Executive AI Assistant
A voice and chat AI agent that runs real business operations, from inbox and calendar to outbound calls.
The problem
The person running a business is the bottleneck. Email triage, scheduling, meeting prep, and follow-ups eat the day, and the threads that slip are usually the ones that mattered most.
What I built
A multi-agent assistant you talk to, by voice or chat, that acts on your behalf across the tools the business already runs on, Microsoft 365 among them. Several agents cooperate through an orchestration layer I wrote in TypeScript on top of the provider agent SDKs, which holds their state, routes their tool calls, and keeps memory between sessions. The assistant triages and drafts email, runs the calendar and task list, writes a brief before each meeting, and places outbound calls. Voice is always on, over the OpenAI Realtime API and WebRTC. Every action that changes something is confirmed before it fires, and credentials never touch the model.
My approach
The agent reasons in plain language and acts only through typed, validated functions. I measure latency, tool-call success, and cost on every run, and every write waits for a confirmation. That is what makes it safe to point at a real inbox.
The result
The inbox is triaged and the day's briefs are written before the working day starts, and every action the assistant took is in an audit trail.
Stack
Connect
Building something, or thinking about it? Book a 1:1 and let's dig in.
Gbolagade Ishola