Unmind logo

Mental health days, subsidized therapy and mental health apps compared as workplace support

visualelectric-1753280887614.png

Content

  • What mental health days do and do not do
  • What subsidized therapy does and does not do
  • What mental health apps do and do not do
  • How to judge the quality of a mental health app
  • Layers rather than competitors
  • The real question is routing and measurement
  • A practical way to decide your mix

HR and Benefits leaders who ask which of these three options works best are usually starting from the wrong question. Mental health days, subsidized therapy and mental health apps do different jobs. The first gives time, the second gives clinical care, and the third gives support that is available at any hour. The useful decision is how to combine them, who each one reaches, and how you will know whether people are using them.

This post sets out what each option does and does not do, how to judge the quality of a mental health app, and how to think about the mix as a set of layers with routing between them.

What mental health days do and do not do

A mental health day gives an employee time away from work and, just as important, permission to take it without explaining themselves. That permission matters. A common barrier to rest is the worry that taking it will look like a lack of commitment, and a clearly worded policy removes some of that worry.

What a mental health day does not provide is care. It has no clinical content, no guidance about what to do with the time and no route to anything further. Someone who is simply depleted may need exactly this. Someone with a persistent difficulty can take the day and return to the same problem on Monday.

Two practical limits are worth planning around. Uptake can depend on manager behaviour, so if managers quietly discourage people from using the time, the policy exists on paper and nowhere else. And the benefit is hard to measure. Absence data tells you that days were taken, but it cannot tell you whether anyone was supported.

Mental health days belong in the mix as a foundation of time and permission, and they work best when something sits behind them.

What subsidized therapy does and does not do

Subsidized therapy is the option that provides clinical care. Talking with a qualified therapist is the right response to many of the difficulties employees bring, and a subsidy lowers one of the biggest obstacles to starting. It is the part of any support offer that treats people.

It also has structural limits that are easy to overlook once a benefit has launched. Access depends on therapist availability, language, specialization and location, so a subsidy does little for an employee who cannot find a suitable person. Stigma keeps some people away, because booking therapy is a deliberate and visible step and many who would benefit do not yet think their situation is serious enough. And self-selection shapes who appears in your data. Usage can skew towards people who already recognise they need help and are comfortable asking for it.

That last point affects measurement. If only the already-motivated come forward, usage figures can look healthy while part of the workforce is never reached. Therapy is essential for those who need clinical care, but it cannot be the whole offer, because it is poorly suited to the earlier, quieter stages when people are not ready to book an appointment.

What mental health apps do and do not do

Apps fill the space the first two leave open. They are available at any hour, they are private, and they let someone start with something small, such as a short exercise, a course or a conversation, without booking or disclosing anything to anyone. For employees who would never contact a therapist, that low barrier is the point.

The weaknesses are structural too. Quality varies widely. Engagement can tail off over time without prompting. And a self-guided tool cannot provide clinical care. If an app presents itself as a substitute for therapy, treat that as a warning sign.

That variation in quality is the reason apps deserve more scrutiny from buyers than the other two options.

How to judge the quality of a mental health app

You do not need a clinical background to run a sensible first check. Three areas cover most of it.

Start with public evidence. Look for published research, named authors or institutions, and a clear description of what was measured. Marketing statistics with no source do not count. There is a practical walkthrough on how to determine whether a mental health app is evidence-based in about ten minutes, and the wider context for why buyers are asking harder questions is covered in a piece on why mental health apps are under scrutiny.

Next, check whether the claims are clear and appropriately modest. A well-run product says plainly what it is for and what it is not for. Be cautious of any app that implies it can diagnose, treat or replace professional care. Look for a stated boundary between self-guided support and clinical help, and a defined route to human care when someone needs it.

Finally, ask about safety and privacy practices. How does the product handle risk? What happens when a user raises something serious? Who can see usage data, and how is employer reporting anonymised? If the product uses AI, ask how it is trained, audited and constrained. Unmind sets out its approach on its AI trust and safety page, and any vendor you consider should be able to give you the equivalent in plain language.

If a vendor cannot answer these three areas clearly, that tells you something before any demo does.

Layers rather than competitors

Seen side by side, the three options are not alternatives for one budget line. They cover different points on the path from feeling under pressure to needing clinical help. Time and permission come from mental health days. Private, immediate, self-guided support comes from apps and learning content. Clinical care comes from therapy.

Each layer compensates for a gap in another. Days off do nothing for someone who needs treatment, therapy does not reach those who will not book it, and apps cannot treat a clinical condition. A mix with only one layer leaves a predictable group of employees without anything that fits.

The need for range rises when an organisation is under strain. Unmind holds that 72% of transformations fail not because of strategy or technology, but because they ignore the human cost, and periods of restructuring, new technology and shifting priorities are exactly when the earlier, quieter layers matter most.

The real question is routing and measurement

If the layers are complementary, the difficulty moves. The question becomes how an employee finds the right layer for their situation, and how you will know whether that is happening.

Routing matters because many people may not know which layer they need. Someone with a low mood and a heavy workload may not realise that a conversation with a therapist would help more than another afternoon off. A well-designed offer gives them one place to start and a way to step up when that is warranted.

Measurement matters because availability and use are different things. The average EAP sees under 2% engagement, and a legacy EAP running at 2 to 4% utilisation provides legal cover rather than care. A benefit that exists but is rarely used protects the employer on paper and does little for employees. Whatever mix you choose, ask for utilisation by layer, how people move between layers, and whether the pattern changes over time.

This is the thinking behind the approach at Unmind. Nova is an AI mental health agent that gives confidential support and routes people to human care when they need it. It does not diagnose, does not treat and is not a crisis service, so its job is to be an easy first step and a clear pointer onward. Therapy & Coaching is on-demand access to qualified therapists and coaches, which the employee books with filters for language and specialization. Content and Courses provides science-backed, self-guided learning for people who want to build practical skills. Together they cover the self-guided and clinical layers, with a route between them.

The platform-wide figure Unmind states is a 4.6x return per $1 invested. Separately, in a six-month study at one global consumer-goods employer, the projected return was 3.49x, with an estimated productivity impact of around $1M. Among high-stress employees in that study, stress fell by 16% and work impairment fell by 10%. The study figures are separate from the platform-wide figure and are not interchangeable with it, and your own workforce will differ, which is why you should measure your own results rather than assume them.

A practical way to decide your mix

Before choosing anything, work through five questions with your team.

  1. Who is not being reached today? Look at who uses your existing benefits and who does not, by team, location and seniority.
  2. What does an employee do on day one? Trace the path from "I am struggling" to "I have spoken to someone". If it takes more than a few steps, simplify it.
  3. Do managers know how to point people onwards? A policy that managers do not support rarely works.
  4. How will you separate use from availability? Agree the measures before launch, including utilisation by layer and movement between layers.
  5. Can each vendor answer the quality questions? Public evidence, clear claims, and safety and privacy practices should all be answerable without a long delay.

With those answers, the shape of the mix tends to follow. Time and permission as a baseline, a low-barrier self-guided layer for early support, and clinical care that is easy to book for those who need it.

If you are weighing how to route employees to the right level of support, start with Nova from Unmind to see how a confidential first conversation leads on to human care. For a wider view of how AI fits into this picture, the handbook on the AI advantage for mental health is a good next read.