Used properly, AI can help make the hiring process more objective, drawing on multiple data sources to increase precision of fit with indifference to volume of candidates in an even yet expeditious manner.

As AI makes powerful strides in leveling market bumps and assisting stretched TA teams, AI tools have understandably caught the attention of regulators. As of April 2023, employers in New York City will be required to conduct an annual AI bias audit of automated employment decision tools and publicly disclose related AI hiring metrics. Other state and local governments will likely follow suit.

Their goals are to create more transparency in how AI is used in hiring and reveal any apparent disparate impact. These are important goals—and here’s the good news: responsible, human-centric AI is easily up to the additional scrutiny. Well-trained, ethically validated AI can indeed make hiring more efficient, effective and fair, resulting in both a better experience and better outcomes for everyone, including candidates.

More Awareness Needed

Before AI’s promise can be fully realized, more education is needed. According to a recent survey of senior HR leaders responsible for talent acquisition (sponsored by my company), nearly three-quarters (73%) of organizations increased their investment in TA technology solutions in 2022, and 60% are using more TA tech solutions than they did pre-Covid. Yet, even as investments in TA technology increased, nearly half (46%) of respondents polled—all of whom were senior talent acquisition and HR professionals—didn’t know whether their companies were using AI! Humans are still doing many tasks that could be supported through AI, such as screening and interviewing. Greater awareness and knowledge will be needed for broad adoption of AI in hiring.

How To Use AI Effectively And Ethically

As stated above, not all AI is created equal. Organizations must take conscious steps to use AI effectively and ethically.

Begin by understanding your goals for AI in TA. Do you want to increase efficiency? Improve the quality of hire? Increase DEI? Identify candidates with deeper, more specific experience? Unless you have clear goals and deploy AI with intent, you’re unlikely to see significant benefits you can demonstrate.

Make sure you identify the data you need to collect so you can measure the outcomes. AI requires good data and lots of it. Work with your AI vendor to determine what data you need and how you can leverage automation to collect it. It’s also critical to ensure you can measure the effectiveness of AI in TA. Set benchmarks for your goals and then measure them over time to demonstrate results.

The next step is to ensure that you’re using AI ethically, and this requires strong vetting of your TA AI vendor. (Full disclosure: My company offers AI-enabled hiring solutions, as do many others.) First, make sure the platform only collects job-relevant responses from candidates. Don’t collect social media information, facial recognition data or speech pattern analysis (also known as prosody). And don’t neglect to disclose to candidates that you’re using AI, to gain informed consent. As regulations become more prevalent, gaining informed consent will become par for the course with AI-assisted TA. Establishing processes for obtaining it now will put your organization ahead of the curve.

Make sure the vendor’s AI has been validated to show that its assessments predict job and organizational outcomes. More than that, however, the AI must minimize adverse impact against protected classes. Require full technical reports from the vendor to describe the process they use for validation in addition to their results.

Vendors must regularly monitor their tools for bias. Bias can develop in tools over time, even if they are initially bias-free, so it’s critical to work with a vendor that regularly assesses validity and fairness.

Finally, don’t use black-box AI. Many AI algorithms produce results without giving any indication of how they arrived at those conclusions. The inner workings of complex algorithms must be at least somewhat transparent; otherwise, there’s no way to demonstrate that AI is not basing scores and recommendations on irrelevant relationships, biased factors or other unethical and, increasingly, illegal criteria. Use a vendor whose AI is explainable, because as regulations tighten and spread, the use of black-box AI will soon become completely untenable for TA.

Read the full report here