AI in Mental Health Forum 2026 Recap: Keynotes and Panels
- Will Zhang
- 2 days ago
- 11 min read
Introduction
The AI in Mental Health Forum 2026, held on 25 June at the Clinical Education Centre, Auckland City Hospital, was the second annual gathering of the AI Health Innovation Hub (AHIH), presented by EmoEx. Around 100 clinicians, researchers, policy leaders and practitioners came together for a day of keynotes, panels and seminars.
The questions have stopped being hypothetical. AI is already in mental health settings, and the question this forum is asking is not whether or when, but how and by whose standards.
This is the first of two posts covering the 2026 Forum. This piece covers the morning keynote addresses and the two panel discussions. The four practitioner seminars are covered in a companion post, The Seminar Sessions.
Keynotes and Panels:
Welcome Address
Will Zhang, Co-founder and CEO, EmoEx | Co-chair, AI Health Innovation Hub
Will Zhang welcomed attendees to the second annual AHIH gathering, noting the contrast with last year, when many invitees had replied saying they were not really doing anything with AI. "This year is very different. A few of those people are here today. So either the follow-up emails worked, or the world has changed very quickly. I think it is probably both."
He was direct about pace: "That shift has been fast. Maybe too fast in some places. And that is why a forum like this matters. We need spaces where the conversation is not driven by hype, fear, or sales pitches, but by practice, evidence, ethics and human need."
Will described AHIH as a collaborative ecosystem bringing together clinicians, researchers, universities and technologists, with EmoEx providing core technology infrastructure for clinical teams to explore, test and evaluate AI-enabled solutions before anything reaches the public. The governing principle: "Clinical knowledge should shape the technology, not the other way around."
On Heartline Connect, he explained that EmoEx's early focus on AI companions for everyday wellbeing made clear that digital support cannot be the end of the journey. Heartline Connect is a consent-based care coordination layer helping people move from digital support to matched human care, giving clinicians relevant context when the user chooses to share it.
He closed with the day's purpose:
Somewhere in this room, someone has worked through a problem you are still sitting with. In mental health, the cost of working in isolation is not just inefficiency. It is the person still waiting, the clinician still guessing, the family still searching.
AI Triage in Practice: Building the Front Door Without Losing the House
Anna Campbell, Chief Support Services Officer, Whakarongorau Aotearoa
Anna Campbell opened by reframing a commonly used metaphor: "Building a front door is really easy. Making a door that opens for everyone to be able to walk in, in a house that has multiple rooms that serve all of those people — that is the real challenge."
Whakarongorau Aotearoa operates New Zealand's national telehealth infrastructure, including the 1737 mental health line, handling around 7,000 contacts every day. Anna was clear that volume is not the core problem: "It's complexity, and the time it takes to try and serve those people who need help on their wellbeing journey." Contacts over the last few years now take fifty percent longer to resolve, placing significant cognitive load on kaimahi who are simultaneously holding a live conversation, conducting a clinical risk assessment, and documenting in real time.
The organisation's guiding question: "How do we help the helpers to be more human in their interactions, using technology to support them?"
Anna described two AI builds currently in development. The first, Welcome, is a conversational AI tool for the 1737 SMS and Family and Sexual Harm webchat channels. It greets people when they contact the service, gently captures some demographics, and holds them until they are connected with a counsellor. It is deliberately nonclinical and does not triage or alter queue priority, though it does prompt people to call 111 if immediate risk words are detected. Service user data has not been used to train the tool. The second, the Mental Health Navigator, is a broader navigation platform funded through the Mental Health Innovation Fund, designed to help people find the right support across the sector, retaining their choice and control throughout.
On governance, Anna was emphatic: "If I had ten hours to spend on something, I would probably spend eight of those on getting the governance and the structures right. Because once you've got those, everyone's aligned and the innovation can just flow."
She closed with three principles she encouraged the room to take away: design digital first but stay human where it matters; treat governance as an enabler of innovation rather than a blocker; and treat equity as architectural rather than aspirational.
"If you don't design with the people these tools are designed to serve, you could leave populations behind. And we already know there is a significant digital divide in New Zealand."
The Mental Health and Wellbeing Outlook for Aotearoa
Hon Matt Doocey, Minister for Mental Health and Associate Minister of Health
Minister Doocey had been due to attend in person but was called away at short notice to represent the government at another event. He recorded a video address for the Forum, which was played during the morning session.
His address focused on the government's digital health investment programme and its implications for mental health services. He pointed to the practical barriers still facing the health system: disconnected infrastructure, paper-based records in some settings, and information that cannot flow reliably between primary care, hospitals and community services. Against that backdrop, he outlined the Health Digital Investment Plan, a 10-year roadmap with $300 million committed to the first three years, structured across three phases: stabilise, modernise, and transform. The stabilise phase, where the system sits now, focuses on making critical systems reliable, removing paper from hospitals, and getting information moving between providers. The modernise and transform phases will follow, ultimately enabling AI-supported clinical work and patient-centred care at scale.
