AI-Assisted Applications: How to Verify Job Skills Fairly

Start with the work that is changing

The International Labour Organization's research published on 20 May 2025 estimates that one in four workers globally is in an occupation with some exposure to generative AI. It identifies job transformation as the most likely overall effect, because many tasks still require human input. This measures potential exposure, not observed job losses or the proportion of applicants using AI.

The hiring implication is to examine tasks rather than rely on job titles. A customer-support employee might use AI to draft a reply but still need to identify an incorrect refund promise. A reporting assistant might generate a summary but remain responsible for checking totals against the underlying spreadsheet.

Describe those responsibilities before choosing an assessment tool. Broader staffing decisions can sit within a workforce strategy and planning approach.

References: [1]

Replace vague skill claims with a small work sample

The U.S. Office of Personnel Management describes work samples as exercises that resemble tasks performed in the job. Its guidance also cautions that this approach is appropriate for competencies applicants should already possess when entering the role, rather than activities the employer plans to teach after selection.

Keep the exercise focused. A Bangladesh-based data assistant could review a fictional spreadsheet containing inconsistent dates and duplicate records. A Gulf customer-service applicant could draft a response to a fictional delivery complaint. An international recruitment coordinator could prioritise interview requests across time zones.

Choose the example that matches the vacancy, then define what a successful response contains. Score accuracy, prioritisation and explanation separately. Keep the same core task and scoring criteria for comparable applicants. Avoid turning the exercise into useful unpaid client work: use invented records and a clearly bounded assignment.

References: [2]

State the AI rules before the assessment

Applicants should not have to guess whether translation, spellchecking, search or generative AI is permitted. Explain the allowed tools, expected output, time allowance and assistance candidates should disclose before they begin.

Keep human review and data protection visible

NIST's Generative AI Profile, published on 26 July 2024, identifies risks including confidently incorrect output, data privacy, harmful bias and over-reliance on AI. It is voluntary risk-management guidance, not a Gulf or Bangladesh recruitment law.

Our recommendation is to use AI-generated summaries as material for verification, with a named reviewer responsible for the assessment decision. Check important claims against the application and appropriate supporting evidence. Give candidates a route to explain discrepancies. Do not make rejection depend solely on an AI-detector score or writing style.

Use fictional assessment data. Avoid asking applicants to upload former employers' confidential documents. Restrict access to submitted work and follow the organisation's approved retention arrangements. Before sending candidate records to an external AI service, review its processing arrangements and obtain the necessary organisational authorisation.

References: [3]

Give applicants a clear preparation path

Candidates can use permitted editing tools while remaining accountable for every statement in their applications. Describe what you personally delivered, the tools you used and a result you can explain. Remove invented qualifications, projects and performance figures, even when they sound convincing.

Prepare to discuss how you reached an answer. A concise explanation of a spreadsheet correction can reveal more than an impressive list of software names. Ask about language requirements, device access and available adjustments in advance. Where Bangla and English are both relevant, employers should specify which part of the role each language assessment represents.

For wider preparation, read our guide to hiring international talent to understand the employer's cross-border hiring considerations.

A practical checklist for the next vacancy

Before launching an assessment:

• Identify the essential tasks and competencies required on entry.
• Select one short, realistic exercise using fictional information.
• Publish the permitted tools and assistance-disclosure rules.
• Agree scoring criteria before reviewing applicants.
• Explain language, equipment and accessibility arrangements.
• Assign a human reviewer and a discrepancy-review route.
• Set access permissions and retention arrangements.
• Review completion rates, candidate feedback and later job performance before expanding the approach.

Begin with one role and refine the process using evidence. Do not promise a reduction in hiring time or an improvement in retention until your own results support it.