On this page
The Utilization Problem
Start with the number that should reframe every benefits conversation you have this year: the most reliable primary data on Employee Assistance Program uptake — a Business Group on Health survey of large employers — found a median EAP utilization rate of just 5.5% (2018 data). That means, in a typical large employer, roughly 19 out of every 20 employees never use the mental-health benefit their company pays for in a given year.
This isn’t a niche problem or a one-off bad survey. It’s the baseline the entire employee-benefits industry has been trying to solve for over a decade, and it is the number every new vendor — AI-powered or not — is implicitly promising to beat. Before you evaluate any tool against it, it’s worth understanding why the number is so low in the first place, because the answer determines whether a new delivery channel (an app instead of a phone line) actually addresses the cause.
Low utilization is also a budgeting problem, not just an engagement problem. A benefit that a company pays for on a per-employee or per-seat basis but that 19 out of 20 employees never touch is, from a pure cost-per-use standpoint, one of the least efficient line items in a typical benefits package. That’s part of why HR and People teams keep re-evaluating this category every few years — the underlying need clearly exists (see the workforce mental-health cost figures in our companion HR buyer’s guide), but the delivery model keeps failing to reach the people it was built for. Understanding why it fails is the only way to evaluate whether a new delivery model — AI included — is actually solving the right problem.
Why Employees Don’t Use the Benefit
A 2023 peer-reviewed systematic review by Long & Cooke synthesized the research on why EAP and workplace mental-health benefits go unused and converged on three recurring barriers:
- Stigma. Employees worry that seeking help signals weakness, instability, or that they can’t handle their job — a concern that persists even where companies have publicly committed to destigmatizing mental health.
- Confidentiality doubts. Employees are not convinced that what they disclose stays private from their manager, HR, or the company more broadly — regardless of what the benefits handbook says about anonymity.
- Lack of awareness. A meaningful share of employees simply don’t know the benefit exists, don’t remember how to access it, or don’t understand what it actually offers versus their health insurance.
Notice what’s not on that list: none of the three leading barriers is “the benefit was hard to schedule” or “there was no evening availability.” The dominant barriers are trust and knowledge problems, not logistics problems. That distinction matters a great deal for evaluating whether an app actually fixes anything.
It also means the usual first-round fixes — extending phone-line hours, adding a mobile check-in feature, simplifying the intake form — are treating symptoms rather than the underlying cause. An employee who is worried their manager will find out they called the EAP is not helped by a faster intake form; they’re only helped by a credible reason to believe the call stays private. An employee who has never heard of the benefit is not helped by a shorter session — they’re helped by communication that reaches them more than once. Any new tool, AI-based or not, has to be evaluated against these three specific barriers, not against generic usability improvements.
What Employees Actually Fear
It’s tempting to treat “stigma” as an abstraction. The data makes it concrete. The American Psychological Association’s 2023 Work in America survey found that 43% of workers worry that disclosing a mental-health condition would negatively affect them at work. That’s not a fringe concern — it’s nearly half the workforce operating on the assumption that being honest about their mental health carries a professional cost.
That fear has measurable behavioral consequences. A 2025 NAMI/Ipsos workplace poll found that 25% of employees considered quitting a job over a mental-health issue in the past year, and 7% actually did. The same poll found that only 13% told their manager about it. Put those two figures side by side: employees are more willing to quit their job over a mental-health issue than to tell their own manager about it. That is the gap a benefit has to close, and it explains why raising awareness of a hotline number rarely moves the needle — the barrier isn’t not knowing the number exists, it’s not trusting what happens after you dial it.
Does AI/Digital Access Close the Gap?
The logical pitch for AI and self-guided digital tools is straightforward: if stigma and confidentiality doubts are the barrier, then a private, always-on, no-human-in-the-loop channel should lower the barrier to entry. There is real survey evidence supporting the comfort half of that argument. A 2026 Bipartisan Policy Center survey found that roughly 70% of US adults say digital mental-health tools feel “more comfortable” than a direct conversation, and about 50% cite cost as a reason they turn to digital tools over traditional care. Both findings track cleanly with the stigma and access barriers identified above — employees who won’t tell their manager may still be willing to type into an app that no one else sees.
