No-Code AI Automations
Design a small flow that classifies fictional messages; learn the fee, error, retry, and human review steps together.
Starter task: classify the five fictional messages in the customer message guide as order, return, product information, or unclear. The first trial runs manually/with a test button and produces only draft output. The cost of a Make or n8n account/setup and the AI connection is assessed separately; enabling a paid service is not required to read this content. The linked external courses are not all included in the stated practice time.
Trigger and expected output
Read the M1–M5 messages in the guide. For each record, define the message ID, text, class, rationale, and human review fields. M4 is unclear, and because M5 contains more than one topic, it should be set aside for human review. First prepare the expected table manually.
First test flow with Make or n8n
Before choosing a tool, compare hosting, connections, and terms of use. Then give a single fictional record to a manually triggered flow and review its draft output. Do not enable scheduling, real email, or customer connections. When an error occurs, you should be able to see which step stopped.
AI connection and cost
Determine whether the connection you use is AI access provided by Make/n8n or your own provider account. Do not assume that a subscription replaces API usage. Record the number of requests, credit/token consumption, retries, and test budget. Do not put the key in a shareable table or prompt text.
Try the same message twice and with an error
Submit the M1 ID twice; design an ID check so that only one record remains in the draft target. Also try missing text and a connection error. The error log should preserve the message ID; unlimited retries or automatic real sending must not occur. Record the total number of attempts/failures.
Mini project: controlled flow with five messages
Run the five fictional messages through. M1 should be order, M2 return, M3 product information; M4 and M5 should be unclear/human review. Compare the classes with the expected table. The result package is the flow description, five-row output, error/retry logs, and cost note. Separate approval and access design are required for real customer communication.