Imagine it’s 4 p.m. on a Friday when an employee sends you a message. She’s having knee surgery in a couple of months, and she wants to know whether she could get a few weeks off under the Family and Medical Leave Act (FMLA). You don’t know off the top of your head, so you pull up ChatGPT.
ChatGPT is quick with a response: Of course she’s eligible for FMLA leave! She’s logged well over 1,250 hours working for your employer, where she’s worked for the last two years. Your employer is definitely covered, too, since it meets the FMLA’s threshold of employing 50 or more employees in 20 or more workweeks in the current or previous calendar year.
Can I use AI for FMLA management? Read on to find out.
You narrow your eyes and ask yourself a few questions. Is it okay to use ChatGPT for FMLA? Could ChatGPT be wrong? Its answer is straightforward and confident, but you wonder if you can trust it.
General-purpose AI tools are helpful for many leave management tasks. You can use ChatGPT, Gemini, and Claude to draft training materials, analyze non-confidential data, and brainstorm accommodations. But you can’t trust these tools to make leave decisions that require nuance, compliance, confidentiality, judgment, and expertise.
When you use generic AI for HR compliance, you can run into serious consequences. In this blog, we’ll show you how to find AI-powered platforms that maximize your productivity while minimizing the risk of costly compliance errors.
How HR Already Uses AI For Leave Management
If you’re in leave management, there’s a good chance you’re already using AI to get some of your work done. And if you haven’t tried out a tool like ChatGPT, Claude, or Gemini, someone else on your team probably has. According to AbsenceSoft research, 45% of HR leaders are already using AI during the leave process. And just over half (51%) say they’ve used it to manage accommodations.
Our survey also revealed how HR is applying AI to leave and accommodations. The top uses include tasks like reporting and analytics, drafting communications, and looking up applicable laws.
Generally speaking, tasks like these are appropriate for general-purpose AI tools. If a large language model hallucinates and makes a mistake, the consequences will be small. Repercussions are even more minimal when managers review and edit AI’s output.
But HR is also leaning on AI to handle tasks of greater importance. About half of respondents said they use the technology to understand employee eligibility, make eligibility decisions, and calculate entitlement amounts. Often, leave managers tap AI without the guardrails of policy or oversight. This can lead to errors, security violations, compliance mistakes, and more.
It’s worth noting that, while many HR leaders say they’re already using AI, a significant chunk has stayed away from the technology. Of those, 40% are uncertain about whether they’ll use it in the future.
This hesitancy highlights an appropriate sense of caution among HR leaders. While AI can help leave managers reduce administrative work and boost productivity, it can also create costly unintended consequences.
Why FMLA Challenges General-Purpose AI Tools
Can you have Claude handle leave of absence? Can you just ChatGPT FMLA? It may not be as simple as it seems.
Large language models like Claude and ChatGPT are trained on huge amounts of data. The tools draw on that data to answer questions and respond to prompts. Using data, the tools can help HR teams come up with possible accommodations for a disabled employee or draft communications for employees taking a leave of absence.
For purposes like these, a tool needs to be reasonably up to date, drawing on information that was current in the last year or two. When HR teams use AI tools to make compliance-related decisions, however, their tools must be nothing short of precise.
Here’s why. When leave managers are administering FMLA leave, they’re often managing state-level leave laws, too. The specifics of these laws differ in every state, and they change almost constantly.
LLMs that lack real-time legal updates cannot be relied on when it comes to questions about an employee’s eligibility or entitlement for leave under the FMLA, state law, or local statutes. Generic AI tools are trained on data that may not be up to date on changes in the law.
For instance, the PWFA, PUMP Act, and expanded ADA interpretations have happened recently enough that some LLMs will surface outdated information regarding their implementation. When questioned, a model may generate an answer that sounds authoritative and sound, but its response could feature information that was accurate two years ago but is wrong today.
What’s more, eligibility calculations require more than an accurate knowledge of rules and regulations. They rely on specific information about an employee’s work history.
Generative AI Compliance Risk: When AI Is “Confidently Wrong”
Leave managers often tap general-purpose AI tools when they’re unsure about something. They may need help remembering the nuances of Colorado’s Family and Medical Leave Insurance program. Or they may not be able to recall how Massachusetts’ Paid Family and Medical Leave interacts with the FMLA.
In these moments of uncertainty, leave managers want quick, confident answers. LLMs deliver. Tools like ChatGPT and Gemini are quick to produce complete, well-formatted answers that read like expert guidance.
Here’s the problem: AI’s confidence doesn’t translate to accuracy. According to one Stanford study, LLMs hallucinate in 58-82% of legal queries. AbsenceSoft benchmarking has confirmed that this rate holds steady for leave classification as well.
Mistakes in matters of compliance can land employers in serious trouble. A wrong eligibility determination, missed state law interaction, or miscalculated entitlement can result in fines, DOL complaints, or lawsuits.
These outcomes can be costly. Employers spend an average of $78,000 defending employment lawsuits, regardless of the outcome. The U.S. Department of Labor regularly collects tens of thousands of dollars from employers that violate the FMLA and other statutes.
Leave management mistakes generate costs beyond the financial, however. According to AbsenceSoft research, 51% of employees who had a negative leave experience said they felt less loyal to their employers upon returning to work. Many said they were less motivated (42%) and others said they started looking for a new job (36%).
What AI Should and Should Not Do in Leave Management
When used improperly, HR AI tools risk mistakes that create legal exposure and damage the employer-employee relationship. But that doesn’t mean AI has no place in the leave management process.
Artificial intelligence can be a major boon for busy leave managers, especially as budgets contract and administrative tasks multiply.
Leave managers need to have a clear idea of the tasks they can delegate to AI and the tasks that need to remain on their own to-do lists.
| Tasks for AI | Tasks for leave managers |
|---|---|
| Drafting communications for human review | Making decisions on eligibility |
| Summarizing case notes | Generating entitlement calculations without review |
| Flagging incomplete documentation | Providing legal interpretation |
| Generating reports | Protecting sensitive employee data |
Unsure whether to trust AI with a particular task? Remember that AI acts as an assistant to a trained professional. It’s not a replacement for compliance expertise.
If your organization is one of the 60% planning to invest more in leave management next year, consider how potential partners and vendors view AI. Responsible leave management platforms and tools like FMLA software and FMLA eligibility software should leverage AI with ample human review. They should keep compliance rules separate from AI-assisted features.
Remember, successful leave management teams aren’t avoiding AI. They’re leveraging the technology for the tasks, problems, and challenges it’s designed to handle. And they’re hanging onto the tasks that need human judgment and oversight.
For Faster, Better FMLA, Tap AbsenceSoft
Can I use AI for FMLA? Yes, but you must use it carefully.
If you’re tempted to use ChatGPT for FMLA management, you’re not alone. Many leave administrators are busier than ever. As more employees ask for leave each year, HR has to find a way to supply more support without extra resources.
But extra bandwidth shouldn’t come from general-purpose AI tools, especially when it comes to compliance-oriented tasks.
You’re right to consider the operational advantages you can create with AI. You’re also right to wonder whether using AI can expose your organization to risk.
Here’s the good news: Your team can use AI to optimize FMLA management without increasing risk. You just need to use AI that’s built specifically for leave management.
At AbsenceSoft, we apply AI around compliance. We use AI to speed up employee intake, create workflows, and make predictions. Our technology provides instant, accurate eligibility determinations, so you never have to ask ChatGPT whether one of your employees is eligible for FMLA leave.
If you’re interested in learning more about how AbsenceSoft technology can optimize your organization’s approach to leave and accommodations, book a demo today.