How can HR leaders avoid 'bad AI hires' in their organisation?
More than half of organisations have made a "bad AI hire" in the past year, according to a new report, as AI fluency takes over domain expertise in employers' hiring priorities.
A "bad AI hire" refers to an individual who is fluent in AI jargon and tools during the recruitment process, but is unable to apply this knowledge in their role, according to TestGorilla.
These bad AI hires emerge amid employers' shifting priorities, with hiring managers now preferring candidates with strong AI fluency over those with deep subject matter expertise.
But this shift alone isn't the problem, said TestGorilla CEO Wouter Durville.
"The right framing isn't AI skills vs. domain skills. It's AI skills applied to domain skills," Durville told HRD.
"Hire for the combination. The 'bad AI hire' problem is what happens when you optimise for one without testing for both."
Issues in the modern hiring process
TestGorilla's report identified three critical issues of modern AI hiring and recruitment processes.
The first critical issue is setting the minimum bar at tool awareness, with the second one leaving AI assessment entirely to the individual discretion of hiring managers without a shared rubric.
The last critical issue in the modern recruitment process is its current design where it observes communication skills among candidates, and not their ability to execute.
This means candidates are being hired for being able to speak fluently about putting AI workflows into practice, despite not ever having audited an output or redesigned a workflow in the past.
Avoiding the 'bad AI hire'
Taking on a bad AI hire could cost employers in lost output, failed projects, and the time and money to rehire, according to TestGorilla.
In order to distinguish candidates who are able to reliably apply their AI knowledge to their roles, Durville suggested the following measures:
- Reframe your questions by asking: How have they handled situations where AI got it wrong and what they did about it? Real fluency shows up in judgement, not vocabulary.
- Test for critical evaluation. Can the candidate spot a hallucination? Can they identify when an AI-generated output is technically correct but contextually wrong?
- Always include structured skills assessments in your hiring process that mirror real work. A candidate who's genuinely AI-fluent will outperform one who isn't on a task that demands it.
For employers who suspect they already have a bad AI hire with them, Durville said they could address the problem by starting with a skills audit.
"Alternatively, use objective assessments, if they weren't part of the hiring process, to establish a baseline now," he told HRD.
"It removes subjectivity from the diagnosis, creates a documented development plan, and — if the gap proves unresolvable — builds the evidence base you'll need to make harder decisions fairly and defensibly."
He also recommended mapping "actual AI touchpoints," including where a person is expected to use AI and how this person is expected to properly use the technology.
"Most 'bad AI hires' fail because no one defined the standard clearly at the point of hire. That's a systems problem as much as a people problem," Durville added.
The CEO also underscored that targeted upskilling is a much faster and cheaper approach to bad AI hires than replacement.
"But only when the gap is clearly defined," he said. "Blanket 'AI training' rarely moves the needle. Role-specific, output-focused development does."
