How Unmind's practitioner network is shaping the way AI shows up in care

People are turning to AI for their mental health. HBR's 2025 Gen AI report stated that therapy/companionship is now the number one use case for AI/ChatGPT. In our own research, close to a third of the people we surveyed told us they'd used a general-purpose AI tool for wellbeing support. But general-purpose AI tools were never built for this job.

That's the reality we're designing for. Rather than leave people with chatbots that weren't made for wellbeing, we built Nova – Unmind's own AI wellbeing coach. Now, to shape how it's used and where it goes next, we're bringing in the people who understand care best: our network of accredited therapists and coaches.

Nova is a complement to therapy and coaching, never a replacement. It's built for wellbeing – a private space for reflection and journaling, exploring values and goals, preparing for a session, or working through everyday workplace challenges like stress, performance, or a difficult conversation. It’s not there to diagnose or treat mental health conditions. When someone needs one-to-one human support, Nova's role is to help them find it. That boundary isn't a caveat bolted on at the end. It shaped the whole design.

Nova's development and usage is led by our own in-house clinical team – Dr Max Major, our Head of Clinical AI, and Becky Minton, our Head of Global Care – working alongside our product and safety teams. That gives Nova clinical rigour from the inside.

But we wanted more than our own view. So we opened it up to independent scrutiny from the practitioners who'd actually use it. We invited our practitioner network to test Nova in the context of their own client work, where appropriate, and to tell us honestly what worked and what didn't – through structured surveys and focus groups. Their feedback fed directly into how we built Nova: the examples we share, the boundaries we set, and the resources we've built to support practitioners in suggesting it.

The practitioners who tested it rated Nova 4.1 out of 5, with more than 9 out of 10 scoring it four or five. The main detractor was wanting more information on how and when to use it - which we’ve now developed.

The practitioners who tested Nova weren't in one place. Those who gave feedback through our surveys and focus group spanned 15 countries across six continents – from the UK and Germany to Mexico, Chile, the USA, the UAE, Singapore, Hong Kong, Kenya and Australia. That range matters. What feels supportive, and how people talk about wellbeing, shifts with language and culture. Testing across borders helps us see where Nova works well and where it still needs to improve – something a single-market trial would miss.

Take-up was driven by curiosity, not obligation. Practitioners had hundreds of conversations with Nova, and, on average, spending longer with it than a typical user does. What came through was genuine enthusiasm for Unmind building this thoughtfully, and for the chance to support their clients with AI in a way they could trust.

Practitioners have woven Nova into their work in different ways: suggesting a grounding or breathing exercise before a session, asking a client to gather their thoughts in Nova beforehand, reinforcing goals and values between sessions, or offering it as continuity as a course of care comes to an end. As one of our practitioners put it:

"I'm really happy with Nova, as it complements my work with clients. I always recommend that they use it. Sometimes I'll give clients a task after our session – for example, asking Nova about ways to become more assertive – and then we discuss what came up in the next session. My clients find it very useful." Coach, UK

Another described a client who tried Nova's suggestions before their first session: even where the suggestions didn't land, it gave them something concrete to explore together, and saved time. A third summed up the appeal simply – "I can direct clients to this rather than other AI-based services." That's the model we're building towards: practitioner-led, user-owned, and always in service of the human relationship at the centre of care.

Nova is one part of a connected Unmind experience – self-guided tools, educational content, an AI wellbeing coach, and human therapy and coaching – designed so people can move between them as their needs change. A practitioner can suggest Nova to extend the work between sessions; Nova, in turn, points people towards booking human support when that's what they need. The aim is one continuous journey, not a set of disconnected tools.

Nova was built to different standards than a general AI tool. It doesn't train on client inputs. Conversations are private to the client and aren't shared with their practitioner. It has built-in safety escalation. And practitioners always retain professional discretion over whether to suggest Nova at all, and for whom.

Anyone can add a chatbot to a wellbeing app. What's harder – and what matters – is developing AI with the clinicians who'll stand behind it, grounding it in evidence, and holding a clear line between wellbeing support and clinical care. That's the difference between a tool people stumble into and one their therapist or coach can thoughtfully stand behind.