💡 In this tutorial, you'll:
- Create an agent to act as your “second brain” when responding to communication — it will know about recent messages you've received and draft responses for you.
- Import data from the apps you use for work.
- Use AI Columns to make your data more useful.
Creating Your First Agent
Start by creating an agent, then use the ... icon in the upper right corner to open the configuration panel.
AI agents thrive when they know who they're helping, and how that person prefers to be communicated with. Use the custom instructions to provide context about your job — for example your name, title, responsibilities, product, and preferred writing style.
Test your agent with a quick question like “Who am I?” to confirm it has picked up your context, and keep tweaking the instructions until the answers feel right.
Attaching Data to Your Agent
Import data from two kinds of sources:
- A communication tool — like email or Slack.
- A knowledge store — Salesforce, HubSpot, Notion, or simpler tools like Google Sheets or PDFs.
Then attach the resulting datasets to the agent. See Create a Dataset for a step-by-step guide.
Improving Your Datasets
- Delete unneeded data to speed up performance and improve answers.
- Use AI Columns to enrich the data — for example, a column like
isDirectQuestionthat flags which messages need a reply.