In mental healthcare, we have become very good at digitising what already exists.

Paper worksheets became apps. Psychoeducation became online content. Appointment letters became text reminders. Face-to-face therapy became video therapy. Structured psychological interventions became guided digital programmes. More recently, artificial intelligence has arrived, promising to transform almost everything again.

Each of these developments has brought progress. It would be wrong to dismiss the gains. Digital mental health has helped services reach people who might otherwise have waited longer, travelled further, or never engaged at all. It has given clinicians new ways to extend care beyond the clinic room. It has created more flexible routes into support. In some cases, it has reduced stigma by allowing people to take a first step privately, quietly, and at a time of their choosing.

From Digitising Care to Rethinking Care

But I increasingly think we are reaching the limits of the first era of digital mental health.

For the last two decades, much of the conversation has been about how technology can make existing models more efficient. How do we deliver therapy online? How do we automate parts of assessment? How do we support self-management between appointments? How do we reduce administrative burden? These are all necessary questions, but they begin from the same assumption: that the existing model of care mainly needs a better delivery mechanism.

The next generation of mental health technology will ask a more uncomfortable question. What if technology does not simply deliver the intervention? What if, in some cases, the interaction with the technology becomes part of the intervention itself?

That is the shift we are beginning to see. The future of digital mental health will not be defined simply by apps, video consultations, AI chatbots, virtual reality, games or digital therapeutics as separate categories. It will be defined by the emergence of interactive mental healthcare: care that is adaptive, personalised, conversational, experiential, responsive to behaviour and increasingly available at the moment people need support.

This distinction matters because mental health systems are now facing a problem that incremental digitisation cannot solve on its own. Demand is rising, complexity is increasing, and the workforce cannot grow quickly enough to meet the scale of need. Where I live in England, the Royal College of Psychiatrists reported in February 2026 that more than 550,000 children and young people referred to mental health services were on waiting lists for treatment. NHS Talking Therapies continues to support very large numbers of people with anxiety and depression, but the service is also operating under significant performance pressure; NHS Digital reported that 45.8% of eligible referrals reliably recovered in December 2025, below the NHS England target of 48%.

None of this is a criticism of clinicians or services. It is the opposite. The current workforce is carrying a level of demand that no analogue system was designed to absorb. The question is not whether mental health services need more staff. They do. The question is whether staffing growth alone can solve the problem at the scale, speed and consistency now required. I do not believe it can.

That means technology cannot remain a marginal add-on to traditional care. It has to become part of how we rethink capacity, quality and access together.

The risk is that we respond to this challenge by commissioning more digital replicas of analogue services. More content libraries. More self-help portals. More static pathways. More systems that digitise the front door but leave the model of care largely untouched. These tools may help, but they will not fundamentally change the care equation. A service with too few clinicians, too many referrals and long waiting lists does not become transformed simply because the waiting room has moved online.

The Rise of Interactive Mental Healthcare

Interactive mental healthcare offers a different possibility.

In an interactive model, the technology does not simply present information. It responds. It can adapt the pace, tone and content of support according to how someone is engaging. It can help someone practise a therapeutic skill rather than just read about it. It can provide repetition without exhausting a clinician. It can create structured experiences that allow people to rehearse coping strategies, emotional regulation, social confidence or relapse prevention. It can gather signals over time, notice changes, and support escalation when risk increases. It can make care less episodic and more continuous.

This is why the convergence of AI, immersive technology, games, digital therapeutics and behavioural science is so important. AI brings personalisation and dialogue. Games bring motivation, feedback, agency and repetition. Immersive technology brings experiential practice. Digital therapeutics bring structure, evidence and intended use. Behavioural science brings the mechanism of change. Taken together, they begin to point towards a new class of mental health intervention.

But this also creates tension, because our systems are still built around old categories.

Healthcare likes things to fit neatly. A product is either a medical device or it is not. It is either treatment or wellbeing. It is either clinician-led or self-guided. It is either a regulated intervention or a consumer product. It is either therapy or not therapy. Interactive mental health technologies do not always behave so cleanly.

