Hiring managers in 2026 face a new challenge: virtually every candidate claims AI proficiency on their resume, but there is no standardised way to verify what that means. The gap between claimed and actual AI skills is enormous, and it is costing organisations real money in mis-hires and underperformance.
The Problem with Self-Reported AI Skills
Self-reported AI skills are deeply unreliable. Studies consistently show that people with the least competence in a domain tend to overestimate their abilities the most, a phenomenon known as the Dunning-Kruger effect. In AI, this is amplified because the technology is new enough that most people have no frame of reference for what expert-level usage looks like. A candidate who has used ChatGPT to write a few emails genuinely believes they are proficient because they have never seen what a skilled user can accomplish.
Certifications help but are insufficient on their own. Many AI certifications test theoretical knowledge rather than practical skills. A person can pass a certification exam about prompt engineering without ever having crafted a prompt that solved a real business problem.
Practical Assessment Approaches
The most effective approach to measuring AI skills in hiring combines several methods. First, use a standardised AI proficiency assessment that adapts to the candidate's level and measures practical competencies across multiple dimensions including prompt engineering, output evaluation, tool selection, and ethical awareness. Second, incorporate AI-focused work samples into your interview process. Give candidates a realistic business problem and access to AI tools, and evaluate both their process and their output. Third, ask behavioural interview questions about how candidates have used AI in their previous work, with follow-up questions that probe depth of understanding.
Building AI Skills Into Job Descriptions
The hiring process starts with the job description. Rather than listing "AI proficiency" as a vague requirement, be specific about what you need. Do you need someone who can build AI-powered automations, or someone who can use AI-powered tools effectively? Do you need experience with specific models or platforms? Do you need someone who can evaluate AI outputs in a domain-specific context? The more specific your requirements, the better you can assess candidates against them, and the more likely you are to attract people with the skills you actually need.