How To Extract Earnings Call Qa Using Quotes Only From Paste
Transcript extractors fail when they invent EPS and a ticker. The paste had two answers. Quote those spans. Not investment advice. Not an 8-K.
The matching generator is the Earnings-Call Transcript Q&A Extractor (Quotes Only from Paste) prompt. Browse related cards in the PromptDig library (Browse more prompts). When a filled run survives, share the version you actually use (Share a prompt).
Banner quotes only, not advice
Extract Q&A pairs from a pasted earnings-call transcript. Quotes only. No invented guidance, tickers, or speakers. Start by filling Inputs, not by asking the model to remember last week's run. If a field is blank, write NONE or NOT IN INPUTS and leave it blank through Generate. The card is built so the model cannot honestly invent a number, owner, URL, or command that you did not paste.
Paste these fields before you hit run:
Issuer and event I may print (as pasted): [Issuer]
Pasted transcript excerpt (Q&A section): [Paste]
Speakers I may label (names as in Paste): [Speakers]
Themes I want to bucket (or NONE): [Themes]
What I must never infer: [Never]
Words I must not use: [Banned]
Max quote characters per answer: [Cap or 400]
How many pairs: [Count or 8]
Advice banner: [Banner]
Ticker I may print (or NONE): [Ticker]
That inventory is the honesty ledger. Anything that does not appear there is forbidden in the draft. If you catch yourself adding a nice-to-have after the run, you are no longer using the card. You are ghostwriting. Put the extra fact in Inputs and run again.
Pairs from the paste, speakers as named
Generate is numbered on purpose. Do not skip a step because the first paragraph looked done. The early steps exist to stop later prose from smuggling claims.
Walk the Generate list in order:
- Banner: quotes only, not investment advice. Quote Banner.
- Honesty ledger: issuer, ticker, every speaker name in Paste, themes. Forbidden: numbers and names not in Paste.
- Pair list: up to Count Q&A pairs. Each answer is a quoted span under Cap. If a speaker is unclear, write NOT IN INPUTS rather than guessing the CFO.
- Theme tags: only Themes. If NONE, skip tags.
- Guidance box: only sentences in Paste that the issuer labeled guidance. Else write no guidance sentence in paste.
- Ticker: only Ticker. If NONE, omit.
- Gaps: questions that were cut off in Paste.
- Compliance pass: quote Banned words, invented EPS, invented speakers. Cut them.
If a step asks for a version lock, quote the version from Inputs in the output. If a step asks for a refuse list, keep the refuse list in the published artifact, not in a sidebar you delete. Reviewers should see what the model was not allowed to do.
No guidance box unless the issuer labeled it
Most failures are the same shape: a missing field gets a confident fill. A conversion rate appears. A Gradle task appears. A flash point appears. A caption appears on a job that asked for slide text only. Your review is to search the draft for numbers, names, and commands, then grep Inputs. No match means cut.
Honor the constraints as hard stops, not vibes:
- Quotes only from Paste. Not an 8-K, not a 10-K, not investment advice.
- Never invent a speaker, ticker, or guidance number.
- Stay under Cap characters per quoted answer.
- If Ticker is NONE, do not guess one.
- No emojis.
When the card says not legal advice, not certification, not an exam dump, or not a caption engine, that sentence belongs at the top of the output. Deleting it to look more finished is how you inherit risk.
If Ticker is NONE, omit it
Finish with the compliance pass the prompt already asks for. Quote the banned-word hits. Cut them. Print character counts when the job has a cap. Print word counts when the job has a budget. List gaps as gaps. Five missing facts are more useful than one smooth paragraph.
Tags on the card (earnings call qa extractor, transcript quotes only, no invented guidance) are a reminder of the job shape, not an invitation to wander into a neighboring cluster. If you need a different surface, open a different PromptDig card rather than stretching this one.
Fill the card, then run
Replace every bracket. Run on ChatGPT, Claude, or Gemini. Read the ledger first, then the artifact. If the model invents a commit, KPI, DOI, PEL, bid, or logo, discard the run. Tighten Inputs. Run again. Share the filled card that survived, not the first draft that sounded done.