Built an AI agent that handles my email triage in Lindy.ai and here is exactly how it works

I run a small consulting practice and my inbox is one of the main things that eats my day. I wanted an AI agent that could read incoming emails, categorize them, draft responses for the routine ones and flag the ones that need my actual attention. Tried building this in a few different tools and Lindy.ai is the one where I actually got it working properly.

The visual flow editor uses trigger and action logic. You define what starts the agent, an incoming email in this case, and then chain the actions that follow. The configuration is done in plain English rather than code or complex rule syntax. I described the decision logic I wanted in natural language and it built the flow around that.

The Human-in-the-Loop feature is what made me comfortable actually deploying it. You can set certain actions, specifically sending emails, to save as draft rather than send automatically. So the agent writes the response, puts it in my drafts, and I review and send. Once I was confident enough in what it was generating I started letting it send certain categories directly.

The Knowledge Base connection is useful for anything beyond simple triage. You can connect it to Google Drive or Dropbox so the agent can reference your actual documents when answering questions rather than working from general knowledge. For client-facing agents that matters a lot.

Model selection lets you choose which underlying AI powers the agent, OpenAI, Anthropic or Google, which is useful if you have preferences or if certain tasks perform better on different models.