Into the Great AI Unknown We Go

There are a lot more employers exploring the use of artificial intelligence (AI) in their recruiting technologies, especially when it comes to screening candidates at the application stage. Anecdotally over the past two years, most employers that have participated in our annual CandE Benchmark Research have told us they don’t use it to screen.

However, this year when we asked more specifically what kinds of AI recruiting tech they’re using, the overall percentages jumped dramatically for screening applicants (see table below). This means “screening, matching, and/or ranking candidate applications for further screening or disposition”.

Overall in North America, 39% of participating CandE employers said they were, with a much higher percentage of large to medium companies (2,500+ employees) utilizing AI screening tech (in their ATS, CRM, etc.). 

In EMEA, only 19% said they’re using AI screening tech. In APAC, it’s 27%, and in LATAM, it’s 25%. These are smaller segments of global to in-country employers hiring around the world in our benchmark research.

32% of North America CandE Winners – those employers with near-to-above average candidate experience ratings in our benchmark research – also said they’re using AI screening tech. 

It is early days, though. Many of these employers have told me that they’re in the beginning stages of utilizing this tech. There are others that aren’t yet, but they are engaged with their legal counsel to understand the current landscape, their risk, and how best proceed. More on this below.

Utilizing this tech makes sense for those with increased applicant volumes, high-volume hiring, and leaner recruiting teams. In fact, multiple conversations with employers we work with have revealed they’ve already reduced their recruiting teams, or they will be. So, they have to be able to vet, implement, and optimize the best AI tools and technologies to help them find the qualified applicants that the humans can then screen further, interview, make offers to, and hire. 

Unfortunately, one of the top 3 negative comment sentiment themes in our global benchmark research last year and this year is the fact that candidates believe AI is screening them out. Even with the increased utilization of these tools, that’s still not true. Job candidates who are dispositioned at the application stage overwhelmingly believe it nonetheless. And today, this is one of the worst candidate markets I’ve seen in the past 25 years. 

During this year’s CandE Benchmark data reviews with participating employers, I’ve been saying that, “The AI train has left the station, and it ain’t coming back.”

Because it ain’t. And it’s going to be super messy across the recruiting and hiring landscape for the next few years while employers vet and pick solution providers to work with to help them mitigate the application screening conundrum. 

The realities are far from perfect. A great NPR TED Radio Hour podcast I recently listened to titled How companies use AI to choose who gets hired and fired highlights the risky inaccuracies we’ve already experienced in using AI recruiting and hiring technologies. Plus, we have pending litigation that includes:

• Mobley v. Workday, Inc. — Age / disability / race discrimination claims via AI screening
• Intuit / HireVue complaint — discriminatory AI in video / evaluation technology
• Sirius XM Radio lawsuit — Similar to Mobley vs. Workday

Here’s why HR and talent acquisition leaders are huddling with their legal counsel today:

• Liability for discrimination: Even if bias is unintentional, an AI system that screens out disproportionately more candidates from a protected class can pose exposure under Title VII, the ADA, ADEA, or similar anti-discrimination laws. 
• Vendor/agent liability: Employers may not avoid liability by pointing to a third-party vendor; courts can treat the vendor as an “agent,” making the employer responsible for outcomes. 
• Emerging regulation: States and localities have started imposing AI-specific requirements (e.g. bias audits, notice obligations) that go beyond traditional employment law. This is true in New York and California. 
• Opacity and explainability risks: Many AI tools are “black boxes,” making it hard to trace how decisions are made, which raises evidentiary challenges in defending against claims. 
• Reputational risk & candidate fairness: Beyond legal risk, poor transparency or perceived unfairness can damage employer brand and candidate trust.

We certainly understand that last one here at the CandEs. I don’t even play an employment lawyer on TV, but we’ve got to get this right. Application volumes have become untenable for larger organizations around the world; it’s just too easy for candidates to apply for multiple jobs at once. Recruiting teams have to have the tools and technologies that accurately help them screen for the most qualified applicants. Especially those qualified professional to management applicants who are struggling to find jobs in this super-rough market. 

Whether they’re using AI screening tech or not, most employers we work with have emphasized that “humans are making the final hiring decisions” – as they should. This is something that they’ve embedded in candidate communications, displayed on their career sites, and emphasized in screening calls and interviews. 

Recruiting is in the business of “no” unfortunately. We say no to a lot more people than we hire in the end. And where the business of no is headed depends on who you’re talking with: the providers, the employers (including legal counsel), and the candidates. Into the great AI unknown we go.

I hope we get it right. For every real person who wants and/or needs a job today and tomorrow.