AI for journalists: from an hour-long interview to the structure of a piece
A transcript of an hour-long interview is thirty-five thousand characters of spoken language, an eighth of which will reach the piece. Following the course of the conversation does not work: the speaker said the important thing in the forty-fifth minute. Below is the order of work, the rules for choosing quotes and the signs by which you cut.
A transcript of an hour-long interview is thirty-five thousand characters of spoken language, an eighth of which will reach the piece. Following the course of the conversation does not work: the speaker said the important thing in the forty-fifth minute. Below is the order of work, the rules for choosing quotes and the signs by which you cut.
Forty thousand characters of transcript and no idea where to begin
The interview went well. An hour of conversation, the speaker opened up, there are figures and living stories.
Then you open the transcript — thirty-five to forty thousand characters of spoken language — and realise there is at most an eighth of an article in there, and that eighth still has to be found.
Then comes the part that eats a day: reading it straight through, trying to follow the course of the conversation, writing out whatever appealed. By evening you have a retelling of the conversation in chronological order — the worst possible format, because the speaker said the important thing in the forty-fifth minute.
The mistake is in the approach itself. Chronology has to be thrown out immediately.
Who breaks the interview down
The Analytical journalist from the catalogue.
What it covers: research and verification, data journalism — turning tables and reports into analytical blocks, structuring and writing material, developing headlines and architecture, finding new angles on a subject.
Worth noting separately is the phrasing on headlines: apt and attention-catching but not clickbait. The constraint is written into the role itself.
And it immediately asks three questions: what the subject is, for which audience and in what format, and whether there is already material to analyse.
The order of work
The task came with no transcript attached:
Explain how to assemble the structure of a piece from an hour-long interview. The order of work, what to extract first, how to choose quotes and by what sign to cut.
The first thing it stated was the principle: abandoning the chronology of the conversation in favour of thematic logic. Everything else follows from that.
Step 1. A rapid pass and tagging
On the first reading, do not edit. Only mark up with coloured tags by theme: figures, personal experience, conflict, forecasts.
This is the key point and everyone breaks it. Trying to edit straight away means getting stuck on page two and losing the overall picture.
Step 2. The main thesis
One question to yourself: what of everything said is most valuable or most unexpected? The answer shapes both the intro and the whole structure.
Step 3. Clustering
Move fragments from different parts of the interview into shared thematic blocks. The assistant's phrasing is precise: the speaker may have answered an important question in the fifth minute and completed the answer in the forty-fifth.
That is exactly why chronology does not work: in speech a thought assembles from pieces scattered across the whole conversation.
Step 4. The frame
Arrange the blocks by priority — from the most current and vivid to context and detail.
What to extract first
Three categories, in order of value:
Insights and exclusives — previously unpublished data, non-obvious conclusions, admissions of mistakes. The last of those is the most underrated: an account of one's own failure gives a piece a credibility no expert assessment can.
Arguments and figures — specific facts, metrics, sums, timescales.
Examples and cases — real stories that illustrate abstract judgements.
How to choose quotes
The rule is simple and almost nobody follows it.
Keep as direct speech: emotion, the speaker's own metaphors, pithy assessments, characteristic phrases and contentious statements. Direct speech conveys character and position.
Convert to reported speech: chronologies of events, statistical summaries, introductory explanations, long descriptive processes. Here a verbatim quote hurts — a journalist will restate it faster and more precisely than the speaker.
A piece with everything in quotation marks reads heavily for exactly this reason: half the quotes carry no voice and only add length.
By what signs to cut
Four signs, and together they remove the bulk of the length:
Duplication and repetition. Speech is cyclical — a speaker returns to one thought two or three times. Keep the clearest formulation.
Digressions. An answer that went off into adjacent but non-essential detail is deleted entirely rather than shortened.
Filler and verbal noise. "You know", "so to speak", "actually", false starts, thinking pauses.
Over-complex constructions. A spoken sentence of forty words with subordinate clauses is broken into two or three short phrases with the emphasis preserved.
What to take from this even without models
Tag before editing. The first reading is markup only. Edits at that stage will reliably eat your day.
Clustering instead of chronology. A thought assembles from pieces across the whole conversation.
Admissions of mistakes first. The most valuable type of material and the least used.
Reported speech for facts, direct speech for voice. That rule alone cuts a piece by a quarter and makes it readable.
Where the assistant helps most
Handling volume. The whole transcript is handed to it, and that is the most reliable mode of working: when all the material comes from you, there is nothing to invent.
Clustering. Finding every fragment on one subject across forty thousand characters is mechanical work, and it is where a person tires and misses things.
Variants of headlines and intros — with the constraint on clickbait written into the role.
And the thing a person does not do — finding contradictions inside the interview: where the speaker said one thing and something else forty minutes later. That is either a transcription error or the most interesting question for a follow-up call.
An obligatory caveat
Everything verifiable gets verified. Figures, dates, names, job titles, quotations — against sources rather than the model's memory.
Separately: the assistant must not invent anything that was not in the interview. Phrasing the request as "rely only on the attached text; if the answer is not in it, say the data is insufficient" is obligatory for this profession.
