AI-Powered HR: Why Data Governance and Auditing Are the Keys to Success

It wasn’t long ago that artificial intelligence was most often encountered in sci-fi movies stoking (misguided) fears of robots conquering the world. Today AI is a practical part of our everyday lives—and most business operations. Organizations have already started incorporating AI technologies to support some of their most important business needs, including automation, data analysis, and streamlining user experience for both customers and employees. 

Human resources can also benefit greatly from AI, specifically with regard to recruiting and hiring talent. And the need for help is looming large, with the job market primed to heat up again by summer. According to the New York Times, Bank of America expects the economy to gain an average of 950,000 jobs a month in the second quarter of 2021 as COVID-19-imposed restrictions loosen and more businesses reopen their doors. And hundreds of thousands of new applicants are going to be seeking those employment opportunities. Luckily, AI can make this task easier and more efficient, helping organizations evaluate candidates and make hiring decisions.

Marketing Technology News: MarTech Interview with Alexander Rublowsky, EVP Marketing at Controlup

But using AI for HR has its share of challenges. And bias and fairness in particular are serious concerns that need to be addressed before going full throttle. Remember when Amazon made headlines in 2018 after its résumé-screening algorithm penalized women applicants for many of its open roles? The issue there wasn’t the AI model itself but rather that the data reflected flaws that existed within the company. Since the majority of Amazon’s employees were male, the algorithm associated successful applications with male-oriented words. AI models are only as good as the datasets they’re trained on, and Amazon is a prime example of how easily bias can creep into the hiring process. 

The first thing to do is take a hard look at your data supply chain. Unlike Amazon, most companies don’t have the resources to build an internal AI system for HR; instead they rely on vendors and out-of-the-box AI solutions. While this is a great way for companies to get started, it’s also important to recognize that these systems are also imperfect: training data may reflect the vendor’s own biases or those of the third-party datasets used. But whether a company outsources AI or has built its own solution, it’s vital for technical and nontechnical stakeholders alike to understand the shortcomings of the technology.

Marketing Technology News: MarTech Interview with Steve Warren, CEO at Mapp Digital

Regardless of who’s powering the AI systems, all organizations must regularly audit them, review the data being collected to train the models, and train, retrain, and test the models constantly to check for biases. Data governance and auditing help ensure your company is practicing fair and responsible AI—they should be ingrained in HR systems from the onset.

And remember, AI should be used to augment, not replace, the decisions of skilled HR professionals. AI is a powerful tool and it’s getting smarter and easier to use all the time, but contrary to the aforementioned doomsday scenario of killer robots, it will always need to be managed by a human. Fortunately, HR professionals have the advantage of being trained to think deeply about factors like diversity, equity, and inclusion. Even without data science experience, they can spot when these are misaligned and take steps to address them in the hiring process and beyond. 

AI will change the way we work, live, and make decisions. But much like humans, the technology is flawed in many ways. Part of responsible AI is understanding its limitations and confronting them head-on. We’re just now scratching the surface of what AI can do, and it’ll be exciting to see how, when used thoughtfully, it will transform the way HR recruits, hires, and retains talent. 

Marketing Technology News: MarTech Interview with Phil Lore, Chief Revenue Officer at Affinity Solutions