Your first job can still teach you in a workplace that uses AI, but learning needs to be part of how the work is organised. Look for feedback, access to the reasoning behind decisions and opportunities to check your own understanding. Producing more work quickly is not the same as becoming more capable.
This week's Hacker News discussion about graduates managing AI, alongside a larger discussion about a computer-science degree, reflects a real worry: if tools do the beginner work, where does a beginner get practice? The answer depends on the job, not a confident prediction about every profession.
What does the evidence say about beginners and AI?
It shows both useful assistance and reasons to pay attention to how entry-level work is changing.
The published Generative AI at Work study examined a customer-support deployment and found larger productivity benefits for less-experienced and lower-skilled workers. It also reported evidence of learning. That is evidence from a particular workplace, not a promise that every chatbot-supported job develops expertise.
Meanwhile, Stanford's revised 2026 employment analysis describes weaker employment patterns for young workers in more AI-exposed occupations, largely through hiring. The authors explicitly present these as descriptive indicators, not causal estimates of AI's effect, and do not find economy-wide displacement.
These findings concern different questions. An assistant may help a person already in a role while the route into similar roles becomes harder. Neither finding tells you to abandon a field from a headline. Both give you better questions to ask about the work available to you.
What should you look for in an entry-level job description?
Look for the decisions and responsibilities you will gradually learn, not only the tools you'll operate.
Descriptions that mention drafting, researching, checking or reporting can cover very different jobs. Ask what happens after a first draft is produced. Who reviews it? What mistakes are beginners taught to recognise? How does responsibility change as someone improves?
An employer may not have a polished training curriculum. That alone doesn't make the role poor. But someone should be able to explain how a new person learns to do the work well.
Be careful with vague promises of "exposure." Exposure to a busy team may mean useful access to decisions, or simply being nearby while other people make them. Ask for an example of something a recent beginner learned to handle independently.
Also look at the conditions you need to learn: a manageable workload, appropriate access and someone you can ask when the stakes are beyond your authority. A tool subscription can't supply those conditions by itself.
Which interview questions reveal whether you'll actually learn?
Ask about feedback and progression using concrete examples.
You might ask what a new hire usually gets wrong, how that is reviewed and what competent work looks like after the initial training period. Ask whether junior staff can observe a decision before being asked to implement it.
For AI-assisted work, ask who is accountable for checking the output. A sensible answer should fit the task's stakes. A marketing draft, a financial analysis and a production software change need different kinds of review.
You don't need to sound suspicious or demand access to confidential client work. Explain that you're trying to understand how you would become useful to the team. The interviewer gets a better picture of your approach, and you get information that a list of technologies won't provide.
Listen for examples rather than slogans. "You'll learn from everyone" is friendly, but "we review a draft together and explain what we changed" gives you something you can evaluate.
How do you use AI without becoming a person who only forwards its answers?
Keep responsibility for understanding the problem, checking the work and explaining the decision.
Before using a tool, write down what a successful result needs to do. After using it, compare the result with those requirements. What did you reject, verify or change? If you cannot explain that, ask for help before passing the work along.
Keep a private learning record that respects your employer's information rules. Record the type of mistake you missed and the check you will use next time. Don't take client data, internal documents or screenshots into a personal portfolio without permission.
Some practice should happen with reduced assistance where the task and your supervisor allow it. Try explaining an approach before generating a draft, or sketching a solution before asking for alternatives. This is a suggested learning routine, not a reason to ignore a team's approved process.
The goal is to become more reliable. Being fast at producing something nobody can safely use doesn't achieve that.
Can projects help before you get the job?
Yes, if they let you demonstrate judgement rather than merely display a finished output.
Choose a small problem and explain the decisions involved. Show what you tried, how you checked it and what remains weak. If AI helped, describe that contribution honestly. You can still take responsibility for work you directed and verified without pretending you wrote every line unaided.
For a design project, that may mean explaining which user need guided a revision. For an analysis, it may mean showing why a dataset could not answer the original question. For software, it may mean reproducing an error and explaining the fix.
Ask people in your intended field what a beginner's work sample should demonstrate. Keep the scope small enough that you can discuss it in detail.
An imperfect project with an honest explanation gives you something to learn from. A highly polished output that you cannot defend mostly gives the interviewer more questions.
What can you do in your first month?
Agree on a learning goal with your supervisor and connect it to a real piece of work.
Choose a task you need to become more independent at. Ask what good work looks like, complete an attempt within the approved process and request specific feedback. Use the next attempt to show what changed.
You can ask to observe how experienced colleagues check a result or handle an exception. Those moments help reveal the judgement hidden behind a quick answer.
If you aren't getting feedback, raise it early and specifically. "Could we review why these two changes were needed?" is easier to answer than "I don't think I'm learning." If the role consistently offers no suitable supervision or development, that is information to weigh when planning your next step.
A first job does not have to teach you everything. It should give you a credible way to become better at something that matters.
How can you identify a useful skill to build next?
For Grade 11 and older students, college students and early-career learners, Stride Ahead's Skill Gap Index can structure a first look at development priorities. Compare its suggestions with real work and feedback. Choose a practice goal you can observe, rather than treating a score as proof that you're ready for every task.
Sources
- Hacker News, graduates managing AI, 4 November 2026, and AI and a CS degree, 3 November 2026; audience discussion signals.
- Brynjolfsson, Li and Raymond, Quarterly Journal of Economics, Generative AI at Work, published 4 February 2025.
- Brynjolfsson, Chandar and Chen, Stanford Digital Economy Lab, Canaries in the Coal Mine?, revised 12 August 2026; descriptive working paper.
