Infrastructure, Not Apps: What This Mental Health Month Is Really Telling Us

May 25, 2026 · Dr. Ramy Azzam

Every May the sector I work in fills with awareness campaigns, and most years I find them genuinely moving. This year I find myself wanting to say something less comfortable. The data underneath the campaigns has turned, and it has turned in a direction that should make us question the story we have been telling ourselves about technology and mental health for the past decade.

The story went like this. Mental health care is scarce, expensive, and stigmatised, so we will use technology to scale it. Apps will reach the people that clinics cannot. AI companions will be there at 3 in the morning when no human is. We will close the gap with software. It was a beautiful story, and I believed a version of it. I still believe parts of it. But the numbers from this year are quietly dismantling the simplest version of it, and we owe it to the people we are trying to serve to notice.

A survey of more than 2,000 adults this year found anxiety up more than 9% and depression up more than 10%, in a single year. More than half of people reported real loneliness. And the share of people actually reaching professional help fell, from 1 in 2 to fewer. We have spent 10 years pouring apps into the gap, and the gap got wider, not narrower. That is not a marketing problem. That is a signal that we have been building the wrong thing, or at least building it in the wrong order.

A therapist listening attentively to a client in a bright office, representing the human professional support that technology should point people toward.

The Finding That Should Stop Us

The piece of research from this period that I cannot stop thinking about came out of MIT, and its conclusion is uncomfortable for everyone in my field. Relying on AI alone for emotional support, the researchers found, can make loneliness worse rather than better for some people. Not neutral. Worse.

When I sit with that finding, it stops being surprising. A system built to be endlessly agreeable, that never tires of you, that asks nothing of you in return, offers something that feels like connection and is missing the ingredient that makes connection nourishing. Real relationship has reciprocity and friction in it. The other person is also struggling, also limited, also reaching back. That mutuality is not the cost of human connection. It is the substance of it. A companion engineered to remove all friction has, without meaning to, removed the medicine along with the discomfort.

This matters for how we build, because a great deal of consumer mental-health technology has been optimised for exactly the wrong thing. Engagement. Time in app. Daily active users. We measured success by how much people leaned on the product, when the honest measure of success for anything in this space should be whether the person is more connected to their actual life, including the people in it, as a result. Those 2 metrics are not the same. Often they are opposites.

One woman supporting and comforting another, representing the reciprocal human connection that technology can point toward but cannot replace.

Why I Keep Saying Infrastructure

When colleagues ask what I think the region should be building in mental health, my answer has become almost monotonous. Infrastructure, not apps. And I want to be precise about what I mean, because it is easy to hear that as a slogan.

An app is a product you download. Infrastructure is the scaffolding that makes the whole system trustworthy and safe, whatever products sit on top of it. In mental health, infrastructure means the things that almost no consumer app provides on its own. Clear pathways from a digital tool to a human clinician when a situation escalates. Standards that distinguish a carefully built product from a carelessly built one, so a frightened person at midnight is not left to guess. Data protections strong enough that people can be honest without fear. Regulatory scaffolding that defines what a tool is allowed to claim and what it must do when someone is in danger. And a workforce of real human beings, adequately resourced, that the digital layer can actually hand people to.

We have been building the top of that stack, the shiny app layer, while leaving the foundation thin. That is precisely backwards. An app that cannot reliably escalate to a human, in a system where the human capacity does not exist anyway, is not closing the gap. It is papering over it, and the paper is starting to tear.

The Gulf Has a Real Choice Here

I think about this most in the context of the Middle East, because the region is building its mental health systems now, in real time, and has a genuine choice that older systems no longer have. The temptation, with significant investment flowing into digital health and AI under Vision 2030 and the wider regional strategies, is to skip straight to the impressive part. To launch the companions and the chatbots and the consumer apps, because they demo well and scale cheaply, and to defer the unglamorous infrastructure for later.

A panorama of a modern Gulf city skyline under a clear sky, representing the region's opportunity to build mental health infrastructure deliberately rather than inherit others' mistakes.

That would be a mistake, and an avoidable one. The region has the rare advantage of watching what happened elsewhere when the app layer ran ahead of the infrastructure. It saw the chatbots that presented themselves as licensed therapists to vulnerable people. It is watching American states ban behavioural-health AI outright because the harms arrived before the guardrails did. It does not have to repeat that sequence. It can build the scaffolding first, the escalation pathways, the standards, the human capacity, the cultural and linguistic fit that a region this diverse genuinely requires, and then let the digital tools sit on top of something solid.

A mental health system built in that order would not be slower to help people. It would be the first system in the world built with the lesson already learned. That is a form of leadership the region is well positioned to claim, if it resists the pull toward the easy, visible win.

What This Means for Those of Us Who Build

I founded CIGMA, and its companion MOA, as a social-impact wellness platform, and I have tried hard to build it against the grain of the engagement trap. MOA is designed to point people back toward their lives and their relationships, not to become a substitute for them. It is built to recognise when something is beyond its scope and to help a person reach a human. Through EthicaLabs I work with organisations on the governance that makes any of this safe, and after 13 years in digital health, with WhatsHealth now taking shape, I have watched enough hype cycles to be wary of the one we are in.

What I would say to anyone building in this space is this. The metric matters more than almost anything else you choose. If you measure success by how dependent your users become on your product, you will build something that quietly worsens the very loneliness you set out to solve, and the data this year suggests a lot of us already have. If you measure success by whether the person is more connected to their actual life, you will build something different, something that sometimes encourages people to close the app and call someone, and that is willing to be needed a little less. The second kind of product is harder to build and harder to monetise. It is also the only kind worth building.

The Sentence Worth Holding Onto

Here is the sentence I keep returning to this month. The goal was never to make people need the technology more. The goal was always to help them need it less.

An app that becomes someone's only relationship has failed, even if every engagement metric is green. A system that hands a frightened person to a real human at the right moment has succeeded, even if that means the person closes the app and does not come back for a while. We confused those 2 outcomes for a decade, and the loneliness data is the bill coming due.

This Mental Health Awareness Month, I am less interested in awareness than in architecture. The awareness is not the problem. Almost everyone now knows that mental health matters. What we lack is the infrastructure to turn that awareness into help that reaches people safely. Building that scaffolding is slow, expensive, and largely invisible work. It is also the work that actually matters, and the region that builds it first will have done something the rest of the world is still struggling to do. If you are working on that foundation rather than the layer that sits on top of it, those are the conversations I most want to be in.