Case study: Australian Suicide Prevention Foundation (ASPF)
Challenge
The Australian Suicide Prevention Foundation (ASPF) had a prototype application designed to help the "inner circle" (friends and family) of an at-risk individual. Its primary challenge was that the app's static, pre-scripted messages lacked the nuance and personalisation needed for high-stakes, sensitive conversations. ASPF needed to evolve this into an interactive tool that could safely guide a user to co-author an effective, medically-approved message, requiring sophisticated AI guardrails to prevent harmful hallucination.
Solution
The 4-week discovery sprint resulted in a validated architecture for a secure, authenticated, responsible AI system on AWS.
Secure Authentication & Data
The system is designed to use Amazon Cognito for user authentication and Amazon Aurora to securely store user profile data, all managed by AWS Lambda functions.
Safety-First AI Guardrails
To eliminate the risk of harmful AI-generated responses, the system's guardrails explicitly restrict its purpose. Its sole function is to perform semantic matching - using Amazon OpenSearch Service and Amazon Bedrock - to connect a user's query to the most relevant, pre-approved message from the ASPF content library.
Future-Proof Safety Roadmap
This foundational AWS architecture creates a future-proof roadmap. While the MVP is strictly limited to semantic matching for maximum safety, the platform is ready to incorporate more advanced AWS AI safety tools as they evolve. Future backlog items, like real-time sentiment monitoring, can be securely explored by leveraging AWS's integrated services.
Scalable Serverless Backend
The entire solution, including the user-facing application and admin panel, is built on AWS Lambda, Amazon S3, and Amazon Route 53, creating a scalable, cost-effective, and resilient platform.
Process
Our 4-week sprint was a highly structured engagement designed to move from concept to a concrete implementation plan.
The first two weeks were dedicated to deep-dive discovery. We began with a project kick-off to align all stakeholders on the sprint's objectives. We then conducted a series of intensive workshops with ASPF's leadership team. These workshops were critical for mapping the precise user journeys, defining the functional scope, and, most importantly, co-designing the essential AI safety guardrails with ASPF's clinical subject matter experts.
In the final two weeks, our team synthesised this information. We designed the solution architecture and developed a phased roadmap. We presented these draft documents to the ASPF team for review, incorporating their feedback to finalise the strategic deliverables.
The engagement concluded with a final executive presentation and a formal handover of all strategic documents, securing sign-off for the discovery phase and providing ASPF with a clear business case for the next phase of development.
Impact
The 4-week discovery sprint successfully concluded the project and provided ASPF with a clear, de-risked path forward for implementation for presentation to funding bodies. The engagement delivered four key strategic documents.
- Project Discovery & Scope Document: Detailed user requirements, personas, and the functional scope for the Foundational MVP.
- High-Level Technical Architecture: A diagram and documentation illustrating the proposed authenticated, serverless AWS solution.
- Defined Success Metrics: A finalised list of KPIs to measure project success.
- Phased Implementation Roadmap: A detailed project plan, business case, and fixed-price investment for the Foundational MVP, estimated at 12-14 weeks.