Make.com AI Agent Tutorial — Build Your First Agent in 20 Minutes

About this video

Kevin builds a first AI agent in Make, a visual drag-and-drop platform, for the fictional Kevin Cookie Company. The agent should answer simple customer emails from an FAQ on its own and escalate special requests to Slack for a human. He creates the agent under AI Agents beta, creates a connection using Make’s AI Provider because it works without extra setup, names it KCC Support Agent, picks the large model, and writes a system prompt: draft a friendly reply if the email matches a known FAQ; if it is complex — a bulk order, refund, complaint, or unclear — escalate to Slack with a summary and suggested reply, and never issue refunds or discounts on its own. He adds context by uploading the Kevin Cookie Company FAQ covering store hours, whether cookies contain nuts, delivery, and other common questions, then tests it with ‘what are the store hours?’ Next he builds a scenario triggered by new Outlook emails (Gmail is mentioned as an alternative), set to the inbox, from now on, and on demand. He adds the Make AI Agents module, maps in the message ID, subject, body, and sender address, and adds a second message instructing the agent to format the reply body in HTML. Tools are separate scenarios the agent can call: an email reply tool taking message ID and reply text and using Outlook’s Reply to a Message, and a Slack tool taking customer email, subject, and body and posting to a customer escalations channel. He links both tools back into the agent with descriptions of when to use each. Testing shows a store-hours email gets an automatic reply, while a bulk order request for 1,000 cookies is escalated to Slack. The Make Grid view shows the whole workflow and its connections.

Key takeaways

• An AI agent differs from a chatbot because it connects to tools and can take initiative, not just respond
• Build the agent in Make with a provider, model, and system prompt that defines FAQ handling, escalation cases, and guardrails like never issuing refunds alone
• Give the agent knowledge by uploading context such as an FAQ file, then test it before wiring up automation
• A scenario trigger (new email) feeds mapped fields — message ID, subject, body, sender — into the agent, with formatting instructions
• Tools are separate on-demand scenarios (email reply, Slack post) that must be linked back into the agent with clear use descriptions

Episode details: published September 11, 2025 · 20:57 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are Kevin Stratvert’s own claims, shared for education — they are not typical or guaranteed.

← Previous: AI Presentation Tools: Skywork vs Gamma? I Put Them Head-to-Head (You’ll Be Surprised!) · Next: How to Add Voice Over to a Video in DaVinci Resolve →

All Kevin Stratvert videos · Kevin Stratvert profile · Kevin Stratvert on YouTube