AFA AI Assistant Powered Helpline.
A custom RAG system gives caregivers 24/7 access to helpline guidance and memory screening resources, in text and speech, in the compassionate tone AFA is known for.
The challenge.
The Alzheimer’s Foundation of America needed to provide caregivers with 24/7 access to helpline guidance and memory screening resources, without requiring full-time staff availability. Caregiving questions do not keep business hours.
The build.
We designed the conversational system prototype, then refined functionality and assets into a production assistant. The core is a custom Retrieval Augmented Generation system with vetted resources, policy memory, tone control, and safety review tools. The system provides instant text and speech responses while maintaining the compassionate, informed tone AFA is known for.
The assistant runs live at alzfdn.org/virtual-helpline-assistant. The avatar and character work behind it is documented in its own field note: Virtual AI Assistant for Alzheimer’s Foundation of America.
Stack.
Python · FastAPI · OpenAI API · ElevenLabs API · PostgreSQL · RAG retrieval · Prompt chains · Vanilla JS · D-ID API · Stable Diffusion models
Results.
- 24/7 availability for caregivers
- Instant answers cut friction for first-time users
- Fewer repetitive questions reaching helpline staff
- Caregivers reach the right resources quicker
- Months of vendor development saved
“We’re very proud of this project, and I know that it’s something we intend to showcase for a very long time as something that is both a great resource, and also something that is a show of innovation of what AFA can do for people nationally.”
Lonnie Ostrow, marketing director, Alzheimer’s Foundation of America
OUTCOME: 24/7 caregiver access with instant text and speech answers, fewer repetitive questions for staff, and months of vendor development saved.
Ready before
the agents arrive.
Tell us what you want to build, we respond within one business day.