The invisible work of journalism: transcription
A reporter has a 45-minute interview with a key source. They record everything. They leave the meeting with great material and a 6-hour editorial deadline. But first they have to transcribe.
Three hours later, they've transcribed half the interview. Three hours remain to write the full piece. The result: a rushed article that doesn't do justice to the interview, and a source whose most interesting quotes remained unused in an audio file.
This scenario repeats thousands of times daily in newsrooms worldwide. And it has a solution.
Before and after: workflows with AI transcription
Traditional workflow
- 45-minute interview → audio recording
- 2-4 hours manual transcription
- 15-20 minutes searching for the exact quote you want
- Writing the piece with whatever time remains
- Total: 3-5 hours of pre-writing work
Workflow with automatic transcription
- 45-minute interview → audio recording
- 3 minutes of AI transcription
- 10 minutes reading the text, flagging quotes to use
- Writing the piece with all quotes at hand
- Total: 15-20 minutes of pre-writing work
Quote accuracy: the standard that matters most
The most legitimate concern journalists have about AI transcription is accuracy. And it's valid: a misquoted source can create legal problems, damage source relationships and undermine a publication's credibility.
The practical reality is that current AI models achieve 95-97% accuracy on clean audio. The review process is far faster than manual transcription. Instead of transcribing everything, the journalist only needs to verify the 5-10 quotes they're going to publish, cross-referencing against the original audio at the exact timestamps AI provides.
Best practice: for quotes that will be published with quotation marks, always verify the verbatim quote against the original audio using the timestamp. AI takes you directly to the exact moment — the verification process takes seconds, not minutes. This meets the verification standard of any responsible publication.
Data journalism: the benefit few anticipate
When you have 20 interviews transcribed as text, you can do something that wasn't previously possible: search for patterns across all interviews simultaneously.
How many sources mentioned the word "bureaucracy"? What language does each spokesperson use when discussing the same topic? In which interviews did the name of the same intermediary appear?
With audio files, this investigation would take weeks of listening. With text transcripts, it's a 30-second search. Automatic transcription doesn't just save time — it opens analytical possibilities that simply didn't exist before.
Source archive management
Every journalist has their sources. And every conversation contains information that might be relevant weeks or months later. Without transcription, that context gets lost in an unlabeled audio archive.
With indexed transcripts, you can search your entire history of conversations with a source, find a specific statement someone made 8 months ago about a topic that's back in the news, and verify whether what they're saying now is consistent with what they said then.