Make.com AI Agent Tutorial – Build a No-Code Assistant That Works 24/7

About this video

Kevin Stratvert builds a no-code AI customer service agent in Make for the fictional Kevin Cookie Company that automatically answers customer questions by email. He creates an agent connected to an AI model provider — he selects Grok in the video to run Meta’s Llama 3 models, choosing the Llama 3 70B versatile model as a balance of speed and quality — and writes a system prompt defining the agent as a friendly, helpful agent for questions about flavors, hours, shipping and pricing, with instructions to answer concisely and refer unanswerable questions to support. He then builds tools as Make scenarios: first, a scenario that gets content from a Google Docs FAQ document and returns it as a scenario output named KCC_FAQ set to run on demand, which he adds as an agent-level system tool with a description; second, a send-email scenario with defined inputs for customer email address, subject and content mapped to an email module. Finally he builds a main scenario triggered by Tally form ‘Watch for new responses’ that runs the agent with the customer’s name, question and email, using the FAQ tool globally and the email tool as a scenario-specific tool, plus additional instructions to greet the customer by name and sign off as Kevin Cookie Company customer support. A test with a Saturday-hours question produces a correct emailed reply citing 10am to 6pm from the FAQ.

Key takeaways

• In Make, an AI agent needs a model connection, a name, a chosen model and a system prompt that acts as its job description
• Tools for a Make agent are built as separate scenarios; the FAQ tool retrieves a Google Doc and returns its text as scenario output on demand
• Agent-level tools are always available to the agent, while scenario-specific tools like sending email are added only in scenarios that need them
• The email tool scenario first defines inputs — customer email address, subject and content — then maps those inputs into the email module fields
• The main workflow uses a Tally form trigger to run the agent, feeding it the submitted name, question and email address to generate and send the reply

Episode details: published April 21, 2025 · 18:45 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: ChatGPT Tips & Features You’re Not Using (But Should Be!) · Next: AWS Tutorial for Beginners – Step-by-Step Guide to Cloud Computing →

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