Build an AI Executive Assistant With n8n: Email, Calendar and Company Knowledge

Josh builds an AI assistant in n8n that takes a Telegram text or voice note, decides what needs doing and hands the work to specialist sub agents for email and calendar. It also answers questions from your own documents, which are added to a vector database automatically when you drop them into a Google Drive folder.

Key takeaways

• One orchestrator, several specialists. The main agent chooses between an email agent, a calendar agent, a document search tool and simple tools like a calculator. Each sub agent has its own prompt and its own small set of actions.
• Voice works out of the box. Voice notes are transcribed with OpenAI's Whisper before they reach the agent, so you can give instructions on the go.
• Short memory makes it conversational. A buffer of recent messages lets it ask for a missing email address and then finish the task once you reply.
• Your documents become its knowledge. New or updated files in a Drive folder are split into chunks and stored in Pinecone. In the demo, Josh uploads a sales letter and the agent breaks down that offer from the document.
• The master prompt does the routing. The system message spells out each tool's purpose, when to use it and the rules to follow, such as asking for a recipient before sending.
• It is modular by design. You can swap the language model, the chat channel or the vector store, and add new sub agents, for example one that looks up contacts in your CRM.

Apply it to your business

Start by listing the admin tasks you repeat every day, such as scheduling, quick email replies and answering the same questions about your offer. An assistant like this can take those off your plate, and the same pattern of one orchestrator plus specialist agents is how customer facing agents like AI setters are built. Load it with your offer documents and FAQs first, since the quality of its answers depends on what it can search.