Discussion

AI skills matter at work, but most workers say nobody tests them

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#13296

Employers are hiring for AI ability without much formal evidence of who has it, according to a survey of US workers and managers. For people using AI in paid search, the practical takeaway is to document what they delegate, check and override, not just collect certificates.

Watch Desk analysis

What happened

Search Engine Journal’s Greg Jarboe reports that CodeSignal surveyed 1,000 US employees and 1,000 US people managers. Seventy-two per cent of workers said their AI skills had never been formally evaluated, while 69 per cent of managers called AI ability a major or moderate factor in hiring.

The article says the surveys were commissioned by CodeSignal and conducted online through Pollfish in September 2026. They measure respondents’ answers, not independently tested skill. CodeSignal sells skills assessments, a relevant interest to keep in view.

Key findings

  • Training does not always travel to the job
    While 62% of employees said they had received employer AI training, only 35% said that training was where they built the AI skills they actually use.
  • Workers say skills are hard to apply
    Only about one in four trained workers said they could apply most or all of their employer’s AI training.
  • Formal assessment is not the whole story
    Sixty-six per cent said their skills in general had never been formally evaluated, suggesting the gap is broader than AI alone.

Why it matters

The gap is not just between people who can and cannot write a prompt. In paid search, AI fluency can mean deciding what to hand to automated bidding, checking the results and knowing when to step in. A certificate may show familiarity with a platform’s playbook; it cannot show what happened when someone chose not to follow it.

The survey also suggests training can miss the everyday work it is supposed to improve. That matters when AI tools are making decisions inside campaigns and mistakes can spend a client’s budget. The figures are self-reported and come from research commissioned by an assessment company, so they are a useful signal, not a final scorecard for the workforce.

Our read

The article’s most useful advice is refreshingly unglamorous: keep a record of AI-related decisions. For each task, note what you delegated, what you checked or changed, and what happened next. Google Ads change history records changes, Jarboe notes, but not why you made them.

That gives workers specific evidence to discuss in reviews, and managers something more concrete than a certificate or a confident claim. It is not a perfect measure of expertise, but it beats asking people to prove good judgement with a job title and a prompt-writing badge.

What to watch

  • Whether employers start evaluating practical AI judgement, rather than treating familiarity as competence.
  • Whether training becomes more specific to the work people actually do.
  • Whether future surveys from outside assessment vendors find the same gap.

Discussion spark: Should employers formally test workers’ AI skills, or would reviewing real work give a fairer measure of judgement?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.