Artificial intelligence (AI) is making its way into the hiring process, promising to help employers identify better candidates, reduce hiring time, and improve decision making.
For many cleaning and restoration companies, separating genuine value from marketing hype can be difficult. Wells Ye, co-founder of EmployJoy.ai and a former cleaning company owner, has spent recent years working to solve the hiring challenge by tapping into the power of AI.
Ye said he sees hope for the ongoing hiring dilemma, though how much AI can help depends a great deal on the job and how a company recruits. “Applicants are using AI, and we must as well,” he said. “But then we also have to know how, and what the best way is.”
Three areas where AI adds value
Ye broke AI’s usefulness in hiring into three categories. The first is efficiency—reducing the time it takes to handle tasks such as scheduling and automated follow-up. The second is analytics, which he said AI handles faster, more accurately, and more consistently than manual tracking, giving recruiters a clearer picture of where candidates are dropping out of the process and why.
“The third one is the most challenging one, but also the most powerful one—that is prediction,” Ye said. “AI can help us analyze and predict our outcomes in a way that has a lot more validity, if we do it correctly.”
What AI cannot fix
Ye was equally direct about the limits of the technology. A weak job offer or below-market pay will not be solved by better software, he said, no matter how the job posting is written.
“If we have a bad job offer, if we’re underpaying people and so on, using AI, however you use that—maybe you can make the job description better—will not do too much to help us, because AI will not be able to change that,” he said.
AI also cannot identify the specific traits that matter for a particular cleaning job, Ye said, since that knowledge is tied to a company’s own culture and operating requirements. And it cannot substitute for a genuine human connection with candidates. “It can go faster, can be warm and fuzzy, but the real human connection, people can still smell that miles away,” he said. Nor will it stop no-shows that stem from a poor job offer in the first place.
Ye said that dynamic cuts both ways. A strong hiring process becomes more effective with AI layered on top of it. A weak one, including underpaying workers or offering little management support after hiring, only gets magnified by the technology.
Fairness and bias remain concerns
Asked about fairness, bias, and transparency, Ye said the level of concern depends on how sophisticated the AI application is. Tools that read resumes, summarize key points, or handle automated follow-up are relatively straightforward and still require testing to confirm accuracy, he said. Prediction tools raise bigger questions, including how the underlying model was trained, whether it draws on industry-specific data, and whether the company using it has strong governance over that data.
Those questions led Ye to pursue a second role beyond running EmployJoy.ai. “I decided I want to become an independent AI and automated systems auditor, because I really want to understand—get into this fairness,” he said, adding that the work requires expertise spanning both AI and emerging legal and regulatory requirements.
Where the legal risk is highest
Ye outlined several practices he considers high risk from a legal standpoint, including fully automated rejection decisions with no human review, automated emotional scoring of candidates, and screening based on zip code or commute distance. He noted that last category carries a caveat for the cleaning industry, where commuting distance is often a legitimate, job-related requirement rather than a proxy for something else.
Lower-risk uses of AI, he said, include scheduling, screening for must-have qualifications, and prediction tools that are tested against industry-specific data with human oversight built in—meaning a person, not the algorithm, still makes the final call.
What the research shows
Ye pointed to 2016 research by Schmidt, Oh, and Schaefer, which found that common practices such as reference checks, unstructured interviews, and screening by years of experience have relatively low predictive validity on their own. Structured interviews perform better, and adding work-sample tests raises validity further, to around 63%.
“If you add AI, using data to really make these decisions, the validity can go to 70 to 80%,” Ye said. He acknowledged the research predates widespread AI adoption in hiring and that data specific to AI-assisted recruiting is still limited. “But the direction is very clear,” he said. “When you use AI to evaluate candidates systematically, you get better hiring decisions.”
A closer look at retention
Ye said EmployJoy.ai offers a complimentary cleaner retention survey that takes about three minutes to complete. The tool compares a company’s results with local competitors in the same market, giving owners a benchmark for how their retention stacks up in their area.