A new Centre for Digital Modernisation of Health has been established to drive the plan, shifting from fragmented projects to an integrated portfolio approach and setting common standards for identity, interoperability and clinical safety across the system.
On mental health specifically, Minister Doocey referenced the Digital Mental Health Roadmap, which he received late last year. It sets out three priority areas: an AI-enabled national navigation tool for personalised access to mental health services; a national directory of validated digital interventions to reduce wait times; and workforce capability tools to enable hybrid care models.
He closed by affirming the government's direction:
We remain focused on delivering a future-ready health system driven by innovation, strengthened through collaboration, and enabled by leading technology.
The Importance of Science-Backed Innovation: What Good Looks Like
Patrick Schweder, Consultant Neurosurgeon | Co-chair, AI Health Innovation Hub
Patrick Schweder framed his address around a single challenge: how do we move from unregulated "doctor chatbots" to evidence-based, clinically validated AI? "We can't just have anyone whip something up and have it scaled out there, because it may have significant clinical risks."
He described the appeal of generative AI in healthcare plainly: fluid, empathetic, available at 3am, and consistent in a way human clinicians cannot always be. But he was equally direct about the risks. Large language models predict the next most likely word, not the next safe word. Without careful design, they can hallucinate, generate harmful advice during a crisis, or foster parasocial dependency in vulnerable users.
The AHIH approach addresses this through scientific integrity: restricting the AI's knowledge base to peer-reviewed, evidence-ranked data; using retrieval-augmented generation to force the model to consult clinical guidelines before responding; training on therapist-patient transcripts; and hard-coding stops that redirect users to emergency services when risk is detected. You should be talking to the world's greatest expert, rather than simply a collection of knowledge.
He acknowledged the screen time paradox directly: "If we're going to encourage more screen time, are we actually undermining all the good that we're trying to do?" His suggested mitigations included hard session limits, disabling late-night push notifications, and an output-to-action principle: for every ten minutes with the tool, spend twenty minutes on real-world action.
On progress since last year, he noted that EmoEx's concussion and mild traumatic brain injury companion is now available and will be delivered through various healthcare systems, and that a multiple sclerosis companion is in development. He pointed to the global burden of disease data as the right compass for what to build next: low back pain, dementia, diabetes, oncology journeys.
He closed with a call for genuine collaboration:
If you only have engineers, development will go in one direction. If you only have doctors, it will go in another. It has to be collaborative.
Panel: Using AI in Practice in Mental Health
Panellists: Jenny Hamilton, Dr Diane Divett, Dr Brigitte Viljoen, Patrick Schweder
Moderator: Andrew Hoggard
The panel opened with a question to Brigitte about what we are getting wrong in how people interact with AI in mental health contexts. She pointed to the gap between general-purpose AI and purpose-built tools: "General-purpose AI models are being used by people for mental health support, but they are not designed for that." She flagged the risk of cognitive offloading eroding relational skills, and the importance of designing AI that asks good questions rather than gives advice: "It should be about saying, how about I ask you a series of questions to help you, as a human end user, get clear on what the issue is, where you stand on it, and what steps you want to take."
Jenny, drawing on twenty years working across grief, PTSD and neurodiversity, spoke to where AI genuinely helps clinicians and where it does not. She described AI as most useful for holding clients between sessions, supporting psychoeducation, and providing 24/7 access for people on waiting lists or dealing with conditions like autism or PTSD where in-the-moment grounding tools have real value. On boundaries: "I would not see it as replacing the therapist. A lot of harm occurs to people through human relationships, but in those cases, healthy human relationships are the greatest source of healing. That healing really has to take place in a human-to-human relationship."
Diane described her Refocusing methodology and its potential application in AI: shifting people away from problem-saturated perspectives toward a safe, resource-based space, incorporating faith and cultural dimensions. "AI could shift them to a safe place, that very defining, faith-filled, hopeful, calm space, whatever it might be for that individual, working with their language, their metaphors, their symbols."
Patrick brought a neurosurgical lens, noting AI's role in providing accessible, validated information at any hour, and calling for a tiered approach: companions for general support, treatment companions for specific rehabilitation, and hard-coded stops when conversations move into territory the AI is not equipped to handle.
On New Zealand's regulatory environment, Brigitte was direct: "New Zealand is terrible at regulating AI. They are giving very soft guidelines. There is no responsibility and accountability for what people are designing, and there are no guardrails for designers." She called for the sector to cultivate nuanced thinking:
We need to hold the paradox the whole time. There will be benefits and risks, and we need to hold both without running to one side or the other.
