A January 2019 article in Stanford Business included widely cited comments by Adina Sterling, an assistant professor of organizational behavior at Stanford Graduate School of Business. Sterling said that "hiring demands a global view of the company and its direction within a shifting market. Computers do not possess such a view." She also said that HR needs to "have the sense that they're held accountable for what algorithms are doing."

A lack of understanding of AI, problems with bias, a rapidly changing global environment and many other factors may mean that AI is suitable for specific functions, but not ready to be a part of many important HR roles.

Benefits decisions

AI can be a beneficial tool for the annual open-enrollment process by helping employees make complicated benefits decisions. AI tools in this capacity use each employee's answers to a set of questions, generating guidance based on logic and programming –and therefore without any emotions. For example, AI can help employees make decisions on:

• Life insurance and disability benefits: Most companies provide a core amount of coverage tied to a group policy. Many employers often offer additional levels of coverage. AI can help employees determine if they need more coverage based on responses to questions about other insurance coverage, medical conditions and cost.
• Medical dental and vision: Making complex decisions can require extensive data to make an informed election. AI creates decision-matrix solutions to aid in the process.
• Retirement planning: Similarly, AI can help employees make decisions using all relevant information, from current pay, current living expenses and other benefit choices.

Talent acquisition

AI is a potential solution to process large volumes of applicants' data as well as streamline and manage time-consuming tasks, but it also has many potential pitfalls.

AI already is being used in HR to automate recruiting and employee development. Many HR vendors have already released, or are about to introduce, products that review data, such as social media posts and cover letters, to help narrow down and determine which applicants are a strong fit for the position and culture.

However, AI can't be the only hiring tool. As Sterling also stated, using only AI can miss "outliers" who may not meet every programmed qualification but offer tremendous value to the organization.

Then there are issues relating to bias, including:

• Gender and race discrimination: There could be biases built into the programming that trigger responses based upon the programmer—which can be problematic because women and minorities are underrepresented in the technology field. Hiring and screening can be fraught with problems, including claims of discrimination. According to the Bureau of Labor Statistics, employment growth in computer science and engineering jobs is more than double the national average; however, women and minorities continue to be underrepresented. Last year, Amazon scrapped an AI recruiting tool after finding out it was eliminating female candidates.
• Age discrimination: AI in the workplace may also expose companies to potential violations related to age discrimination, as the law prohibits age-based discrimination against applicants or employees age 40 or over. The use of AI in the workplace to streamline certain activities could result in a disparate impact on an older workforce and potentially expose a company to discrimination claims.

Legislation may create other roadblocks. For example, in early August the Illinois legislature unanimously passed the Illinois Artificial Intelligence Video Interview Act and if enacted will regulate the increasing use of algorithms, so-called "interview bots," and other forms of AI to analyze applicants' facial expressions, body language, word choices, and vocal tones during video interviews. We expect other states will pass AI-related legislation in the future.

Best practices for using AI

Prior to implementing AI, employers should have a well-thought-out plan with mechanisms in place to manage it. Most importantly, employers should not rely exclusively on the AI results. They also should:

Document the process used to create the AI tool. Before implementing AI in any aspect of HR, carefully document the process, including the factors used in creating the algorithms.

Establish a review process of AI results. Implement a review process, such as full and false inclusion/exclusion tests of those who were selected and not selected. For example, 10 resumes selected for the next step in the recruiting process should be examined (if this is a statistically relevant data point) to determine if in fact they are worthy of the next step. Similarly, 10 resumes rejected should be examined to determine if the rejection was proper. It may be worthwhile to conduct this on a "blind" basis and not indicate which ones were rejected or accepted.

AI can save companies time and money and help employees with benefits decisions, but it shouldn't be the only tool. Hiring the right people takes more than data analysis, and efficiency should not be more valued than fairness.

Elliot N. Dinkin is president of Cowden Associates and a nationally known expert in compensation and employee benefits issues.

Read more: 

• HR capabilities falling short on enterprise goals
• Meet the new names in HR
• Banishing bias in the workplace–and in benefits