- Client
- Avisi
- Documented Outcome
- AI Agent
- Key Metric
- MVP expanded 4× — same timeline, zero delay
The Situation
Avisi wanted an internal AI assistant that could answer business questions from their own data — calendar, email, documents, CRM — without sending sensitive information to external servers. MVP scope defined for 4 weeks.
What We Built
Built LEO: a fully chat-based AI agent running on Avisi's own NVIDIA GPU server. Integrations: Google Calendar, Gmail, Drive, HubSpot CRM. Extended during development with weather, location awareness, agenda-based advice, and document Q&A — well beyond the original MVP scope.
Result
Delivered within the 4-week deadline and far beyond the MVP definition. LEO provides active advice based on calendar, documents, and CRM context — not data display, but real decision support.
Key Metric
MVP expanded 4× — same timeline, zero delay