The problem: the client received dozens of calls a day. Managers listened back to recordings manually and typed the details into a spreadsheet. Information got lost, reports were always late.
What I built:
A pipeline in n8n. The call recording goes to transcription automatically, then an AI model processes the text and extracts structured fields: topic, key details, next action. The result is written to the database and the team receives a ready summary.
Error handling is a separate workflow: if a step fails it retries, and only then sends an alert with the raw data attached, so no call is ever lost silently.
Result: manual listening removed completely, data on every call appears right after the conversation ends.
Stack: n8n, AssemblyAI, OpenAI API, database, webhooks.
#n8n #automation #AI #integration #API
What I built:
A pipeline in n8n. The call recording goes to transcription automatically, then an AI model processes the text and extracts structured fields: topic, key details, next action. The result is written to the database and the team receives a ready summary.
Error handling is a separate workflow: if a step fails it retries, and only then sends an alert with the raw data attached, so no call is ever lost silently.
Result: manual listening removed completely, data on every call appears right after the conversation ends.
Stack: n8n, AssemblyAI, OpenAI API, database, webhooks.
#n8n #automation #AI #integration #API