it. Spend the hours on the evidence first.
The bottom line
Hiring got harder to fake and easier to prepare for. When the filter was credentials, you were largely at the mercy of institutions and grades. Now that more employers screen for demonstrated skill, the things that matter most are things you can build on purpose without anyone’s permission.
You do not need to predict the job market. You need one skill you are genuinely good at, three pieces of real proof, honest competence with AI in your field, one real contact with the working world, and the ability to explain all of it clearly. That set holds up regardless of which titles turn out to be growing in 2027.
Do this week. Not all of it, just the first line.
✓ Write down three people from the list above, by name.
✓ Send one of them the twenty minute ask, using a script above.
✓ Check your module handbook for the rules on AI in assessed work.
✓ Look up internship deadlines for the next intake.
✓ Write up one thing you have already built, using the five headings.
If you are heading specifically into software or IT, the picture there has particulars worth reading separately: our guide to AI and software development jobs in 2026 covers which roles are growing and how the junior end is changing. And if you want help choosing tools to support your studying along the way, the best AI tools for students groups them by the job you need done.
Turn this into a dated plan
Reading a plan is not the same as having one. The free AI Learning Plan Generator builds a day by day path around your goal and the hours you actually have.
Frequently asked questions
What if I do not know anyone with a business or a real problem?
Use your own course. Pick the task every student on your programme does by hand and dreads, such as reformatting lab reports or tracking group project deadlines, and solve that. Your classmates are real users, their complaints are real feedback, and nobody can say you invented the problem. A campus society or a small local charity is the next easiest option.
How do I ask someone for a project without it being awkward?
Keep the ask small, time boxed, free, and easy to refuse. Name a specific thing you have already noticed, ask for twenty minutes to watch them do it once, offer to build it for free, and make clear they can throw it away. Avoid asking can I do a project for you, because that makes them invent the work. Expect roughly one yes for every three people you ask.
Do grades still matter for getting a job?
Yes, but less as a filter than they used to. Some employers and sectors still apply a minimum threshold, and some graduate schemes are strict about it. What changed is that a good grade alone increasingly is not enough to differentiate you, and a mediocre one is less often fatal if you have real evidence. Treat grades as a floor to clear, not the thing you compete on.
Will AI take entry level jobs before I graduate?
It is reshaping them rather than uniformly deleting them. NACE found that just 11% of employers were discussing whether AI might replace positions, and more than half said AI is not reducing the tasks entry level workers perform. Globally the World Economic Forum projects 92 million jobs displaced by 2030 but 170 million created, a net gain of 78 million. The practical implication either way: roles where a human owns judgment and consequences hold up better than roles that are purely output production, so aim your evidence at judgment.
How big should a portfolio project be?
Small enough to finish in two weekends. If your plan does not fit, cut it until it does. An unfinished ambitious project is worth nothing, and half finished work is the most common outcome of student projects. A small tool that one named person actually uses beats an impressive prototype nobody opened.
Are certificates and online courses worth it?
They are useful for learning and weak as evidence. Use them to acquire a skill, then immediately build something with it, and put the built thing on your CV rather than the certificate. A certificate with an accompanying project is respectable. A list of certificates with nothing attached is not.
Should I mention using AI in my application or interview?
Generally yes, if you use it well and can explain your process. Employers increasingly expect AI competence, so hiding it is the wrong instinct. What matters is showing judgment: what you used it for, what you checked, and what you decided yourself. Do check the specific employer’s stance and any assessment rules, since practice varies.
Sources and references
Every figure in this article comes from primary survey research, linked below and checked against the source document rather than a summary of it. These reports are updated annually, so check the current edition if you are reading this a long time after publication. The three worked examples are illustrative scenarios written to show the shape of a finished project, not case studies of specific people.
NACE (National Association of Colleges and Employers), Employer Use of Skills-Based Hiring Practices Grows, January 2026. Source for the 70% skills based hiring figure and the fall in GPA screening from 73% in 2019 to 42% in 2026. Job Outlook 2026 data was collected from 183 employers between August and September 2025.
NACE, Demand for AI Skills in Entry-level Jobs Nearly Triples Since Fall 2025, April 2026. Source for the share of entry level jobs requiring AI skills, the 60% of employers assigning interns AI projects, and the finding that AI is reshaping rather than replacing early career roles.
NACE, The High-Impact Skills College Students Should Showcase on Their Resumes, 2026. Source for teamwork, problem solving, and communication as the attributes employers most want to see.
NACE, Job Outlook 2026 report (PDF), page 30. Source for the influence ratings when choosing between equally qualified candidates, where the two internship attributes rank first and second and GPA ranks sixth.
World Economic Forum, Future of Jobs Report 2025 (PDF), January 2025. Source for AI and big data, networks and cybersecurity, and technological literacy as the top three fastest growing skills, and for the projected 170 million jobs created against 92 million displaced by 2030.
About the author
Muhammad Adnan is the editor and publisher of TwistyApps, with around 17 years of experience in software development and DevOps. The hiring and development notes marked in this article come from his own experience evaluating and working with junior developers, not from survey data. Everything else is sourced above.