A question from the floor on whether New Zealand should regulate general-purpose AI from engaging in high-risk mental health topics drew agreement from the panel, with Brigitte noting the government has a responsibility to act as protector of public harm, and Patrick pointing to FDA regulation of AI as a medical device in the US as a potential model.
Panel: Creating Responsible AI Solutions for Mental Health
Panellists: Will Zhang, Jamie Allen, Stella Ward, Abbas Al-Murrani
Moderator: Andrew Hoggard
The panel opened with a question to Will about the commercial tension between keeping users engaged and what those users actually need. He was direct: "We want to drive our user away from the platform." EmoEx builds in a thirty-minute reminder prompting users to step away and connect with friends, family or a human therapist. He shared a letter from a user who had stopped using the app after six months because it had consistently reminded her to seek professional help. "At that moment I felt so good, because we actually did that."
Jamie Allen, from Taranaki Retreat Trust, offered a frank counterpoint from the NGO context: peer-led, trauma-informed services running on minimal funding, where staff are already using tools like ChatGPT on their phones regardless of policy, and where data sovereignty is not an abstract principle but a lived concern. "What are we doing with this precious taonga that we hold for people, their data?" Through a co-design process with their community, the Trust discovered deep variation in how different groups felt about data use, ultimately leading them to host their AI infrastructure locally within Aotearoa. He also noted a practical risk with AI summarisation tools: "To begin with, the team review those summaries meticulously. Then, little by little, they stop. Who is guarding the guardians?"
Stella Ward, CEO of the Digital Health Association, assessed sector readiness plainly: "I wouldn't say we are ready, which is a shame because it's here." She pointed to creaking infrastructure, ad hoc approval mechanisms and an immature regulatory environment as the key gaps, while noting active work on a medical products bill that may introduce some form of regulation for AI used as a therapeutic tool.
Abbas Al-Murrani framed the economic argument: before asking whether AI is good value, organisations need to understand their baseline, put the right outcome metrics in place, and test whether a solution that works at sandbox scale holds up when deployed more broadly. "Part of my job is taking that sandbox to the beach. Does it still work?"
On cultural responsiveness, Will was candid:
In 2023, we told our users we could speak 140 languages just because ChatGPT can. We soon took that back. Language doesn't mean you understand the culture.
The AHIH model of onboarding clinicians and therapists from diverse backgrounds into the design process is how EmoEx is addressing this, acknowledging there are no shortcuts.
Introducing the Charity Partners — Youth in Transition and The Runway Foundation
Tina Jones MNZM, Director / Counsellor / Co-Founder, Youth in Transition Charitable Trust
Tina Jones opened by acknowledging the reality behind the work: New Zealand's youth suicide rates remain among the highest in the world, and the Hibiscus Coast was identified as a particular hotspot. Youth in Transition was founded in 2014 by Tina and her co-founder Bridget, who had lost her son George to suicide, with a clear mission: "No young person should ever die by not being able to get the quality and timely help they needed before reaching crisis point."
At the heart of the Trust's work is the Journey Back to Awesome programme, a five-step model offering counselling, mentoring, peer support, group wellness coaching and parent support. A retreat gifted by Briscoes in 2022 extended the programme to include animal, garden and art therapy. Nearly a thousand young people have been referred to the Trust from mental health services, GPs, hospitals and police. "We've only lost one young person to suicide. We've completely changed the trajectory on the Hibiscus Coast of young people dying by suicide."
Terese Marr MNZM, Founder and Board Member, The Runway Foundation | Kerry Manders, CEO, The Runway Foundation
Terese Marr MNZM founded The Runway Foundation roughly ten years ago in Pukekohe, initially with a broad community focus before directing almost all resources toward ADHD support. The need, she explained, came directly from her own lived experience as a parent of a child diagnosed with ADHD at age five: "You are diagnosed. And then what? Where is the help?"
The Foundation's ADDventurers programme offers after-school kids clubs, one-on-one youth mentoring through secondary school, parent support groups and community education events. CEO Kerry Manders spoke to the relational core of the work: "We are relationally connected with these kids. People are emailing and ringing daily wanting to join our programme." Both charity partners were beneficiaries of the OpenMind silent auction, with proceeds from the EmoEx user artworks on display in the foyer throughout the day.
Next: The Seminar Recap
The keynotes and panels set out the shape of the debate: governance, regulation, cultural responsiveness, and where the line between AI support and human care should sit. The four practitioner seminars that followed put those questions into practice across neurodiversity, workplace wellbeing, relationships, and parenting.
Read the seminar recap here: The Seminar Sessions.
Comments