Employers want AI-fluent employees. Hiring them is the challenge

[SINGAPORE] Nearly every manpower survey has trumpeted the demand for artificial intelligence skills and fluency among potential jobseekers, with salary premiums over listings with no mention of such requirements.

Yet, finding and hiring the right talent with suitable AI skills appears to be even harder than before.

Companies looking to get an AI wunderkind – who can magically turn a few prompts into instant revenue or savings – would have to temper expectations.

A 2026 KPMG study showed that only 5 per cent of users in companies that have been using AI are sophisticated enough to drive value.

That is not even considering the hiring process that might produce more misses than hits.

A new survey by skill-testing platform TestGorilla, which polled 2,000 employers in the US and the UK, showed that 71 per cent of organisations have defined what AI fluency looks like for their teams, and 50 per cent have built internal measurement criteria.

But, 59 per cent have made a bad AI hire despite supposedly having done this groundwork.

Part of the issue lies in how organisations actually define AI fluency and skills, with some defining it as just using an AI tool.

Now with widespread adoption, using an AI tool can hardly be seen as a sign of proficiency. Rather, it’s the way someone uses an AI tool that will set them apart.

To avoid subjective evaluations of capability, there have been attempts to quantify different aspects of AI competency, such as Anthropic’s AI Fluency Index. The index scored AI interactions based on higher order skills such as description, delegation and discernment.

Anthropic’s research found that while 85.7 per cent of analysed conversations exhibited the description and delegation skills required for continuous iteration, users did not always exercise critical discernment.

When AI tools produce polished outputs, such as code, documents or interactive tools, user evaluation plunged.

In these scenarios, only 5.5 per cent of users would check the facts, 13 per cent questioned the reasoning and 15.8 per cent identified missing context for non-text output.

This behavioural gap is problematic because AI is agnostic to the actual results it is giving, equally able to do either a good or bad job faster than a human. This then hinges on the human user to critically evaluate the output, continually iterating to get the right results.

In the rush to tap AI, businesses will need to take a harder look at their interviewing and evaluation processes for jobseekers. With the multitude of prompting resources available, sounding like an expert user is easier than ever.

Hirers will need to explore more rigorous ways of testing AI proficiency, especially since typical interview processes reward showmanship over competence.

Often, applicants are evaluated based on how well they tell their AI fluency story rather than demonstrating practical execution.

Tackling a problem using AI takes iterations and constant evaluation by the user – intangible qualities that will not come across in a standard interview setting.

For companies to test this effectively, the interview format will need to shift from storytelling towards competency-based evaluation, which could involve applicants auditing AI outputs in real time.

In hiring for AI-fluent users, companies will do well to better define and test applicants on their capacity for human oversight, and other critical AI competency skills needed for their specific job functions.