AI Agents That Screen Clients and Reply for You

About this video

Nick, a freelance training provider who manages projects in Trello and receives client requests by email, builds two linked Make scenarios using Make AI Agents to screen and reply to clients without coding. In the first scenario, a Gmail trigger watches the inbox for new messages. An AI agent, configured with Make’s built-in provider and GPT-5 medium, receives the thread ID, sender address, and full body text as inputs and is instructed to identify only new messages that are not part of an existing thread and that request training services, ignoring unrelated email. The agent is given a Trello tool to create a card in the New Leads list of his freelance projects board, with the card name set to let the AI decide so it can insert the sender name, any company name it identifies, and the thread ID, plus a description. In a test, a training request from Ava produced one card while an unrelated message from Liam was ignored. Once working, the scenario is set to run every 15 minutes. A second pre-built scenario triggers when a card is moved to a specific Trello list; its agent reads the stored thread ID and uses a Gmail tool to reply inviting the client to schedule a meeting. Make’s grid view is shown for visualizing connections across scenarios and checking layers for data volume and AI credit use.

Key takeaways

• Make scenarios always start with a trigger, and running the scenario with only the trigger in place shows exactly which email fields can be passed to later steps.
• The AI agent’s instructions work like a chat prompt, but any fields the agent must actually analyze — such as thread ID, email address, and full body text — must also be added to its input field.
• Agents act in other apps through tools added from the agent’s bottom plus button, for example a Trello Create a Card tool, with fields set to let the AI agent decide when the agent should fill them from its instructions.
• Storing the original email thread ID on the Trello card enables a second scenario to find and reply to the correct thread when the card is moved.
• The creator recommends inspecting the summary above each step after a test run to diagnose problems, adjust the configuration, and retest rather than expecting a scenario to work perfectly the first time.

Episode details: published April 13, 2026 · 12:25 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: KaneAI Tutorial - Test Your Entire App in Plain English · Next: Intuit Intelligence in QuickBooks Online (2026) →

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