The short version
- Find a role with entry requirements you can realistically meet.
- Make a small sample showing your sources, a mistake and a correction.
- Keep applying while you learn; a portfolio is not a prerequisite for every job.
Start with a real opening and a small sample
- Week 1: read three local listings and choose a small relevant task to practise, such as making a visitor guide from a public event page. Use public or fictional information.
- Week 2: ask a practitioner what makes the task difficult and what a useful result would look like.
- Week 3: make a small attempt. Record preparation, checking and correction time, not just generation time.
- Week 4: write a one-page case study showing the source, mistake, correction and remaining limitation. Ask for feedback or a paid next step.
Make learning visible
In a study of 5,172 support agents at one firm, an AI assistant increased issues resolved per hour by 15% on average, with larger gains for less-experienced and lower-skilled agents. This was a staggered rollout, not a randomized trial or a study of every kind of junior work.[3]
Make a first attempt yourself. If you use AI, compare the draft with the original source. Keep what was wrong and how you corrected it. Ask someone familiar with the work for one specific improvement. You want to show learning, not just a polished result.
A June 2026 World Economic Forum and PwC framework likewise focuses on job access, work design, talent pipelines and education. It offers proposals for protecting early-career pathways, rather than evidence that those pathways will protect themselves.[5]
Look for demand, then check access
U.S. 2025–35 projections include strong growth in care, energy and data occupations. BLS cautions that its AI-exposure categories are not forecasts of worker replacement. Growth projections also do not tell you whether you can enter a role without further training.[4][6]
Keep applying while you learn. If income is urgent, put applications and public career support ahead of a long unpaid project. Ask employers about pay, entry requirements and who trains newcomers. If applications go nowhere, ask a practitioner or adviser whether the barrier is a missing skill, your evidence or a lack of openings.
What the research can—and cannot—say
Anthropic’s March 2026 analysis found no systematic unemployment increase in more AI-exposed U.S. occupations, but suggestive evidence of slower hiring of younger workers. That is an early warning, not proof that AI caused the change. The exposure measure partly reflects one vendor’s users.[1]
A Danish study revised in March 2026 linked adoption surveys to administrative records. It found task reorganization without substantial average changes in earnings or recorded hours in the first two years after ChatGPT. Early effects can differ across countries, workers and outcomes.[2]
Neither finding guarantees your first job. Employers can reduce openings, change responsibilities or improve training. A fair response needs employer investment in apprenticeship—not just advice that applicants should work harder.
A first office-support sample
- Before
- A beginner has an AI-written visitor guide, but cannot explain where its answers came from.
- Small trial
- They keep the public event page, check five answers, remove an invented cancellation rule and ask a tutor to review the result.
- What to check
- They share the corrected guide as a practice sample—not a client project—and explain what they learned in a relevant application.
Read the evidence.
The practical steps are our editorial recommendations. Numbered citations link research claims to their sources. Expand a source for its findings, methods and limits.
Labor market impacts of AI: A new measure and early evidenceMassenkoff & McCrory · Anthropic · March 5, 2026
Finding. No systematic rise in unemployment in more-exposed occupations; suggestive evidence of slower hiring of younger workers.
Limit. Association is not causation. One vendor’s usage is not all AI use, and aggregate stability can hide uneven effects.
- Type
- Vendor research · observational
- Period covered
- U.S. labor-market trends since late 2022
- Scope & method
- U.S. occupations, with exposure informed by observed Claude usage.
- Where to look
- Results on unemployment and hiring among young workers.
Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AIHumlum & Vestergaard · NBER Working Paper 33777 · Revised March 2026
Finding. Task reorganization accompanied near-zero average effects on earnings and recorded hours; estimates rule out effects larger than 2%.
Limit. Early Danish outcomes, not a global or long-run prediction. Working paper, not a randomized trial.
- Type
- Working paper · observational
- Period covered
- First two years after ChatGPT’s launch
- Scope & method
- Danish adoption surveys linked to administrative records; difference-in-differences analysis.
- Where to look
- Revised abstract and labor-market outcome estimates. DOI: 10.3386/w33777.
Generative AI at WorkBrynjolfsson, Li & Raymond · Quarterly Journal of Economics · February 4, 2025
Finding. 15% more issues resolved per hour on average; larger gains among less-experienced and lower-skilled agents.
Limit. Not randomized. Results vary by experience; experienced agents had small quality declines. One setting cannot establish economy-wide gains.
- Type
- Peer-reviewed · field study
- Period covered
- Earlier-generation assistant; not a 2026 tool evaluation
- Scope & method
- Staggered rollout to 5,172 customer-support agents at one business-software firm.
- Where to look
- Abstract and heterogeneous effects; QJE 140(2), pages 889–942.
Fastest growing occupations, 2025–35U.S. Bureau of Labor Statistics · August 27, 2026
Finding. Percentage growth and net new jobs highlight different opportunities; care occupations combine several large increases.
Limit. Not current vacancies, annual openings or an AI-immunity ranking. National medians do not establish local pay or access.
- Type
- Official projection
- Period covered
- 2025 baseline → 2035 projection; 2025 wages
- Scope & method
- U.S. national employment projections by detailed occupation.
- Where to look
- Table 1.3: employment change, percent and numeric. Employment figures in thousands are converted to jobs in our chart.
Artificial Intelligence and the Future of Entry-Level WorkWorld Economic Forum & PwC · June 22, 2026
Finding. Proposes responses across job access, job design, talent pipelines and education.
Limit. A framework for action, not causal evidence of layoffs or proof that a particular intervention works.
- Type
- Policy framework
- Period covered
- 2026 synthesis
- Scope & method
- Early-career pathways, education alignment and employer practices.
- Where to look
- Framework’s four dimensions and recommendations.
Artificial Intelligence exposure categoriesU.S. Bureau of Labor Statistics · August 27, 2026
Finding. Categories describe relative exposure, not predicted replacement.
Limit. BLS explicitly excludes interpreting these as forecasts of employment, wages, adoption or productivity.
- Type
- Relative exposure classification
- Period covered
- 2026 release
- Scope & method
- U.S. occupations classified using several theoretical and observed-use measures.
- Where to look
- Definition, methodology and interpretation cautions.
Reviewed September 5, 2026. Publication dates and periods studied can differ. Exposure is not job loss; projections are not outcomes. Reading time covers the main briefing; learning resources, source notes and linked research are additional.
Browse the research library