Though technological innovation is always happening, in the past year there has been an almost inescapable reference to a decades-old science fiction term: AI, or Artificial Intelligence. Seen as a watershed moment of research and development in the past, AI is now a buzzword for content creation, software features, and product design. Whether you are on the internet or watching TV, AI seems to be everywhere. But some employers don’t realize that AI is making its way into the workplace, whether it is wanted or not. This means that regulation of this budding technology is inevitable.
On April 29, 2024, the Department of Labor’s Wage and Hour Division joined the fray to provide its analysis on emerging issues, publishing a Field Assistance Bulletin (“Bulletin”) concerning “Artificial Intelligence and Automated Systems in the Workplace under the Fair Labor Standards Act and Other Federal Labor Standards.” The bulletin discusses how AI can affect the workplace in relation to several federal laws, but dives a bit deeper into how it might interact with the Fair Labor Standards Act (“FLSA”). Specifically how it relates to these two FLSA issues: “Hours Worked” and “Calculating Wages Owed under the FLSA”, and we think both deserve a little more perspective.
Hours Worked and AI’s Impact on Employee Tracking – The DOL Bulletin
The Bulletin provides an overview of how all “hours worked” must be paid under the FLSA, grouping its discussion into AI’s impact on tracking work time, monitoring break time, waiting time, and employee travel. Each discussion point focuses on how AI now makes it easier and more tempting than ever, for employers to track an employee’s work performance.
The FLSA mandates employers pay covered employees at least the federal minimum wage for every hour worked, as well as overtime pay of one and one-half their regular rate of pay for each hour worked in excess of 40 hours in a workweek. See 29 U.S.C. §§ 206-207. In all cases, it is the duty of management to exercise control and ensure that work is not being performed when it does not want work to be performed. See 29 C.F.R. § 785.13. That makes enforcing schedules and tracking employee time essential for compliance and, if the employer knows or has reason to believe that work is being performed, it counts as hours worked. See 29 C.F.R. §§ 785.11, 785.12.
With AI improved employee tracking, it is easier than ever to recognize when employees are working and how much time they are spending on individual tasks. In 2022, Vice published an article exploring Amazon’s detailed warehouse employee tracking system that used employee’s handheld scanners to discern their actions, and which Amazon ultimately used to discipline employees for time spent “off task”. As the Bulletin recognizes, AI and employee monitoring tools are now more widely available and accurate, so that an employer can know precisely when an employee is idle and not performing tasks directly related to their job.
But “hours worked” under the FLSA are not always spent working. Short breaks and waiting time might require payment, even if the employee is not doing any actual work during those periods, depending on the circumstances. See 29 C.F.R. §§ 785.14 (noting that being engaged to wait is compensable time), 785.18 (noting that rest periods of short duration from five to 20 minutes must be counted as hours worked). Though part of the rationale for these rules is that such breaks and waiting time boost productivity and efficiency, the end-result is that measurable productivity and efficiency play no role in the calculation of “hours worked”. So though AI and an increasingly digital world are making it easier to track employees and their work performance down to the minute, employers cannot claw back “idle” time.
So what happens if an employer can definitively prove that an employee was not working at a certain time (using advanced AI tracking) and has a written policy prohibiting idle time while working? From an FLSA perspective, not much. With an increase in wearable smart technology, employers now have more tools that collect real time data that AI could use to classify activities as “work” or “idle” time, regardless of whether employees are working desk jobs or are in the field. Whether an activity is excluded from hours worked under the FLSA is a mixed question of fact and law, meaning that there is no one-size fits all solution and the details matter. But “hours worked” is not changing to reflect that AI might enable more that accurate tracking, so “idle” time remains a performance issue, rather than an FLSA one.
One area of the FLSA that AI may significantly impact is the defense of “de minimis” time. The de minimis time concept arose because sometimes, small periods of time spent on work related tasks could not be accurately recorded for payroll purposes. See 29 C.F.R. § 785.47. De minimis time has impacted cases where employees spent time “donning and doffing” protective gear and equipment, took time uploading data gathered during field work once returning home, and when logging into an employer’s time-keeping program. Under the de minimis doctrine, the employer bears the burden to satisfy a three-part test which considers: (1) the regularity of the additional work, (2) the aggregate amount of compensable time, and (3) the practical administrative difficulty of recording the additional time. But with the capability for increasingly accurate employee tracking and monitoring through AI, we may see the erosion of the de minimis doctrine, as situations involving uncertain and indefinite periods of work time become more rare.
