Something quiet is happening inside your organization. Your employees are not raising their hand to their manager.

They are not calling the EAP line. They are opening ChatGPT, typing out a fear about job security or a night they could not sleep, and treating whatever comes back as support. That is not a hunch. A 1,545-person Harvard Business Review study published this spring backs it up, and it names one of the least-discussed people risks HR is carrying into the second half of 2026 (Hadley and Wright, Harvard Business Review, 2026). If you lead People or HR and do not yet have a response to it, the data below should move you to build one.

What Is the Real Cost of Employees Turning to AI for Support?

74% of employees already use AI for at least one form of support that used to come from a person: career guidance, venting, or plain companionship (Hadley and Wright, HBR, 2026). The instinct to seek support is healthy. The channel most people are defaulting to was never built to carry it.

The HBR Finding Every CHRO Should See

Organizational psychologist Constance Noonan Hadley of the Institute for Life at Work and Professor Sarah Wright of the University of Canterbury surveyed more than 1,545 U.S. knowledge workers. Their finding: 74% already use AI for career guidance, emotional venting, or companionship. Despite that, 52% still report high or moderate loneliness at work, and only 12% said AI made them feel less lonely (Hadley and Wright, HBR, 2026).

Hadley and Wright's conclusion is direct. AI cannot replace human connection, and over time it may erode the collaboration, trust, and social skills that high-performing teams depend on.

This is not a technology failure. It is a design problem. Employees reach for the most frictionless tool available because the human support infrastructure at most companies is invisible, hard to access, or quietly stigmatized. Where there is a vacuum, people fill it. Right now, they are filling it with AI.

Workday's Human Connection Workplace Index, surveying 2,150 workers across large enterprises, found the same pattern from a different angle. Among Gen Z workers, 20% took time off this year due to loneliness or isolation, and more than one in five said AI tools had made their relationships with colleagues worse (Workday, 2026). Gen Z workers are also 12 times more likely than Gen X to say they feel completely disconnected from their colleagues. That is not a soft issue. It shows up as attrition risk, engagement risk, and culture risk, all at once.

The burnout data compounds it. Spring Health's 2026 Workplace Mental Health Annual Report, drawing on more than 1,500 full-time employees, found that 74% have experienced burnout and 61% of HR professionals say burnout increased over the prior year. HR leaders estimate that roughly 30% of employees are living with "silent burnout," the invisible kind where someone looks fine and is privately running on empty (Spring Health, 2026). These are not people one wellness webinar away from recovery. They need real support built into how their week already works.

Does AI Coaching Actually Work?

AI coaching tools genuinely have gotten better. But the research is consistent on where they hit a ceiling: the relational, values-based work that drives lasting change still belongs to a skilled human (Passmore, Tee, and Rutschmann, 2025).

The Trust Gap No Algorithm Has Closed Yet

Professor Jonathan Passmore of Henley Business School, working with University of Cambridge colleague David Tee, ran ICF-trained assessors against transcripts of an AI coaching agent. The agent demonstrated elements of both ICF Associate and Professional Certified Coach-level competencies in structured, task-focused sessions (Passmore, Tee, and Rutschmann, Coaching: An International Journal, 2025). That's a real result. It also is not the whole story: Passmore's broader research argues that AI is well-suited to habit tracking and progress monitoring, while the deeper transformational work still requires a human across the table (Passmore et al., International Journal of Training and Development, 2025).

Amber Barger of Teachers College, Columbia University, tested something more specific: can a person build genuine working alliance, the strongest known predictor of coaching effectiveness, with what they believe is an AI coach? In her randomized controlled trial, one group of 26 talked to a human coach. Another group of 26 talked to what they believed was an autonomous AI, though it was actually a trained human coach speaking through an interface. Both groups built comparably strong alliance, human average 74.5, believed-AI average 72.7, a difference that was not statistically significant (Barger et al., Frontiers in Psychology, 2025).

That finding matters more than it first appears. It suggests people can form real rapport with something they think is AI, at least in a single structured session. But it does not tell us what happens over months, under real pressure, when the conversation gets harder than a first session ever does.

A separate, sharper test came from Passmore and colleagues in a quasi-experiment with 63 white-collar workers, 28 assigned to an AI coaching agent and 35 to a human coach (Passmore et al., Journal of Work-Applied Management, 2025). Here is what the data actually showed:

That is a more interesting finding than a clean "humans win everything" headline. People trusted the AI coach and disclosed to it about as readily as they did with a human. Where the human coach pulled ahead was in alliance and commitment, the two measures most tied to whether someone actually sticks with the work over time. Trust gets someone through the door. Alliance and commitment are what keep them coming back when the work gets uncomfortable.

The Conference Board reached a compatible conclusion at scale: AI can provide up to 90% of day-to-day coaching functions, but human expertise remains critical for emotionally charged, political, or values-based conversations (The Conference Board, 2025). That remaining 10% is frequently the one conversation that changes something for a person.

Why Human Coaching Builds Capacity and Composition That AI Cannot

You already believe your people deserve real support, not a chatbot standing in for one. Most HR teams already have that conviction. What they are missing is the system that makes it operational: visible, fast to access, and staffed by someone who actually knows the person on the other end.

What Mental Fitness Actually Trains

That is what Mental Fitness training is built to be: a proactive, trainable set of skills that strengthen how people think, regulate, recover, and perform under real pressure, week after week, not only in a crisis. It is not therapy, and it is not a meditation app. Forte trains four permanent pillars, and they require the same thing physical training requires: a real coach, a real relationship, and real accountability over time.

What a Forte Coach does in every session that an AI tool cannot: hold the full context of a Member's life, work, and growth over time. Notice the shift in tone when someone says they are "fine" and does not mean it. Push back when a stated goal is not the real goal. That is what makes a session bookable within 24 hours feel urgent instead of procedural. The person on the other end already knows the Member, rather than processing them fresh each time.

To see how Forte puts human coaching at the center of a full Mental Fitness system, the how Forte works page is a useful next stop.

What Should HR Leaders Do Right Now?

Employees are looking for support. They are finding AI. AI is not delivering connection, and the silence that follows shows up in Gallup's finding that global employee engagement fell to 20% in 2025, its lowest point since 2020 (Gallup, 2026). That silence is landing in your lap as a people leader, whether anyone names it out loud or not.

Here is what the research, and the HR leaders we work with, point to.

1. Name the gap out loud. Your employees do not know what they do not have. If your organization has not actively communicated that real, human, on-demand coaching is available, and made it genuinely frictionless to access, you built the vacuum AI is now filling.
2. Do not ask a technology tool to solve a human problem. AI wellness apps can play a supporting role in a well-built Mental Fitness strategy. They cannot be the center of it. The moment an AI tool sits at the center of your wellbeing program, you have optimized for accessibility at the cost of impact.
3. Invest in Capacity before you need it. The organizations that hold performance through the next wave of AI disruption and workforce change are the ones building people's resilience now, not scrambling to respond after it breaks. Capacity builds in a sustained coaching relationship over time, with someone who knows you, not in a single session.
4. Measure what actually matters. Not utilization. Not logins. Ask whether people feel more grounded, more adaptable, more able to sustain real performance. That is the honest output of a Mental Fitness program, and it shows up in qualitative outcomes and manager observation more than in a dashboard.