Applied AI and integrations
AI sales assistant
A functional prototype that supports sales calls, transcribes them live, and executes post-call work without manual capture.
- Responsibility
- End-to-end architecture, development, and integration
- Period
- 2026
- Client
- Saludsa
The problem
The sales team needed to validate whether an AI layer could support calls, capture relevant information, and trigger follow-up without adding more operational load for the advisor.
The solution
I built a web panel with Chrome audio capture, Deepgram transcription, incremental OpenAI coaching, session closeout, quote generation, HubSpot notes and tasks, and WhatsApp follow-up. A simulation mode made demos repeatable without live audio.
Outcome
The team received a repeatable internal validation workflow, real integrations, a Windows executable, an operating guide, and reviewable post-call deliverables.
What I owned
- Call flow and session-state design
- Audio capture and live transcription
- Incremental coaching and assisted closeout
- HubSpot and WhatsApp integrations
- Quote generation, packaging, and documentation
Deliverables
- Functional web panel
- Lightweight Chrome extension
- Windows executable
- Simulation mode
- Operating guide and delivery package
Technology