AI Can Screen Resumes, But It Can’t Screen for Culture Add

By Ejieme Eromosele

A 2024 University of Washington study tested three leading AI resume screening tools across more than three million comparisons. Same resume. Same qualifications. Same experience. The only variable was the name at the top of the page. AI tools favored white-associated names 85% of the time. Black-associated names were preferred less than 9% of the time. For Black male candidates, the AI disadvantaged them in nearly 100% of direct comparisons.

If you work in tech or SaaS and you hire CS or GTM talent, this is not a distant corporate problem. According to Deloitte’s Human Capital Trends report, 85% of recruiting processes in top tech firms now incorporate AI. The tools that produced these results are the same category of tools your company is likely already using, or evaluating.

And here’s what makes it harder to ignore: 67% of companies openly acknowledge their AI hiring tools could introduce bias. They know; yet adoption is accelerating anyway.

This is the quiet danger of AI in hiring: it doesn’t discriminate with intent. It discriminates with efficiency.

The Problem Isn’t New. AI Just Made It Faster.

Long before AI entered the picture, hiring was already shaped by a concept that sounded neutral but rarely was: “culture fit.” You’ve probably heard of the airport test. Would I want to be stuck in an airport with this person? It sounds harmless. But it was never really a measure of cultural alignment. It was a measure of familiarity. Of comfort. We naturally gravitate toward people who look, sound, and move through the world like we do. Left unexamined, that instinct produces homogeneous teams with matching blind spots.

When screening tools are trained on historical hiring data, who got hired before, whose resume got flagged as strong, they learn to replicate whatever patterns shaped those past decisions. A model trained on “successful hires” at a homogeneous company will keep selecting for homogeneity. This is not because it’s intentionally biased. It’s because it’s doing exactly what it was designed to do: find more of what already worked.

And removing names from resumes won’t solve it. The UW researchers found that AI can infer identity from educational history, geographic location, word choices, even membership in affinity organizations. The bias isn’t just in the name at the top of the page. It’s woven into the texture of how different people describe their lives and careers.

From “Fit” to “Add”: The Shift That Matters

Forward-thinking organizations have started replacing “culture fit” with “culture add.” The question shifts from “Does this person conform to what we already are?” to “What does this person bring that makes us stronger?” It’s a small wording change with a big implication. And the business case is clear. McKinsey research shows that companies with more than 30% women in executive roles outperform peers, and top-quartile companies for ethnic diversity outperform the bottom quartile by 36% in profitability.

But AI, in its current form, is not built for that question. It can screen for culture fit. It cannot screen for culture add. It cannot identify the candidate who will ask the question no one else thought to ask, or catch the blind spot the rest of the team shares.

So What Should We Do?

I’m not arguing against AI in hiring. I’m arguing for using it with intention.

For hiring managers: use AI to handle the administrative lift, scheduling, sorting, initial organization. But resist letting it make the cut. Before you open a search, define what “fit” actually means for the specific role, the team, and your company. Make it concrete and objective so that your criteria, not the algorithm’s training data, drives the decision. Build structured interviews with standardized questions. And ask yourself the harder question: are you hiring for what already worked, or for what could work better?

For candidates: you may be navigating AI screening before you ever reach a human. Optimize for relevant keywords and clear structure. But don’t let that be your whole story. Articulate your unique value proposition, the specific combination of experience, perspective, and skill that only you bring. When you do reach a human, make sure they understand not just what you’ve done, but also what you add.

The Blind Spot We Can’t Automate Away

AI will keep getting better at screening for culture fit. It will get faster, more sophisticated, more confident in its output. Our job, as leaders and hiring managers who care about building teams that actually perform, is to stay relentless about something AI cannot do: screen for culture add.

Want to put this into practice? Download the free Culture Add Hiring Checklist, a step-by-step resource for both hiring managers and candidates, based on the Fit Framework from The Customer Success Talent Playbook. Get it at cstalentplaybook.com.