Somewhere tonight, an employee who just had a hard conversation with their doctor is anxiously typing questions into a chatbot looking for help. They’re trying to figure out if they qualify for FMLA and how much time they can take. The answers arrive quickly and sound certain—whew. But most of them will be wrong.
This moment illustrates why AbsenceSoft is clear about how we think about AI, and why we want our trusted audience to know. Leave and accommodations are where some of the most personal moments in a person’s life run into some of the most complex regulations an HR team will ever manage.
Sure, AI can make this work lighter. But when it’s used carelessly, it can make this work risky. We think it’s important to be clear about how you can hold both of these truths at once.
What is AbsenceSoft’s position on AI?
We like to think of ourselves as AI-forward. We’ll continue to implement AI where it can help make leave and accommodations easier, more efficient, and more accurate. We prefer to keep AI out of any places where compliance has no room to flex.
Our mission is the same we’ve held for more than a decade. Our commitment to customers has always focused on automation, compliance, thoughtful employee experiences, and insight leaders can act on. AI gives us a better way to deliver on them. We’re adopting it because it serves that work, and we wouldn’t do it otherwise.
Why is “confidently wrong” the biggest AI risk in leave compliance?
In compliance, a careful answer will always serve people better than a confident wrong one.
Our team conducted firsthand testing of large language models’ (LLMs) accuracy in compliance. When we benchmarked major general-purpose LLMs on leave classification, every model we tested was severely overconfident—and severely wrong. Each was incorrect while sounding completely sure of itself.
And no, better prompting won’t solve that problem. Leave eligibility depends on overlapping federal, state, and employer rules (on tenure and hours thresholds) and how they interact. General models were never trained on that body of knowledge, and they have no compliance logic to check their answers against.
Getting this information wrong has very real costs. An employee can lose job protection they were entitled to, and an employer can face a wage-and-hour audit or an ADA complaint. For someone looking for answers in an LLM, a bad one sends them in the wrong direction during a season that’s already hard enough.
What should AI do in leave management, and what should stay human?
AI should carry the weight while people keep their hands on the wheel.
AI is well suited to handle some heavy, repetitive parts of this work. It can help an employee describe what’s happening in their own words, organize intake, check documents for completeness, and flag what needs to happen next. Those tasks can take hours and rarely require judgment. Handing them off to AI can give leave teams time back for cases that need them most.
Decisions, however, should stay with people. Eligibility determinations, accommodation decisions, and conversations with an employee who is scared, grieving, or finding their way back to work belong with leave administrators and HR managers. AI can help them prepare for those conversations, but that’s the most it should be doing.
AbsenceSoft cares about this well beyond our own platform. Leave professionals are some of the hardest-working and least-thanked people in HR. They’ve always brought judgment and care to the job. Good technology should make more room for both.
How does AbsenceSoftAI work with its compliance engine?
We made a deliberate architectural choice to have AI work around the AbsenceSoft Compliance Engine™ (ACE™) while leaving the compliance logic inside it untouched.
ACE is rules-based and governed by people. It covers more than 200 federal and state leave and accommodation laws, and our in-house compliance team keeps it current as legislation changes. Whenever AbsenceSoft calculates eligibility or entitlement, it follows those rules every time and maintains an audit trail.
AI makes the experience around the engine smarter, from the moment an employee starts a request to the way leaders see trends across their program. We believe that’s the right design principle for AI in any regulated function. And it means our customers get the benefits of AI without needing to become experts or take on risk inside their compliance program.
The AI we’re building is proprietary and trained specifically for this domain, shaped by over a decade of experience in the leave and accommodations industry. We built it to be confident when it’s accurate, and honest when it isn’t. In work like this, knowing the limits of what you know is one of the most valuable things a tool can offer.
What’s the difference between automation and AI in leave management?
Automation and AI are different, and it’s worth understanding before you evaluate any vendor’s claims.
Automation follows a set of rules. It does the same thing, the same way, every time. For example, when a system calculates FMLA eligibility, it applies defined logic the same way each time. Much of what AbsenceSoft delivers today comes from this kind of automation.
Rules can’t do everything. AI can handle what rules can’t, like making sense of how an employee describes their situation or spotting patterns across thousands of cases. AI runs on probability, which makes it a helper to your compliance rules; it should never be the one deciding. When a vendor says its compliance is AI-powered, it’s worth asking which parts use AI (and how) and which parts use rules.
Why not just use ChatGPT for leave and accommodations questions?
This is a fair question, and plenty of employees already do this. Accuracy is the first issue to consider. Our benchmarking showed general models get leave wrong with high confidence and high inaccuracy. By design, LLMs are probability machines, not answer engines.
There’s also no compliance engine inside an LLM. Matching what ACE does would mean rebuilding an entire compliance operation, including the people who research, interpret, and test every law.
A human needs to keep this information current. Leave laws change every quarter, so when California expands CFRA or Massachusetts amends its paid family and medical leave law, the logic has to be updated, tested, and documented. AI can’t do that, and homegrown tools fall out of date without anyone noticing. These paths present compliance risks that teams can’t afford to carry.
Can organizations that are cautious about AI still use AbsenceSoft?
Yes, of course. We think all perspectives on AI are reasonable. Some are eager to put it to work, while others want to keep it out of employee-facing tools for now.
We build for both. Every organization decides how AI shows up in its experience. Traditional paths like form-based intake remain available for teams that aren’t ready. Customer data is protected by contract, and we don’t use it to train AI models. Whichever path a team chooses, the compliance engine underneath works exactly the same way.
What this means for HR and leave teams
Seth Turner, AbsenceSoft’s co-founder, likes to say that in a few years the conversation will move past AI itself and settle on the experiences we created with it.
That’s where we like to keep our focus—on the employee experiences we help you create. We picture an employee who gets a clear, accurate view of their options on one of the hardest days of their year. And we picture a leave team who gets the time and confidence to show up for them well. To get there, we’ll use AI intentionally where it makes sense, and leave it out where it doesn’t.
FAQ
-
Yes. AbsenceSoft uses proprietary, domain-trained AI to make intake, document review, the employee experience, and program insights faster and clearer. We apply it where it improves accuracy and efficiency, and we keep it out of compliance determinations.
-
No. Eligibility and entitlement come from the AbsenceSoft Compliance Engine™ (ACE), which is rules-based and governed by people. Leave administrators and case managers make the final decisions.
-
General-purpose AI tools aren’t reliable for leave eligibility. AbsenceSoft’s internal benchmarking found that every general-purpose model we tested was severely overconfident in leave classification and often wrong, because these tools lack a compliance engine and current state-law logic.
-
Automation applies defined, traceable rules, such as calculating FMLA eligibility or sending notices on schedule. AI works on probability and handles things rules can’t easily capture, like plain-language requests or pattern detection. In AbsenceSoft, automation and ACE handle compliance logic, and AI supports the surrounding experience.
-
Yes. Each organization chooses how AI shows up in its experience, and traditional options like form-based intake remain available. The compliance engine works the same way either way.
-
No. Customer data is protected by contract and isn’t used to train AI models.