Raw meeting transcripts can waste RAG agent tokens

A pipeline is taking in many recorded meetings and voice calls. Whisper turns the speech into text, but the result is a huge, unstructured block of words.

The transcript includes filler sounds such as “um” and “uh,” along with empty pauses. When that text is split and stored in a , the RAG agent spends many on low-value text.

Without speaker labels or clear document structure, the agent can lose the thread of the conversation or . The practical issue is whether audio should be cleaned and structured before they are used, instead of paying OpenAI or Anthropic to process messy input.

Key points

  • Whisper can transcribe audio, but the raw output may still be hard for an agent to use.
  • Unstructured can waste inside a RAG workflow.
  • Missing speaker labels make it harder to follow who said what.
  • Cleaning before storage can reduce model cost and errors.
  • Recorded meetings need an ingestion step that removes noise and adds structure.
Read original