Modes and chains
The answer ends with an offer to assess the work and suggest improvements. There is a focus group mode — how the piece reads through your audience's eyes — plus harsh assessment and refinement against criteria.
Chains: this one assembled the structure and the quotes — from there a copywriter makes versions for social media, and a presentation assistant makes cards for channels. One hour of interview becomes an article, a series of posts and a clip.
How a journalist earns from this
First — speed. A day of work with a transcript compresses to an hour, and the structure comes out better because it is built by priority rather than by the course of the conversation.
Second — as a service. Processing interviews and podcasts for experts and companies: transcription, structure, article, cuts for each platform. From our bot's feed of 5 986 jobs over four months, this is work at the junction of text and content, and it sells as a package rather than by the piece.
Where to start
The "Media" category holds nine assistants: analytical journalist, universal scriptwriter, copywriter, academic editor, TED presentation master, TED editor for messengers, Minto-method business presentations and others. In total the platform holds over 200 AI assistants across 14 categories.
Registration is free and opens three days of full Basic access — all the assistants and six foundational express courses, including "Perplexity as an investigative tool" for fact-checking.
Do one thing today: take your most recent transcript and tag it without editing anything. Twenty minutes. Usually at that step alone it becomes visible that the main thesis was not where you thought.
The second approach: not a chat but a conveyor
Everything above is work in dialogue with an assistant. There is another format, and for volume it suits better.
Cascading prompts in spreadsheets
The course "Earning with ready templates" contains spreadsheets with an integrated model: a large task broken into micro-steps, each living in its own cell and taking the previous result as input. The formula can be dragged down a thousand rows — and a thousand tasks are processed at once. How this works step by step is covered separately: Thirty posts in one launch.
What is available for a journalist:
A copywriting template — from subject to finished text along a chain. A social media template — one piece of material unfolds for every platform.
The course is included in the Basic plan.
When a chain needs images, video, sound or external services
When a chain needs images, video or sound, a spreadsheet is not enough: that is what Cerebrum is for — you describe the task in ordinary words, the right models are picked automatically, and the price is visible before launch. For regular business automation that calls external services, use the n8n and Make templates (Creator and above).
Ready pipelines for content
To repackage a piece into other formats and to keep regular selections on a subject, use the n8n and Make automation templates (Creator and above).
Plus the course "Editing" — eighteen lessons, of which two matter most to a text writer: why 7 clips out of 10 achieve maximum reach and an analysis of the first two seconds. What determines whether you get read through works the same way in text and in video.
Automation templates — Creator, Make, N8N and Full; the "Editing" course — Plus.
Nine assistants in the "Media" category
The catalogue, Basic plan:
- Analytical journalist — subject analysis and working with sources
- Universal copywriter and AI copywriter — producing text
- Academic editor — when the material demands precision of wording
- Universal AI scriptwriter — for video formats
- TED presentation master and TED editor for messengers — the structure of a talk and its adaptation for a channel
- Minto business presentations — the pyramid principle: the main point first, the reasoning after
- Provocative content writer
Source verification as a separate skill. The express course "Perplexity as an investigative tool", Basic plan and above, or $9 one-off. The key point from it: the "search first" approach — long detailed requests work worse than short ones, because they mix search intent with output intent. Plus precise search operators and working with databases rather than the general web.
Skills for verification and text
All six are advanced skills: Creator and above, or $9 one-off, or 1000 XP.
- Fact verification — extracts numbers, dates, names, statistics and quotations from a text and checks each statement separately, with a source and a status: confirmed, discrepancy, unverifiable
- Final proofreading — spelling, punctuation and grammar, but with an important caveat: it checks against a description of your style so it does not correct authorial devices as errors
- Research and writing with citations — structure, sources, refining the first paragraph, section-by-section work. The author's voice is preserved
- Copy editing in seven passes — seven separate passes: clarity, tone, "why does the reader care", evidence, specificity, emotion, polish. Every change is justified
- Removing signs of machine text — strips inflated phrasing and repetitive constructions to encyclopaedia standards. An obligatory step if a model was involved in the work
- Digest curation — with a useful distinction: restating a headline against an editorial view. The formula is what happened, why it matters to the reader, your position
On the first item separately. It does exactly what this article demands of a journalist: it breaks a text into verifiable statements and checks each one. Not "assess the credibility" but a list of facts with a status for each.
What to take on the platform for this task
- Gen AI, the "File → text" mode (section) — the "Whisper · speech transcription" model: audio or video to text, including subtitles with timecodes; $0.15 per run
- Whisper — an open speech-recognition model for transcribing interviews locally; the description is open to all, the repository link from Creator
- Buzz — offline audio and video transcription — a desktop app built on Whisper, no internet or subscription; repository link from Creator
- Castmagic — transcription and content from audio — a Claude Code skill that connects Castmagic to your workflows; files from Basic
- Telegram voice message transcription — an n8n template: voice messages become text automatically; Creator, Make, N8N and Full
What to read next
Transcription and file analysis — the reverse direction of working with AI in full.
How to find trends — what to write about besides what the speaker already said.
Editing expert videos — what else can be got out of the same recording.
How to verify a fact with AI search — the obligatory step before publication.
AI agents: what they are and what they pay — the full guide to agents, with market rates.