Anonymised under NDA
WhatsApp personal assistant with a private memory
Consumer · A consumer personal-assistant product
A personal assistant you talk to on WhatsApp, backed by your own localised database and a retrieval layer over everything you have told it: identity documents, insurance, birthdays, reminders, and anything else worth remembering.
The problem
- Important personal information (Emirates ID details, insurance policies, renewal dates, birthdays) lived across photos, emails, and memory, and was never to hand when needed.
- Reminder apps needed the user to structure everything up front; nobody does that for the long tail of things worth remembering.
- Sending personal documents to a hosted assistant with no control over where they were stored was not acceptable.
The architecture
- WhatsApp as the only interface: users send text, photos, and documents in natural language, and the assistant extracts, confirms, and files what matters.
- A per-user localised database holding structured records (Emirates ID, insurance, birthdays, contacts, notes) alongside the original documents.
- Every stored item is also indexed into a per-user RAG layer, so the user can ask anything about their own data in plain language and get a cited answer.
- Scheduled agents run reminders, birthday prompts, insurance and document expiry warnings, and a daily report; a web-search tool lets the assistant answer questions about the outside world too.
WhatsApp Business APIClaudePostgreSQLpgvectorTemporalWeb search APIs
The outcome
- Users store things by simply telling the assistant, then retrieve them the same way, without ever opening a form.
- Expiries and renewals stopped being surprises: the assistant raises them ahead of time, on the channel the user already lives in.
Services drawn on
A similar problem on your roadmap?
A Discovery Sprint is the fastest path to a scoped, costed proposal.