Mental health care has a personalisation problem.
Traditional therapy is deeply personalised — a skilled therapist adapts their approach to each client's needs, culture, personality, and progress. But traditional therapy doesn't scale. There aren't enough therapists, sessions are expensive, and access is wildly uneven across geographies and demographics.
Digital mental health tools solved the scale problem but created a new one: most digital interventions are generic. The same CBT module for everyone. The same meditation library. The same chatbot script regardless of whether you're a 22-year-old in Jakarta dealing with work anxiety or a 45-year-old in Sydney processing grief.
AI is changing this equation. And the implications are enormous.
The Three Layers of Personalisation
At Intellect, we think about AI-driven personalisation in three layers:
Layer 1: Content matching. This is the simplest form — using data about a user's presenting concerns, goals, and engagement history to surface the right content at the right time. Instead of a generic library, users see a curated pathway that adapts as they progress.
Layer 2: Interaction adaptation. This is where AI gets more sophisticated. The way a self-guided exercise is framed, the tone of a chatbot interaction, the pacing of a learning module — all of these can be adapted based on user behaviour and preferences. A user who responds well to direct, practical advice gets a different experience than one who needs more reflective space.
Layer 3: Clinical augmentation. This is the frontier. AI that helps human clinicians personalise their care — by providing session prep summaries, tracking progress patterns, suggesting treatment plan adjustments, and flagging risk indicators that might otherwise go unnoticed.
What Makes Mental Health Personalisation Different
Personalising mental health isn't like personalising a Netflix recommendation. The stakes are fundamentally different:
Wrong recommendation in entertainment = mild annoyance. Wrong recommendation in mental health = potential harm.
This means mental health personalisation requires:
Safety guardrails. Every personalisation decision must pass through a safety layer. AI should never recommend that someone with suicidal ideation try a mindfulness exercise instead of connecting with a crisis professional.
Cultural sensitivity. Mental health is profoundly cultural. The way grief is expressed, the role of family in treatment, attitudes toward medication, stigma around seeking help — all of these vary dramatically across cultures. Personalisation that ignores culture isn't personalisation. It's projection.
Clinical grounding. Personalisation should be informed by evidence-based clinical frameworks, not just behavioural data. The goal isn't to give users what they want — it's to guide them toward what helps.
Real Examples
Here's what AI-powered personalisation looks like in practice at Intellect:
A user in Singapore logs in after a stressful day. Based on their history — they respond well to structured exercises and tend to engage in the evening — the app surfaces a 10-minute guided journaling session focused on work stress, with a check-in prompt adapted to their communication style.
A therapist preparing for a session receives an AI-generated summary: the client's mood trend over the past two weeks, key themes from their self-guided exercises, and a flag that their sleep patterns have changed. The therapist walks into the session better prepared, more attuned.
An HR leader receives a population-level insight: engagement with anxiety-related content has increased 40% in the engineering team this quarter. No individual data is shared — but the signal enables proactive organisational support.
The Ethical Tightrope
Personalisation in mental health walks an ethical tightrope. We have to balance:
Helpfulness vs. surveillance. Users need to feel supported, not watched. Personalisation vs. filter bubbles. We don't want to trap users in echo chambers of their own distress. Data utility vs. privacy. Better personalisation requires more data. But mental health data is the most sensitive data imaginable.
Our approach: radical transparency. Users know what data we collect, how we use it, and can control their personalisation preferences. We never sell data. We never share individual data with employers. And we build our AI with clinical oversight, not just engineering optimism.
What's Next
The future of AI in mental health personalisation isn't just better algorithms. It's better systems — where AI, human clinicians, self-guided tools, and organisational insights work together as a coherent care ecosystem.
We're early. But the trajectory is clear: mental health support that truly meets each person where they are — culturally, emotionally, clinically — at a scale that human systems alone could never achieve.
That's what we're building. And I've never been more excited about what's possible.
Originally published on Medium.