An AI companion may begin as guided self-help but become a risk interface when a person discloses suicidality. A virtual reality mindfulness tool may begin as wellbeing support but become clinically relevant when deployed as an intervention for depression or anxiety. A game-based intervention may look like a way to simply drive engagement, but if it changes behaviour, mood or functioning, then it starts to raise therapeutic questions. A system that personalises its output over time may be more useful than static content, but it is also harder to validate, harder to explain and harder to govern.

That does not mean we should avoid these technologies. It means we need to become much more sophisticated in how we commission, evaluate and regulate them.

The global policy environment is already moving in this direction for AI more broadly. The WHO Regional Office for Europe has highlighted that AI is reshaping health systems and that countries are developing governance models, data strategies, legal frameworks, workforce plans and regulatory approaches to support adoption. Its 2026 report found that all 27 EU Member States considered improving patient care a major or moderate driver of AI innovation, while 26 of 27 considered reducing workforce pressure a major or moderate driver. The same report also emphasised the need for regulatory sandboxes, assurance laboratories, clearer accountability rules, ethical design and real-world monitoring to enable safe implementation.

According to a recent IPSOS study, Mental health is now the number one health problem, ahead of cancer and coronavirus. Mental health should therefore be at the centre of this conversation, not at the edge of it. Few areas of healthcare are more dependent on relationships, trust, language, context, engagement and behaviour change. Few areas are more vulnerable to harm if technology is deployed badly. But equally, few areas have more to gain if technology is developed and commissioned responsibly.

Why Commissioning Must Evolve

This is where commissioners need to become bolder.

Bold does not mean reckless. It does not mean ignoring clinical safety, data protection, medical device regulation or evidence. Quite the contrary. It means recognising that if we only commission technologies that fit perfectly into yesterday’s compliance categories, we will never build tomorrow’s models of care. Regulation and governance should protect patients, but they should also create safe ways for systems to learn. The answer to uncertainty cannot always be delay. Sometimes the answer has to be a well-governed pilot, a real-world evaluation, a sandbox, a staged deployment, a clear escalation model, and a willingness to learn in partnership with innovators.

At the moment, too much promising technology gets stuck in the gap between “interesting” and “commissionable”. Innovators are told to generate more evidence, but without a route to procurement they cannot generate the evidence that commissioners need. Commissioners want reassurance, but the reassurance often depends on another commissioner having gone first. Regulators want clarity of intended use, but the most innovative products often emerge precisely because old use categories no longer quite apply. The result is a familiar trap: everyone is interested, everyone is cautious, and the patient waits.

Tend: A Glimpse of What’s Possible

Tend is a useful example of why this matters.

Tend is not interesting simply because it uses virtual reality. It is interesting because it represents a more interactive way of thinking about access, engagement and therapeutic delivery. The company has adapted Mindfulness-Based Cognitive Therapy into an immersive format, aiming to deliver structured mental health support through virtual environments that people can use at a time and place that suits them. Health Innovation East describes Tend VR as a novel way to deliver MBCT, a NICE-recommended therapy for depression, with potential to reduce costs and staff time while improving access and helping to reduce waiting lists. More recently, Tend VR has worked with the Mental Health Research for Innovation Centre in Liverpool to involve service users and carers in shaping a VR-based mindfulness and therapeutic relaxation tool for people experiencing mental distress, including difficult-to-treat depression. The most interesting point to make though is how Tend compares to traditional human-led therapy. On all metrics,Tend shows signs of equal or better clinical results. Patients engage longer on Tend Programmes meaning reduced wait lists. Patients have higher reliable recover rates with Tend meaning better quality care. 

The important point is not that Tend is the whole answer. No single company is. The important point is that it shows the kind of boundary-stretching innovation mental health systems need to learn how to support. It is not a conventional therapy appointment. It is not simply an app. It is not just content. It is an interactive therapeutic experience that raises precisely the questions commissioners now need to become comfortable asking. What is the intended use? Who is it for? Where does it sit in the pathway? What level of clinical oversight is needed? What evidence is proportionate? What risks need to be monitored? How does it escalate? How does it improve quality, access or productivity compared with the current model?

