Mental health charities support people from all walks of life, everyday, on their journey to better health and wellbeing. Our workforces are in person, over the phone, online, supporting people when they need it most. So where does AI come in? 

At our recent Members’ meeting, we discussed how we can make the most of AI as mental health charities, without compromising on the quality of our services, the safety of the people we support, or the human connection which makes mental health support work. 

For the first blog in our series on AI in mental health charities, we caught up with Lee Hall, Technology & Digital Director from our member Everyturn Mental Health, to find out what they’ve learnt so far about AI usage within their charity. 

Everyturn is a national charity providing specialist services on behalf of the NHS, including talking therapies, crisis support, dementia care, services for children and young people, specialist nursing and hospital step-down, and community wellbeing support.

How are you using AI at your mental health charity and what have you learnt? 

Lee: At Everyturn Mental Health, we’re currently using two main AI platforms: Limbic and Microsoft Copilot. Limbic supports our referral routes, while Copilot is being used across the organisation to help with productivity, insight and reducing administrative burden.  

A key factor for us with Copilot is the availability of Enterprise Data Protection (EDP). Given the nature of our work and the sensitivity of the data we hold, it’s critical that any AI tooling we use operates within a secure, enterprise‑grade environment where our data is not used to train public models or leave our control. That has been a non‑negotiable for us and I would strongly recommend any provider considering adopting an AI platform ensures this is the case.  

We’ve also ensured that AI adoption can’t be informal or ad hoc. As a result, we’ve recently developed an organisational AI Policy. This is deliberately pro‑AI, recognising the genuine opportunities it presents, but it’s equally clear about the need for safe, ethical use, strong governance, and clear approval processes. We’re very controlled in terms of which products are permitted, how they’re assessed, and how they’re monitored over time. We actively encourage use and innovation in this space, but via our approved routes and our within our professional guardrails. 

How have you made sure that your use of AI tools is safe and ethical? 

Lee: For us, safety and ethics start with governance. Our AI Policy sets clear expectations around acceptable use, data handling, risk assessment and accountability, and we’ve been explicit that AI tools must be approved before they’re used with organisational data. Governance isn’t the thing that slows AI down; it’s the thing that makes it usable at scale. 

On a personal level, my own thinking has evolved over the past few years, with a particular change in direction over the past 3-6 months specifically. I’ve recently become much more cautious, particularly as awareness has grown around the political interests, commercial incentives and broader power dynamics of some of the largest AI providers, many of which are US‑based. The evolution of social media offers a cautionary lesson here, particularly in how powerful technologies can be leveraged for influence and manipulation. Fundamentally, AI is no different here, it doesn’t arrive value-free – it brings the incentives, power structures and blind spots of the organisations that built it. 

That doesn’t mean we’re anti‑AI, but it does mean we’re deliberate. We’re taking a careful approach to ensure that the data entrusted to us is not used in ways that could be exploitative, misaligned with our values, or effectively “weaponised” through opaque models or secondary uses we don’t control. In mental health charities especially, trust is fundamental, and that has to extend to the technologies we choose to adopt. 

Ultimately, we see AI as a tool to support people and staff, not something that replaces judgement, relationship or responsibility. Being clear about that boundary has been essential. 

What do you find promising about the future of AI in your work as a mental health charity? 

Lee: What excites me most about AI is its ability to turn complex datasets into practical insight. This isn’t about chasing novelty, but about strengthening the everyday work that organisations rely on – things like data analysis, reporting, planning, assurance, and sense-checking information across complex systems. For me, AI becomes genuinely useful when it stops trying to be overly clever and starts being dependable. 

I’m also so intrigued as to how we can start using AI to extend access to support outside traditional working hours, providing people with timely information, guidance, and reassurance when services are closed or capacity is limited. It also offers opportunities to personalise interventions, adapting content and approaches to individual needs, preferences, and learning styles. Used thoughtfully, this can support more inclusive and equitable access by helping organisations develop approaches that better reflect the diversity of how people engage with and process support. The real value lies in using AI to broaden reach and responsiveness, while ensuring that human judgement, compassion, and accountability remain central to care. 

What are your biggest concerns going forward around the use of AI within mental health charity service provision? 

Lee: Beyond the very real concerns about sensitive data being misused or repurposed, one of my biggest worries in a mental health context is the growing narrative around full autonomy – the idea that AI systems can be left to operate unmanaged or unchecked. We’re seeing rapid developments in agentic AI, alongside claims that these systems can quietly run in the background of organisations, taking on everyday tasks with little or no human involvement. In practice, that can range from managing diaries and correspondence, through to drafting responses, prioritising work, or making recommendations based on large volumes of information – all without direct human oversight. 

In a mental health organisation, that trajectory should give us pause. When AI systems begin to act autonomously around information that may include clinical notes, risk assessments, safeguarding concerns, or communications with people in distress, the stakes are fundamentally different. My concern is that, so far, evangelists of AI have tended to overestimate the maturity of these technologies and underestimate their risks, driven more by hype and the promise of efficiency than by evidence or lived experience. Allowing AI to act independently without robust governance, clear boundaries, and continuous human oversight introduces significant danger. 

In real terms, that risk can manifest as inappropriate responses to sensitive communications, subtle but harmful errors in how information is interpreted or prioritised, unintended data exposure, or decisions being made at speed without the contextual judgement that mental health work depends on. These risks are not theoretical – they arise precisely because accountability is being removed from systems that are still fundamentally probabilistic, opaque, and incapable of understanding human distress, complexity, or nuance in the way practitioners do. 

I believe given the AI-hype, we’re seeing many organisations drive towards achieving AI offered efficiencies quickly and at scale, without the necessary guardrails in place. In mental health services, that combination can be particularly dangerous whereby small errors, misclassifications, or poorly governed automations can rapidly become systemic, affecting large numbers of people and undermining trust, safety, and care quality. That is what many currently underestimate. Autonomy without accountability is, in my view, the point at which AI stops being a supportive tool and starts becoming actively harmful. 

Stay tuned for more blogs in this series, discussing the future of AI in mental health charities. 

To join the discussion, become a member of the Association of Mental Health Providers today.