The Silent Majority.

The mental health conversation in benefits has long been organized around a clinical model: identify symptoms, connect to care, treat the condition. That model made sense when the primary concern was acute illness. It makes much less sense for the workforce of 2026.

According to the 2025 Aflac WorkForces Report, American workforce burnout has reached a six-year high — with nearly three in four employees facing moderate to very high stress at work. Only 48% of employees now feel confident their employer cares about their mental health, down from 54% in a single year.

Layered on top of pre-existing pressure is something newer: AI anxiety. In a 2025 EY survey of 1,000 US workers, 65% reported being anxious about AI replacing their job. A 2026 Modern Health study found that nearly a quarter of employees report AI is already negatively affecting their mental health today.

The share of employees feeling 'very stressed' nearly doubled from 2024 to 2026. Employees are more stressed about the state of the world than at any point in recent measurement.

There's a persistent instinct in benefits to frame low utilization as an awareness problem. If employees just knew about the EAP, they'd use it. The data doesn't support that framing.

According to NAMI's 2025 Workplace Mental Health Poll, only 13% of employees who felt their mental health suffered due to work actually told their manager. Three in four employees say it's appropriate to discuss mental health at work. They know it's okay. They're still not talking.

Awareness isn't the barrier. Access design is.

To access traditional mental health support, an employee must recognize they have a problem, decide it warrants professional help, navigate provider networks, find someone available (often weeks out), show up to a session, and repeat. That sequence demands precisely the psychological resources that are hardest to access when someone is most depleted.

Employees aren't going without support — they're finding alternative paths. A peer-reviewed study published in Practice Innovations in 2025 found that nearly half of US adults with mental health challenges who use AI are turning to general-purpose tools like ChatGPT for support. A 2026 survey found that 58% of employees feel safer talking to a chatbot about their mental health than to their workplace's HR or mental health personnel.

Nearly half of respondents said an AI chatbot is the first place they turn when mental health issues arise — ahead of doctors, friends, or family.

The WHO estimates depression and anxiety alone cost the global economy $1 trillion annually in lost productivity. In the US, untreated mental illness is projected to cost approximately $477.5 billion in 2024 — and nearly $14 trillion cumulatively by 2040.

But these figures flatten a more important truth: the costs aren't primarily crisis-related. They're distributed across the entire workforce as low-level performance drag.

Presenteeism — showing up while depleted — consistently outpaces absenteeism and turnover as a cost category. Because it doesn't appear in absence data, it rarely registers as a mental health cost at all. It shows up instead as missed deadlines, quality issues, and team friction.

The answer to the utilization problem isn't more awareness campaigns or a better phone tree. It's a different model of care — one that meets people where they are, before they've decided they need help.

Thoughtful is your mental health benefit, rebuilt for how people actually want support. Every Thoughtful member has access to a licensed provider — matched, booked, and supported through T, your AI guide. T's first job is to get you to the right therapist. Between sessions, T keeps the work going. For employees who aren't ready for a provider yet, T can take care of them entirely.

Best-in-class care, from day one

Thoughtful is built in partnership with Spring Health — which means every Thoughtful member has access to the same provider network that powers one of the most clinically rigorous mental health platforms in the world.

Most workforce mental health programs are designed around symptoms. They count utilization when someone is already in pain — already missing days, already filing a claim, already in crisis. By the time the benefit fires, the cost has already accrued.

With earlier interventions through technology we believe that there is an opportunity to move a different set of measures on top of the clinical ones. The ones that change before symptoms appear when assessed.

In a randomized controlled trial run in partnership with Spring Health, people using Thoughtful improved meaningfully more than people who didn't — across nearly every measure of psychological functioning the trial assessed.

Plus statistically significant gains in positive mental health (d = 0.30), emotional self-efficacy (d = 0.23), behavioral activation (d = 0.22), and reductions in experiential avoidance (d = 0.18). Nine validated psychological functioning measures. Eight moved with small-to-moderate effects.

In the clinical literature, these are called process measures. They're the upstream mechanisms — the things that change before symptoms change. When someone engages more with their life, avoids less, believes more in their capacity to cope, and becomes more open to support, that's the substrate on which symptom change is built.

