Key Takeaways:
For facility managers struggling to fill skilled positions, a stack of applications should be good news. Until it’s time to sort through the resume pile.
Enter artificial intelligence. AI can turn a stack of 200 job applications into a shortlist of 10 in seconds. But what if the best candidate is in the 190 applications the tool rejects? And what happens if the technology shows bias as it screens applicants?
Those questions sit at the center of Mobley v. Workday, a lawsuit alleging that Workday’s AI-powered hiring technology discriminated against job applicants. In June, a California federal judge allowed much of the case to continue, including claims involving age and disability discrimination.
This case emphasizes that as employers leverage AI in hiring, they are still responsible for overseeing how the technology screens candidates.
Facility managers may never select or configure the organization’s AI recruiting platform. Those decisions typically rest with Human Resources (HR), IT or corporate leadership.
But when it comes time to hire an HVAC technician, electrician or maintenance professional, HR turns to supervisors to define the job, establish qualifications, interview candidates and make final hiring recommendations. To make confident hiring decisions, facility leaders need to know that the AI tool is putting forward qualified applicants.
And if not, the Workday case raises questions about who is responsible if a third-party technology influences employment decisions. Stacy Thompson, a labor and employment litigator at DarrowEverett, warns that using an outside system doesn’t insulate an employer from liability.
Thompson compares using an AI hiring tool to hiring a consultant to evaluate candidates. If that consultant engages in discriminatory practices, the employer cannot shrug off responsibility because someone else performed the initial screening.
She asserts that same principle applies when software performs the work.
“If there is bias to be found, it can and will be imputed to the employer as the final backstop and actual decision maker,” Thompson says.
That makes understanding how candidates are evaluated more than an HR or technology concern. Facility managers should know whether AI is part of the recruiting process for their positions and what criteria it uses to move candidates forward or weed them out.
AI doesn’t independently determine what makes a good HVAC technician, building engineer or maintenance supervisor. Its conclusions depend on the information it has been trained on and the criteria an employer asks it to evaluate.
Those inputs can introduce problems, Thompson shares.
“AI makes decisions based on the data it’s given. So, if there is already bias in an employer’s hiring practices, that will continue,” she says.
Bias can also arise through the algorithm’s design or the criteria used to evaluate candidates. Even requirements that appear neutral could inadvertently favor or disadvantage particular groups.
This fact amplifies the importance of quality job descriptions.
A vague job posting asking for an “experienced” technician, for example, leaves considerable room for interpretation. What does “experienced” actually mean? Five years in the field? A particular certification? Experience servicing certain equipment? Familiarity with a building automation system?
Thompson recommends replacing ambiguous requirements with concrete skills and qualifications whenever possible. For example, the job description might name a specific HVAC certification, require electrical knowledge, or identify equipment, systems or technology the employee must know how to service.
“Make sure that you’re designing your job description parameters around those things rather than amorphous concepts such as experience. Define what experience actually means,” she says. “Without those parameters, you might get involved in age discrimination issues because no one really knows what experience means.”
Clearer job descriptions also help applicants better understand the position and give AI screening systems more relevant criteria on which to base their recommendations.
One of the biggest risks shows up when employers run AI in the background and the system operates unchecked. The tool might screen 200 applications and flag 20 potential candidates. Those 20 applications move forward, while the remaining 180 disappear.
That’s handing AI the keys and never checking the direction it’s headed, according to Thompson.
She explains if no one examines why those decisions were made, the organization may not know if AI excluded qualified candidates or a screening criterion introduced unintended bias.
For this reason, she says employers must remain actively involved.
“Don’t just put something out there and let it do its thing,” she says. “This is a tool in your toolbox. It’s not something that you should let go wild without human oversight.”
Facility managers, she says, must understand the practical decisions these tools make, such as:
Then they must be able to understand the basis for those decisions. “Why did the AI tool make that decision? You really need to understand that,” she says.
Employers also shouldn’t assume a hiring system works as intended because it came from an established and reputable technology provider. Thompson recommends testing AI screening tools before relying on them.
Facility leaders could work with HR to run sample applications through the system and examine the results. If comparable applicants receive different treatment, managers should investigate why before using the tool to screen actual candidates.
How often this testing occurs depends partly on hiring volume, Thompson says. Organizations that process volumes of applications or experience substantial turnover may need more frequent reviews.
At a minimum, she suggests testing the system with each new position posting.
“It makes sense to test every new job before you post it. Do a blind test to see how the tool responds to an application,” Thompson says.
HR and facility leaders should also ask vendors questions about their hiring technology before entrusting it with applicant screening. These questions may include:
“Depending on how the algorithm is coded or designed, there can be implicit bias because it all stems from human decisions,” Thompson explains.
Policies governing AI-assisted hiring are another important safeguard, but Thompson warns, “there can be a disconnect from policy on paper and what’s actually happening.”
Facility managers should understand their organization’s hiring policies and work with HR to ensure those requirements carry through to their own recruiting.
This involves documenting necessary qualifications, defining who will review AI recommendations, specifying when human oversight is needed, and verifying adherence to established policies.
“The legal landscape surrounding AI-assisted employment decisions continues to evolve,” Thompson adds. “Existing discrimination laws still apply while legislatures and courts determine how those protections translate to automated decision-making. This makes documentation, oversight and consistency very important.”
AI technology can process applications faster than a person, but speed means little if the process eliminates people who should have received consideration.
Facility managers must understand how AI technology affects the candidates who sit across the interview table and those they never see at all.
For now, Thompson’s gives some straightforward advice.
“Never treat AI as an automated process you can let run in the background with no oversight,” she says. “It’s a tool in your toolbox. It’s does not replace a human being.”
Her advice may be the most important hiring rule of all in the age of AI. Use the tool, but don’t hand it the keys.
Ronnie Wendt is the owner of In Good Company Communications and a freelance writer specializing in articles for the facilities management, aviation, RV and automotive, meetings and events, security, logistics and business technology industries.