So, while the FLSA may eventually change to reflect improved employee tracking abilities, for now the traditional rules on “hours worked” continue to apply.
Payroll and AI
The Bulletin also discusses how AI might impact the calculation of wages under the FLSA, aka payroll, due to increased automation. With the ability for AI systems to track and assess the amount or type of work an employee is performing in detail, AI systems can also recalculate and adjust pay rates in real time to reflect when an employee is completing certain tasks. For instance, if an employee earns a premium during work time spent in a certain location, or working out of class, an AI system might be able to track and log that without relying on an employee or supervisor to manually tag such an adjustment.
Despite the jump in technology, there are a few reasons to be wary about adopting an AI payroll system without close oversight. AI automation can be extremely useful for employers (and employees) to tackle time-heavy and rote tasks but can also have unintended consequences when complicated variables are at play. For instance, hiring systems using AI have faced backlash for unintentional discrimination. And during the early days of image generative AI systems, AI-generated images regularly displayed stereotypical and offensive things that it had not yet “learned” was incorrect. Unless you designed it, there is a lot unknown about how a given AI system works – it is a “black box” where inputs go in and outputs come out. AI systems may not properly account for variables, or can have a “bias” built in due to its designer’s own misunderstandings. So there are several reasons why an AI system might fail to address an unforeseen problem that requires a more “human” touch.
AI payroll systems, while hopefully not as prone to discrimination-related mistakes, still pose the danger of oversimplifying or failing to account for legal requirements, especially when their designers do not know what to watch out for. One hallmark of the FLSA is that for non-exempt employees, employers must pay employees overtime, at one and one-half times their “regular rate of pay”, for any time worked in a workweek in excess of 40 hours. Beyond the difficulties of tracking time, discussed above, the “regular rate of pay” causes multiple complications in the public sector because of the need to comply with local rules, collective bargaining agreements or memoranda of understanding, the FLSA, and more. And because the “regular rate of pay” includes “all remuneration” with certain exclusions, it is not always easy to determine which payments made to employees need to be included. If a payroll vendor is trying to craft a payroll system from the ground up, there are a lot of legal variables that aren’t readily obvious, and quite a few employers are going to have situations where legal advice is needed to ensure proper calculations.
So an AI payroll system vendor attempting to build their system without fully understanding or accounting for those changing variables may ultimately produce a payroll system that suffers from similar problems that traditionally offered automated payroll systems face. Even in “man-made” automated payroll systems, we regularly see payroll errors happen frequently due to human error, such as when a bonus is coded incorrectly and not counted in the regular rate of pay, when overtime is due for hours worked, and when certain cash in lieu payments are not included in the regular rate of pay. Add onto these issues that the designers of AI payroll systems might not know how to properly address both FLSA and MOU problems at the same time, and you can quickly have unintended underpayments (or overpayments) happening with an AI payroll program that provides a quick fix. And even if an AI payroll system was capable of “learning” what variables or documents to consider, and then seek them out, there is simply no guarantee that the AI payroll system would still be legally accurate without review.
In short, AI is likely to provide additional automation that makes it tempting to quickly adopt, but employers are going to want to double-check the work for accuracy. We heavily recommend, just like we do for all current payroll systems, regular audits and oversight of any payroll system using AI.
Future AI Uses in the FLSA Sphere
AI has the potential to drastically improve the productivity of workers, alter employee monitoring, and quite possibly eliminate the need for certain positions. One possibility is that AI will allow employees to hold multiple jobs, as rote tasks and data analysis become simpler and quicker to perform. If an employee works multiple job classifications for a single employer with different rates of pay, maybe the AI payroll system could automatically detect and accurately pay the employee the correct rate of pay for the time spent on each classification, as well as calculate their regular rate of pay to properly pay overtime.
Another possibility is that AI could become widely incorporated into wearable trackers issued to all employees. We already have law enforcement personnel using body-worn cameras, is it a stretch to have all employees wear some sort of similar tech that allows constant tracking during the workday? There are already hi-tech eyeglasses on the market that allow for video recording without much added bulk, and we would anticipate it will not take long for AI to become incorporated. Will the FLSA need to adapt so that there are more protections for “hours worked”? With more advanced AI technology creeping into everything, we might find out the answers to the above answers sooner rather than later.