"Improvements in these process-relevant constructs may represent meaningful early indicators of benefit and could plausibly precede measurable reductions in depressive or anxiety symptoms."— Spring Health RCT

That's not a softer intervention. It's a different category of intervention entirely. One that operates on the population traditional benefits never reach, at the stage of need they never address.

For an HR leader, the question isn't whether a study moved a clinical scale. It's what the result means for next year's leave volume, productivity, claims line, and benefits spend.

Each of the metrics Thoughtful moved is a leading indicator of an operational outcome you already track and pay for. Here's the literature.

Behavioral activation ↑ (d = 0.22, p = .028)

Behavioral activation is the clinical mechanism behind depression's largest workplace cost: presenteeism. The foundational Stewart et al. study in JAMA (2003) found that workers experiencing depressive symptoms lost an average of 5.6 hours per week of productive time, with the majority of the loss coming from reduced output while at work, not from absence. Behavioral activation is the active ingredient in the most evidence-based behavioral treatments for depression (Dimidjian et al., 2006). When it moves up, that productivity loss reverses through the same mechanism.

Loneliness ↓ (d = 0.36, p = .003)

The most directly costed workforce variable in the secondary outcomes. Cigna's 2020 Loneliness and the Workplace report — the largest U.S. workforce loneliness study to date, n = 10,441 — found that lonely workers had 5.7 more sick days per year than non-lonely peers, lower work-quality ratings, and significantly higher attrition risk. Ozcelik and Barsade's 2018 study in the Academy of Management Journal established the causal link between workplace loneliness and reduced job performance.

"Lonely workers think about quitting more than twice as often as non-lonely workers."— Cigna, Loneliness and the Workplace, 2020

Emotional self-efficacy ↑ (d = 0.23, p = .037)

Self-efficacy is, alongside cognitive ability, the most-studied predictor of work performance in industrial-organizational psychology. Stajkovic and Luthans's meta-analysis in Psychological Bulletin (1998), spanning 114 studies, found a corrected correlation of 0.38 between self-efficacy and work-related performance — a stronger effect than goal-setting, feedback, or most management interventions. When emotional self-efficacy moves, individual ownership, execution speed, and reduced manager dependency move with it.

Help-seeking attitudes ↑ (d = 0.25, p = .009)

The line item: total cost of care. Untreated mental illness costs U.S. employers an estimated $300 billion annually in absenteeism, turnover, and reduced productivity (American Psychiatric Association Foundation, Center for Workplace Mental Health). Earlier engagement with care reduces episode intensity and duration. The RCT showed Thoughtful shifts people toward formal support even in a community sample not actively seeking treatment — exactly the population that drives the long-tail of leave, disability, and high-cost claims when conditions escalate untreated.

Emotional well-being ↑ (WHO-5, d = 0.31, p = .004)

The WHO-5 is one of the most widely validated instruments in occupational health research. Low WHO-5 scores predict subsequent sickness absence, long-term disability claims, and turnover across multiple longitudinal studies (Topp et al., Psychotherapy and Psychosomatics, 2015). When the WHO-5 moves up, downstream absence rates move down.

Most workforce mental health programs are designed to meet people once symptoms are present — through utilization, sessions delivered, and crisis support. These are essential. They're also, by design, measured at a stage of need when costs have already begun to accumulate.

The metrics Thoughtful moves sit earlier in that progression. They are leading indicators rather than downstream recordings — which means they offer a different kind of leverage: the chance to influence a cost curve before it forms, rather than respond to one already in motion.

That's what catching people before they're symptomatic actually means, in dollars.

1. Who is actually using what you offer? Most employers have a utilization figure. Few have visibility into whether engagement is concentrated in a small group of high-engagers, or distributed across the workforce.
2. What happens between awareness and action? Most HR programs have no visibility into the journey from "I need help" to a first appointment. The drop-off point matters more than the awareness rate.
3. Are you reaching people before crisis? EAP engagement tends to cluster around acute moments. The subclinical population is rarely captured in utilization data.
4. What is the counterfactual for non-engagers? A growing proportion turn to general AI tools with no clinical oversight. Understanding that counterfactual matters for benefits strategy.
5. Are you measuring the right outcomes? Symptom-only measurement misses the psychological functioning shifts that predict recovery.