So on paper, AI access addresses two of the three documented barriers directly: it removes the exposure risk that drives stigma and confidentiality doubts, and it’s available the moment someone thinks to look for it, which helps with the awareness problem too — provided the benefit is actually communicated and easy to find. That’s a real, evidence-backed reason to expect some utilization lift from adding a digital layer. It is not, however, the same as proof that employees will actually adopt it at scale, which is a separate question with its own, less encouraging data.
Design matters here as much as the underlying model. A tool that is procured at the organizational level but still routes any signal — even aggregate usage counts broken down by team — back to a manager or HR dashboard reproduces the exact confidentiality doubt that suppresses EAP usage in the first place. If the goal is to use AI or digital access to address stigma and confidentiality specifically, the procurement decision has to protect that anonymity as a hard requirement, not a nice default. Our HR buyer’s guide covers the specific data-privacy questions worth asking a vendor before you sign.
The Honest Limits
Comfort with the idea of a digital tool and actual intent to use an AI chatbot for mental health are two different things, and the gap between them is the most important number in this guide. The same 2025 NAMI/Ipsos research found that only 12% of US adults were “likely” to use an AI chatbot for mental-health treatment in the next six months, and just 1% already did.
Read that alongside the 70% comfort figure above and a real tension emerges: a large majority say digital tools feel more comfortable than a direct conversation in the abstract, but only a small fraction currently intend to use an AI chatbot specifically for mental-health treatment. “Digital mental-health tools” in the Bipartisan Policy Center survey is a broad category — think meditation apps, mood trackers, and self-guided programs — while “AI chatbot for mental-health treatment” is a narrower and, for many people, a more clinically loaded ask. Employers who assume comfort with digital tools in general translates into demand for an AI mental-health chatbot specifically are extrapolating past what the data supports.
The honest summary: AI access plausibly helps with the stigma and confidentiality legs of the utilization problem, but adoption intent for AI-specific mental-health tools remains low across the broader population as of the most recent data available. A digital layer is a reasonable hypothesis for lifting utilization above 5.5% — it is not a proven fix, and vendors who present it as one are getting ahead of the evidence.
What Actually Moves Utilization
Given that the dominant barriers are trust and awareness rather than access mechanics, the levers most likely to move utilization — whether or not you add an AI layer — follow directly from the barrier research:
- Address confidentiality doubts explicitly and repeatedly. A single line in the benefits handbook does not overcome a well-documented trust gap. Managers and leadership need to say, in plain language and more than once, exactly what HR does and does not see.
- Treat awareness as a communications problem, not a one-time launch. If lack of awareness is a top-three barrier, a benefit that’s announced once at open enrollment and never mentioned again will underperform regardless of how good the tool is.
- Reduce disclosure exposure wherever possible. Anything that lowers the number of humans an employee has to go through to get help — anonymous intake, no-manager-visibility design, self-serve first contact — targets the stigma barrier directly.
- Don’t assume a new channel replaces the trust-building work. If you add a digital or AI layer, treat it as one input into a utilization strategy, not a substitute for the confidentiality and awareness work above. The adoption-intent data above suggests the tool alone won’t carry the gap.
Bottom Line
Utilization is low because employees don’t trust that using the benefit is safe or don’t know it exists — not because the benefit is hard to reach. AI and digital access address part of that equation by removing exposure and offering always-on entry, and there is real survey evidence that employees find digital tools more comfortable than a direct conversation. But the current adoption-intent data for AI mental-health chatbots specifically is still low, so treat AI as one lever in a broader utilization strategy — built alongside real confidentiality guarantees and ongoing communication — rather than as a standalone fix for a 5.5% baseline. This page is general information for benefits decision-makers, not legal or clinical advice.
Related Reading
- AI Mental Health Tools for Employee Benefits: An HR Buyer’s Guide — the full evaluation framework for AI vendors, privacy, and crisis safety
- EAP vs. AI Mental Health App — how the two benefit models compare on access, cost, and clinical scope
- Crisis Escalation for Employers — what has to be in place before you route employees to a self-guided AI tool
- Is AI Therapy Effective? What the Research Shows