These are not reasons to avoid commissioning. They are the right questions to ask before commissioning responsibly.

In fact, this is exactly the kind of innovation that may help solve the workforce problem without pretending that technology can or should replace clinicians. The future is not a choice between human care and digital care. That is a false and unhelpful binary. The future is about using interactive technologies to extend the reach, consistency and quality of human-led systems. A clinician should not have to personally deliver every moment of psychoeducation, skills practice, relaxation training, relapse prevention or between-session support. Their time should be focused where human judgement, relational skill and clinical responsibility matter most.

If interactive tools can help people practise skills earlier, stay engaged longer, receive support between appointments, step up or down safely, or avoid deterioration while waiting, then they are not peripheral. They are part of the capacity strategy.

Innovation Requires Better Governance

But quality matters. There is a real danger that the current excitement around AI and digital mental health leads to a flood of poorly evidenced, poorly governed tools entering the market. Recent public debate around AI therapy chatbots has rightly raised concerns about safety, over-reliance, weak evidence and the handling of serious risk. Those concerns should be taken seriously. Mental health technology is not like consumer productivity software. People using these tools may be distressed, isolated, traumatised, neurodivergent, suicidal, medicated, or in the middle of a crisis. A poor user experience is not merely inconvenient. It can be clinically significant.

That is why the next phase must be both more ambitious and more disciplined.

We need stronger clinical governance from the beginning of product design, not as a procurement afterthought. We need clearer evidence pathways that recognise the difference between low-risk wellbeing tools, structured psychological interventions, AI-supported care navigation and higher-risk therapeutic systems. We need real-world monitoring that looks beyond whether a product “works” in a study and asks how it behaves in messy service conditions. We need commissioners to evaluate impact not only in terms of symptom change, but also engagement, adherence, escalation, workforce productivity, access equity, dropout, deterioration and patient experience.

We also need to stop treating compliance as a static hurdle. For interactive technologies, compliance has to become more like a living discipline. Products that adapt, converse or respond to user behaviour need ongoing assurance. They need post-deployment monitoring. They need clear boundaries around what they can and cannot do. They need clinical safety cases that evolve as evidence grows. They need procurement models that allow learning without leaving innovators trapped in endless pilots.

The next frontier in mental health technology will not be defined by whether something is AI, XR, a game, a chatbot or a digital therapeutic. Those labels are already becoming too narrow. The more important question is whether the technology creates a better form of care. Does it help someone access support earlier? Does it help them practise skills more effectively? Does it give clinicians better information? Does it reduce avoidable deterioration? Does it improve quality while enhancing capacity? Does it allow services to support people who would otherwise remain unseen?

If the answer is yes, then our job is not to force that innovation back into an old category. Our job is to build the governance, evidence and commissioning pathways that allow it to be tested safely and scaled responsibly.

The Courage to Build What’s Next

Mental healthcare cannot afford to be timid. The current model is not meeting need at the scale required. Waiting lists are too long. Staff are too stretched. Too many people receive support late, inconsistently or only once they have deteriorated. We should not use regulation as an excuse to preserve a system that is already failing too many people. Equally, we should not use innovation as an excuse to weaken safeguards that exist for good reason.

The task is to hold both truths at once.

We need more safety, not less. More evidence, not less. More clinical governance, not less. But we also need more courage. Courage to commission technologies that do not look exactly like the services we already know. Courage to create sandboxes and testbeds where regulators, clinicians, patients and innovators can learn together. Courage to fund real-world evaluations rather than waiting for perfect evidence to emerge from nowhere. Courage to accept that the future of mental healthcare may not resemble the past.

The first era of digital mental health digitised existing care.

The next era should make care more interactive, adaptive, human-centred and scalable.

If we get this right, interactive mental healthcare will not replace clinicians. It will help clinicians deliver better care to more people, earlier, with greater consistency and reach than today’s systems can manage alone.

That is the opportunity.

Not simply to digitise mental healthcare.

To reimagine what mental healthcare can become.