Your AI Recruiting Tools Are Probably Missing Hidden Talent — Here's How to Fix That
The News
Speaking at SHRM26 in Orlando, Jacqueline Grant — founder and CEO of workforce development organisation The Management Academy — warned that AI recruiting tools risk overlooking an entire category of “hidden talent” that could otherwise fill persistent vacancies. This includes career switchers, military veterans, graduates of workforce programmes, and adult learners whose credentials and experiences don’t neatly map to traditional recruitment parameters.
Grant outlined a three-part review process TA teams should adopt: a visibility review to identify which candidates are being considered and why, an interpretation review to translate nontraditional experiences into business-relevant capabilities, and an employer confidence review to analyse which candidates advance versus fall out of the pipeline. “It’s really about seeing the value of the person, what they have to offer and what their experience might bring to the table,” she said.
The timing is acute: 93% of TA professionals planned to increase their AI use in 2026. Grant cautioned that human oversight remains essential, recommending that employers identify “human review trigger points” — moments in the hiring process where candidates reference nontraditional experiences or career paths — and ensure governance standards extend to AI and recruitment vendors alike. A 2025 University of Washington study found that human recruiters tend to adopt the biases of the AI tools they use, underscoring the risk of unchecked automation.
Why It Matters
This hits at a tension every TA leader is navigating right now: you’re being asked to fill roles faster with AI, but the pools you’re drawing from are shrinking. Applications fell 11% year over year according to iCIMS, and frontline roles saw an 18% drop. If your AI is filtering out candidates whose CVs don’t match conventional patterns — veterans who led logistics operations, adult learners with industry credentials rather than degrees, career switchers with adjacent expertise — you’re artificially narrowing a pool that’s already too small.
Grant’s framework is useful because it’s operational, not aspirational. Visibility, interpretation, and confidence reviews are audits you can actually run — not vague commitments to “inclusive hiring.” And the warning about human recruiters adopting AI bias cuts both ways: the tools aren’t neutral, and neither are the people using them.
Taking Action
All action is situated in your corporate context. These are starting points, not prescriptions — choose what fits.
Connect
Who to talk with in your enterprise
Your head of talent operations — schedule a walkthrough of your ATS screening rules and AI matching criteria. Specifically, ask: which keywords, degree requirements, or career-path signals are we using that might exclude veterans, bootcamp graduates, or career switchers before a human ever sees them?
Your diversity & inclusion lead — partner on a “familiarity trap” audit. Map the last 50 hires against the screening criteria your AI uses. If your tools are optimising for signals associated with past hires, they may be replicating homogeneity rather than expanding your pool.
Your legal and compliance team — review your AI vendor contracts and governance framework. Grant explicitly recommended extending governance standards to recruitment vendors. Ask: what bias auditing does each vendor perform, how often, and do you receive the results?
Inspect
What to take a closer look at
Run a visibility review on your last 30 days of rejected applicants. Pull all candidates who were filtered out at the screening or AI-ranking stage, then manually review 50 at random. How many had credentials or experiences that were relevant but didn’t match your standard keyword or degree requirements? Count them. That’s the rough size of your hidden talent gap. Specific tool to look at: your ATS’s automated screening rules engine — most platforms (Workday, SuccessFactors, iCIMS, Greenhouse) let you export screening rule configurations; review your knockout question logic and AI matching score thresholds.
Reflect
What to think about more deeply
The most interesting question isn’t ‘how do we make AI less biased?’ — it’s ‘if our AI tools systematically exclude people who don’t look like our past hires, and those people consistently prove to be strong performers when they do get through, what does that tell us about the actual predictive power of everything else we’re measuring?’