AbsenceSoft’s new survey of 1,200 HR professionals found that 70% already use AI on accommodations work, and that only a third have a formal AI policy for it. HR teams should write that policy before they buy anything else.
The gap in the numbers
AbsenceSoft, which sells leave and accommodations software, published its Business Impact Report on 29 October. The survey ran through Pollfish between 10 and 23 July 2026 and covered US HR professionals at organizations with 500 or more employees. It is a vendor’s survey, and I read it that way. The AI figures still line up in a pattern worth acting on.
Of the respondents, 70% already use AI for accommodations tasks. Only 31% are very confident their AI tools reflect current accommodations law, and for leave law the figure is 27%. Concerns about AI are near-universal: 77% for accommodations and 79% for leave, citing data privacy, human judgment, compliance risk and accuracy. Yet only a third have a formal AI policy for accommodations, and 28% have one for leave.
Meanwhile 35% are actively evaluating AI solutions for accommodations, and 22% already have an expansion plan. Use is ahead of confidence, and confidence is ahead of written rules.
Why the policy comes first
An accommodation request is a legal process with deadlines. The survey says 60% of HR teams see accommodation cases fall behind schedule or miss deadlines at least occasionally, and that 64% saw requests rise over the past two years. A tool that drafts a response or summarizes a medical note touches that process directly. If nobody has written down what it may draft, what a person must check and where the case record lives, each HR partner decides alone.
Seth Turner, Founder and Senior Advisor at AbsenceSoft, put the stakes in business terms: “Replacing an employee can cost far more than providing the accommodation in the first place.” Half of the surveyed organizations say they have lost people who might have stayed with a better accommodation experience. A wrong or late AI-assisted answer lands on exactly that experience.
The policy I am arguing for is short. It names which tasks AI may do, such as drafting a form letter or answering a policy question from an approved handbook. It names which it may not, such as deciding eligibility. It says who reviews the output, how the review is logged, and who updates the tool when a state paid leave law changes. It also covers what data may be entered, since privacy is among the concerns respondents cited.
The counter-argument
The strongest objection is speed. Half of the surveyed teams call forms and paperwork their biggest time drain in accommodations, and 36% point to manager and employee questions. Those are the tasks AI handles well, and requests are climbing. A policy process looks like one more delay.
The survey answers part of this. Only 7% of HR teams generally act on AI output without human review, so most teams already apply the rule I am asking them to write down. A formal policy turns a habit into something a new hire, an auditor or a plaintiff’s lawyer can read. It also fixes who is accountable when the tool is wrong, which matters more once a vendor’s feature is switched on for the whole department.
The objection also assumes the policy takes months. A one-page version can be drafted in an afternoon from the points above. A tool bought without one starts producing case records on day one, and those records are the ones that get read later. I made a related argument when I wrote that HR should tell employees how far its AI agents can act, and the same logic applies inside the HR team itself.
What to do this quarter
Start with an inventory of where AI already touches leave and accommodations work, including features inside the HR platform the team already owns. In the survey, 44% say their all-in-one HR platform covers the basics but leaves compliance gaps, so some of that AI may arrive through a vendor update and not a purchase. I made the same case for hiring tools in my piece on inventorying hiring AI before January.
Then write the one-page policy, have legal read it, and ask each vendor in an evaluation to show how its product supports the review step. Our coverage of two EEOC pregnancy suits shows how a leave policy can become the center of an accommodation dispute, and a written AI policy is part of the record in that kind of case.
The 77% of organizations planning to raise accommodations investment over the next 12 months will buy something. The ones with a written policy will know what they are buying it to do.
